net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/dst_metadata.h>
  58 #include <net/xfrm.h>
  59 #include <net/netevent.h>
  60 #include <net/netlink.h>
  61 #include <net/nexthop.h>
  62 #include <net/lwtunnel.h>
  63 #include <net/ip_tunnels.h>
  64 #include <net/l3mdev.h>
  65
  66 #include <asm/uaccess.h>
  67
  68 #ifdef CONFIG_SYSCTL
  69 #include <linux/sysctl.h>
  70 #endif
  71
  72 enum rt6_nud_state {
  73         RT6_NUD_FAIL_HARD = -3,
  74         RT6_NUD_FAIL_PROBE = -2,
  75         RT6_NUD_FAIL_DO_RR = -1,
  76         RT6_NUD_SUCCEED = 1
  77 };
  78
  79 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
  80 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  81 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  82 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  83 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  84 static void             ip6_dst_destroy(struct dst_entry *);
  85 static void             ip6_dst_ifdown(struct dst_entry *,
  86                                        struct net_device *dev, int how);
  87 static int               ip6_dst_gc(struct dst_ops *ops);
  88
  89 static int              ip6_pkt_discard(struct sk_buff *skb);
  90 static int              ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  91 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  92 static int              ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  93 static void             ip6_link_failure(struct sk_buff *skb);
  94 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  95                                            struct sk_buff *skb, u32 mtu);
  96 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  97                                         struct sk_buff *skb);
  98 static void             rt6_dst_from_metrics_check(struct rt6_info *rt);
  99 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
 100
 101 #ifdef CONFIG_IPV6_ROUTE_INFO
 102 static struct rt6_info *rt6_add_route_info(struct net *net,
 103                                            const struct in6_addr *prefix, int prefixlen,
 104                                            const struct in6_addr *gwaddr, int ifindex,
 105                                            unsigned int pref);
 106 static struct rt6_info *rt6_get_route_info(struct net *net,
 107                                            const struct in6_addr *prefix, int prefixlen,
 108                                            const struct in6_addr *gwaddr, int ifindex);
 109 #endif
 110
 111 struct uncached_list {
 112         spinlock_t              lock;
 113         struct list_head        head;
 114 };
 115
 116 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 117
 118 static void rt6_uncached_list_add(struct rt6_info *rt)
 119 {
 120         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 121
 122         rt->dst.flags |= DST_NOCACHE;
 123         rt->rt6i_uncached_list = ul;
 124
 125         spin_lock_bh(&ul->lock);
 126         list_add_tail(&rt->rt6i_uncached, &ul->head);
 127         spin_unlock_bh(&ul->lock);
 128 }
 129
 130 static void rt6_uncached_list_del(struct rt6_info *rt)
 131 {
 132         if (!list_empty(&rt->rt6i_uncached)) {
 133                 struct uncached_list *ul = rt->rt6i_uncached_list;
 134
 135                 spin_lock_bh(&ul->lock);
 136                 list_del(&rt->rt6i_uncached);
 137                 spin_unlock_bh(&ul->lock);
 138         }
 139 }
 140
 141 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
 142 {
 143         struct net_device *loopback_dev = net->loopback_dev;
 144         int cpu;
 145
 146         if (dev == loopback_dev)
 147                 return;
 148
 149         for_each_possible_cpu(cpu) {
 150                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 151                 struct rt6_info *rt;
 152
 153                 spin_lock_bh(&ul->lock);
 154                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 155                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 156                         struct net_device *rt_dev = rt->dst.dev;
 157
 158                         if (rt_idev->dev == dev) {
 159                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 160                                 in6_dev_put(rt_idev);
 161                         }
 162
 163                         if (rt_dev == dev) {
 164                                 rt->dst.dev = loopback_dev;
 165                                 dev_hold(rt->dst.dev);
 166                                 dev_put(rt_dev);
 167                         }
 168                 }
 169                 spin_unlock_bh(&ul->lock);
 170         }
 171 }
 172
 173 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 174 {
 175         return dst_metrics_write_ptr(rt->dst.from);
 176 }
 177
 178 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 179 {
 180         struct rt6_info *rt = (struct rt6_info *)dst;
 181
 182         if (rt->rt6i_flags & RTF_PCPU)
 183                 return rt6_pcpu_cow_metrics(rt);
 184         else if (rt->rt6i_flags & RTF_CACHE)
 185                 return NULL;
 186         else
 187                 return dst_cow_metrics_generic(dst, old);
 188 }
 189
 190 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 191                                              struct sk_buff *skb,
 192                                              const void *daddr)
 193 {
 194         struct in6_addr *p = &rt->rt6i_gateway;
 195
 196         if (!ipv6_addr_any(p))
 197                 return (const void *) p;
 198         else if (skb)
 199                 return &ipv6_hdr(skb)->daddr;
 200         return daddr;
 201 }
 202
 203 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 204                                           struct sk_buff *skb,
 205                                           const void *daddr)
 206 {
 207         struct rt6_info *rt = (struct rt6_info *) dst;
 208         struct neighbour *n;
 209
 210         daddr = choose_neigh_daddr(rt, skb, daddr);
 211         n = __ipv6_neigh_lookup(dst->dev, daddr);
 212         if (n)
 213                 return n;
 214         return neigh_create(&nd_tbl, daddr, dst->dev);
 215 }
 216
 217 static struct dst_ops ip6_dst_ops_template = {
 218         .family                 =       AF_INET6,
 219         .gc                     =       ip6_dst_gc,
 220         .gc_thresh              =       1024,
 221         .check                  =       ip6_dst_check,
 222         .default_advmss         =       ip6_default_advmss,
 223         .mtu                    =       ip6_mtu,
 224         .cow_metrics            =       ipv6_cow_metrics,
 225         .destroy                =       ip6_dst_destroy,
 226         .ifdown                 =       ip6_dst_ifdown,
 227         .negative_advice        =       ip6_negative_advice,
 228         .link_failure           =       ip6_link_failure,
 229         .update_pmtu            =       ip6_rt_update_pmtu,
 230         .redirect               =       rt6_do_redirect,
 231         .local_out              =       __ip6_local_out,
 232         .neigh_lookup           =       ip6_neigh_lookup,
 233 };
 234
 235 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 236 {
 237         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 238
 239         return mtu ? : dst->dev->mtu;
 240 }
 241
 242 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 243                                          struct sk_buff *skb, u32 mtu)
 244 {
 245 }
 246
 247 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 248                                       struct sk_buff *skb)
 249 {
 250 }
 251
 252 static struct dst_ops ip6_dst_blackhole_ops = {
 253         .family                 =       AF_INET6,
 254         .destroy                =       ip6_dst_destroy,
 255         .check                  =       ip6_dst_check,
 256         .mtu                    =       ip6_blackhole_mtu,
 257         .default_advmss         =       ip6_default_advmss,
 258         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 259         .redirect               =       ip6_rt_blackhole_redirect,
 260         .cow_metrics            =       dst_cow_metrics_generic,
 261         .neigh_lookup           =       ip6_neigh_lookup,
 262 };
 263
 264 static const u32 ip6_template_metrics[RTAX_MAX] = {
 265         [RTAX_HOPLIMIT - 1] = 0,
 266 };
 267
 268 static const struct rt6_info ip6_null_entry_template = {
 269         .dst = {
 270                 .__refcnt       = ATOMIC_INIT(1),
 271                 .__use          = 1,
 272                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 273                 .error          = -ENETUNREACH,
 274                 .input          = ip6_pkt_discard,
 275                 .output         = ip6_pkt_discard_out,
 276         },
 277         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 278         .rt6i_protocol  = RTPROT_KERNEL,
 279         .rt6i_metric    = ~(u32) 0,
 280         .rt6i_ref       = ATOMIC_INIT(1),
 281 };
 282
 283 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 284
 285 static const struct rt6_info ip6_prohibit_entry_template = {
 286         .dst = {
 287                 .__refcnt       = ATOMIC_INIT(1),
 288                 .__use          = 1,
 289                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 290                 .error          = -EACCES,
 291                 .input          = ip6_pkt_prohibit,
 292                 .output         = ip6_pkt_prohibit_out,
 293         },
 294         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 295         .rt6i_protocol  = RTPROT_KERNEL,
 296         .rt6i_metric    = ~(u32) 0,
 297         .rt6i_ref       = ATOMIC_INIT(1),
 298 };
 299
 300 static const struct rt6_info ip6_blk_hole_entry_template = {
 301         .dst = {
 302                 .__refcnt       = ATOMIC_INIT(1),
 303                 .__use          = 1,
 304                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 305                 .error          = -EINVAL,
 306                 .input          = dst_discard,
 307                 .output         = dst_discard_out,
 308         },
 309         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 310         .rt6i_protocol  = RTPROT_KERNEL,
 311         .rt6i_metric    = ~(u32) 0,
 312         .rt6i_ref       = ATOMIC_INIT(1),
 313 };
 314
 315 #endif
 316
 317 static void rt6_info_init(struct rt6_info *rt)
 318 {
 319         struct dst_entry *dst = &rt->dst;
 320
 321         memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 322         INIT_LIST_HEAD(&rt->rt6i_siblings);
 323         INIT_LIST_HEAD(&rt->rt6i_uncached);
 324 }
 325
 326 /* allocate dst with ip6_dst_ops */
 327 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 328                                         struct net_device *dev,
 329                                         int flags)
 330 {
 331         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 332                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 333
 334         if (rt)
 335                 rt6_info_init(rt);
 336
 337         return rt;
 338 }
 339
 340 static struct rt6_info *ip6_dst_alloc(struct net *net,
 341                                       struct net_device *dev,
 342                                       int flags)
 343 {
 344         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 345
 346         if (rt) {
 347                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 348                 if (rt->rt6i_pcpu) {
 349                         int cpu;
 350
 351                         for_each_possible_cpu(cpu) {
 352                                 struct rt6_info **p;
 353
 354                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 355                                 /* no one shares rt */
 356                                 *p =  NULL;
 357                         }
 358                 } else {
 359                         dst_destroy((struct dst_entry *)rt);
 360                         return NULL;
 361                 }
 362         }
 363
 364         return rt;
 365 }
 366
 367 static void ip6_dst_destroy(struct dst_entry *dst)
 368 {
 369         struct rt6_info *rt = (struct rt6_info *)dst;
 370         struct dst_entry *from = dst->from;
 371         struct inet6_dev *idev;
 372
 373         dst_destroy_metrics_generic(dst);
 374         free_percpu(rt->rt6i_pcpu);
 375         rt6_uncached_list_del(rt);
 376
 377         idev = rt->rt6i_idev;
 378         if (idev) {
 379                 rt->rt6i_idev = NULL;
 380                 in6_dev_put(idev);
 381         }
 382
 383         dst->from = NULL;
 384         dst_release(from);
 385 }
 386
 387 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 388                            int how)
 389 {
 390         struct rt6_info *rt = (struct rt6_info *)dst;
 391         struct inet6_dev *idev = rt->rt6i_idev;
 392         struct net_device *loopback_dev =
 393                 dev_net(dev)->loopback_dev;
 394
 395         if (dev != loopback_dev) {
 396                 if (idev && idev->dev == dev) {
 397                         struct inet6_dev *loopback_idev =
 398                                 in6_dev_get(loopback_dev);
 399                         if (loopback_idev) {
 400                                 rt->rt6i_idev = loopback_idev;
 401                                 in6_dev_put(idev);
 402                         }
 403                 }
 404         }
 405 }
 406
 407 static bool __rt6_check_expired(const struct rt6_info *rt)
 408 {
 409         if (rt->rt6i_flags & RTF_EXPIRES)
 410                 return time_after(jiffies, rt->dst.expires);
 411         else
 412                 return false;
 413 }
 414
 415 static bool rt6_check_expired(const struct rt6_info *rt)
 416 {
 417         if (rt->rt6i_flags & RTF_EXPIRES) {
 418                 if (time_after(jiffies, rt->dst.expires))
 419                         return true;
 420         } else if (rt->dst.from) {
 421                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 422         }
 423         return false;
 424 }
 425
 426 /* Multipath route selection:
 427  *   Hash based function using packet header and flowlabel.
 428  * Adapted from fib_info_hashfn()
 429  */
 430 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 431                                const struct flowi6 *fl6)
 432 {
 433         return get_hash_from_flowi6(fl6) % candidate_count;
 434 }
 435
 436 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 437                                              struct flowi6 *fl6, int oif,
 438                                              int strict)
 439 {
 440         struct rt6_info *sibling, *next_sibling;
 441         int route_choosen;
 442
 443         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 444         /* Don't change the route, if route_choosen == 0
 445          * (siblings does not include ourself)
 446          */
 447         if (route_choosen)
 448                 list_for_each_entry_safe(sibling, next_sibling,
 449                                 &match->rt6i_siblings, rt6i_siblings) {
 450                         route_choosen--;
 451                         if (route_choosen == 0) {
 452                                 if (rt6_score_route(sibling, oif, strict) < 0)
 453                                         break;
 454                                 match = sibling;
 455                                 break;
 456                         }
 457                 }
 458         return match;
 459 }
 460
 461 /*
 462  *      Route lookup. Any table->tb6_lock is implied.
 463  */
 464
 465 static inline struct rt6_info *rt6_device_match(struct net *net,
 466                                                     struct rt6_info *rt,
 467                                                     const struct in6_addr *saddr,
 468                                                     int oif,
 469                                                     int flags)
 470 {
 471         struct rt6_info *local = NULL;
 472         struct rt6_info *sprt;
 473
 474         if (!oif && ipv6_addr_any(saddr))
 475                 goto out;
 476
 477         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 478                 struct net_device *dev = sprt->dst.dev;
 479
 480                 if (oif) {
 481                         if (dev->ifindex == oif)
 482                                 return sprt;
 483                         if (dev->flags & IFF_LOOPBACK) {
 484                                 if (!sprt->rt6i_idev ||
 485                                     sprt->rt6i_idev->dev->ifindex != oif) {
 486                                         if (flags & RT6_LOOKUP_F_IFACE)
 487                                                 continue;
 488                                         if (local &&
 489                                             local->rt6i_idev->dev->ifindex == oif)
 490                                                 continue;
 491                                 }
 492                                 local = sprt;
 493                         }
 494                 } else {
 495                         if (ipv6_chk_addr(net, saddr, dev,
 496                                           flags & RT6_LOOKUP_F_IFACE))
 497                                 return sprt;
 498                 }
 499         }
 500
 501         if (oif) {
 502                 if (local)
 503                         return local;
 504
 505                 if (flags & RT6_LOOKUP_F_IFACE)
 506                         return net->ipv6.ip6_null_entry;
 507         }
 508 out:
 509         return rt;
 510 }
 511
 512 #ifdef CONFIG_IPV6_ROUTER_PREF
 513 struct __rt6_probe_work {
 514         struct work_struct work;
 515         struct in6_addr target;
 516         struct net_device *dev;
 517 };
 518
 519 static void rt6_probe_deferred(struct work_struct *w)
 520 {
 521         struct in6_addr mcaddr;
 522         struct __rt6_probe_work *work =
 523                 container_of(w, struct __rt6_probe_work, work);
 524
 525         addrconf_addr_solict_mult(&work->target, &mcaddr);
 526         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
 527         dev_put(work->dev);
 528         kfree(work);
 529 }
 530
 531 static void rt6_probe(struct rt6_info *rt)
 532 {
 533         struct __rt6_probe_work *work;
 534         struct neighbour *neigh;
 535         /*
 536          * Okay, this does not seem to be appropriate
 537          * for now, however, we need to check if it
 538          * is really so; aka Router Reachability Probing.
 539          *
 540          * Router Reachability Probe MUST be rate-limited
 541          * to no more than one per minute.
 542          */
 543         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 544                 return;
 545         rcu_read_lock_bh();
 546         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 547         if (neigh) {
 548                 if (neigh->nud_state & NUD_VALID)
 549                         goto out;
 550
 551                 work = NULL;
 552                 write_lock(&neigh->lock);
 553                 if (!(neigh->nud_state & NUD_VALID) &&
 554                     time_after(jiffies,
 555                                neigh->updated +
 556                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 557                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 558                         if (work)
 559                                 __neigh_set_probe_once(neigh);
 560                 }
 561                 write_unlock(&neigh->lock);
 562         } else {
 563                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 564         }
 565
 566         if (work) {
 567                 INIT_WORK(&work->work, rt6_probe_deferred);
 568                 work->target = rt->rt6i_gateway;
 569                 dev_hold(rt->dst.dev);
 570                 work->dev = rt->dst.dev;
 571                 schedule_work(&work->work);
 572         }
 573
 574 out:
 575         rcu_read_unlock_bh();
 576 }
 577 #else
 578 static inline void rt6_probe(struct rt6_info *rt)
 579 {
 580 }
 581 #endif
 582
 583 /*
 584  * Default Router Selection (RFC 2461 6.3.6)
 585  */
 586 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 587 {
 588         struct net_device *dev = rt->dst.dev;
 589         if (!oif || dev->ifindex == oif)
 590                 return 2;
 591         if ((dev->flags & IFF_LOOPBACK) &&
 592             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 593                 return 1;
 594         return 0;
 595 }
 596
 597 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 598 {
 599         struct neighbour *neigh;
 600         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 601
 602         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 603             !(rt->rt6i_flags & RTF_GATEWAY))
 604                 return RT6_NUD_SUCCEED;
 605
 606         rcu_read_lock_bh();
 607         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 608         if (neigh) {
 609                 read_lock(&neigh->lock);
 610                 if (neigh->nud_state & NUD_VALID)
 611                         ret = RT6_NUD_SUCCEED;
 612 #ifdef CONFIG_IPV6_ROUTER_PREF
 613                 else if (!(neigh->nud_state & NUD_FAILED))
 614                         ret = RT6_NUD_SUCCEED;
 615                 else
 616                         ret = RT6_NUD_FAIL_PROBE;
 617 #endif
 618                 read_unlock(&neigh->lock);
 619         } else {
 620                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 621                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 622         }
 623         rcu_read_unlock_bh();
 624
 625         return ret;
 626 }
 627
 628 static int rt6_score_route(struct rt6_info *rt, int oif,
 629                            int strict)
 630 {
 631         int m;
 632
 633         m = rt6_check_dev(rt, oif);
 634         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 635                 return RT6_NUD_FAIL_HARD;
 636 #ifdef CONFIG_IPV6_ROUTER_PREF
 637         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 638 #endif
 639         if (strict & RT6_LOOKUP_F_REACHABLE) {
 640                 int n = rt6_check_neigh(rt);
 641                 if (n < 0)
 642                         return n;
 643         }
 644         return m;
 645 }
 646
 647 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 648                                    int *mpri, struct rt6_info *match,
 649                                    bool *do_rr)
 650 {
 651         int m;
 652         bool match_do_rr = false;
 653         struct inet6_dev *idev = rt->rt6i_idev;
 654         struct net_device *dev = rt->dst.dev;
 655
 656         if (dev && !netif_carrier_ok(dev) &&
 657             idev->cnf.ignore_routes_with_linkdown)
 658                 goto out;
 659
 660         if (rt6_check_expired(rt))
 661                 goto out;
 662
 663         m = rt6_score_route(rt, oif, strict);
 664         if (m == RT6_NUD_FAIL_DO_RR) {
 665                 match_do_rr = true;
 666                 m = 0; /* lowest valid score */
 667         } else if (m == RT6_NUD_FAIL_HARD) {
 668                 goto out;
 669         }
 670
 671         if (strict & RT6_LOOKUP_F_REACHABLE)
 672                 rt6_probe(rt);
 673
 674         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 675         if (m > *mpri) {
 676                 *do_rr = match_do_rr;
 677                 *mpri = m;
 678                 match = rt;
 679         }
 680 out:
 681         return match;
 682 }
 683
 684 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 685                                      struct rt6_info *rr_head,
 686                                      u32 metric, int oif, int strict,
 687                                      bool *do_rr)
 688 {
 689         struct rt6_info *rt, *match, *cont;
 690         int mpri = -1;
 691
 692         match = NULL;
 693         cont = NULL;
 694         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 695                 if (rt->rt6i_metric != metric) {
 696                         cont = rt;
 697                         break;
 698                 }
 699
 700                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 701         }
 702
 703         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 704                 if (rt->rt6i_metric != metric) {
 705                         cont = rt;
 706                         break;
 707                 }
 708
 709                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 710         }
 711
 712         if (match || !cont)
 713                 return match;
 714
 715         for (rt = cont; rt; rt = rt->dst.rt6_next)
 716                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 717
 718         return match;
 719 }
 720
 721 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 722 {
 723         struct rt6_info *match, *rt0;
 724         struct net *net;
 725         bool do_rr = false;
 726
 727         rt0 = fn->rr_ptr;
 728         if (!rt0)
 729                 fn->rr_ptr = rt0 = fn->leaf;
 730
 731         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 732                              &do_rr);
 733
 734         if (do_rr) {
 735                 struct rt6_info *next = rt0->dst.rt6_next;
 736
 737                 /* no entries matched; do round-robin */
 738                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 739                         next = fn->leaf;
 740
 741                 if (next != rt0)
 742                         fn->rr_ptr = next;
 743         }
 744
 745         net = dev_net(rt0->dst.dev);
 746         return match ? match : net->ipv6.ip6_null_entry;
 747 }
 748
 749 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 750 {
 751         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 752 }
 753
 754 #ifdef CONFIG_IPV6_ROUTE_INFO
 755 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 756                   const struct in6_addr *gwaddr)
 757 {
 758         struct net *net = dev_net(dev);
 759         struct route_info *rinfo = (struct route_info *) opt;
 760         struct in6_addr prefix_buf, *prefix;
 761         unsigned int pref;
 762         unsigned long lifetime;
 763         struct rt6_info *rt;
 764
 765         if (len < sizeof(struct route_info)) {
 766                 return -EINVAL;
 767         }
 768
 769         /* Sanity check for prefix_len and length */
 770         if (rinfo->length > 3) {
 771                 return -EINVAL;
 772         } else if (rinfo->prefix_len > 128) {
 773                 return -EINVAL;
 774         } else if (rinfo->prefix_len > 64) {
 775                 if (rinfo->length < 2) {
 776                         return -EINVAL;
 777                 }
 778         } else if (rinfo->prefix_len > 0) {
 779                 if (rinfo->length < 1) {
 780                         return -EINVAL;
 781                 }
 782         }
 783
 784         pref = rinfo->route_pref;
 785         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 786                 return -EINVAL;
 787
 788         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 789
 790         if (rinfo->length == 3)
 791                 prefix = (struct in6_addr *)rinfo->prefix;
 792         else {
 793                 /* this function is safe */
 794                 ipv6_addr_prefix(&prefix_buf,
 795                                  (struct in6_addr *)rinfo->prefix,
 796                                  rinfo->prefix_len);
 797                 prefix = &prefix_buf;
 798         }
 799
 800         if (rinfo->prefix_len == 0)
 801                 rt = rt6_get_dflt_router(gwaddr, dev);
 802         else
 803                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 804                                         gwaddr, dev->ifindex);
 805
 806         if (rt && !lifetime) {
 807                 ip6_del_rt(rt);
 808                 rt = NULL;
 809         }
 810
 811         if (!rt && lifetime)
 812                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 813                                         pref);
 814         else if (rt)
 815                 rt->rt6i_flags = RTF_ROUTEINFO |
 816                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 817
 818         if (rt) {
 819                 if (!addrconf_finite_timeout(lifetime))
 820                         rt6_clean_expires(rt);
 821                 else
 822                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 823
 824                 ip6_rt_put(rt);
 825         }
 826         return 0;
 827 }
 828 #endif
 829
 830 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 831                                         struct in6_addr *saddr)
 832 {
 833         struct fib6_node *pn;
 834         while (1) {
 835                 if (fn->fn_flags & RTN_TL_ROOT)
 836                         return NULL;
 837                 pn = fn->parent;
 838                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 839                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 840                 else
 841                         fn = pn;
 842                 if (fn->fn_flags & RTN_RTINFO)
 843                         return fn;
 844         }
 845 }
 846
 847 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 848                                              struct fib6_table *table,
 849                                              struct flowi6 *fl6, int flags)
 850 {
 851         struct fib6_node *fn;
 852         struct rt6_info *rt;
 853
 854         read_lock_bh(&table->tb6_lock);
 855         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 856 restart:
 857         rt = fn->leaf;
 858         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 859         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 860                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 861         if (rt == net->ipv6.ip6_null_entry) {
 862                 fn = fib6_backtrack(fn, &fl6->saddr);
 863                 if (fn)
 864                         goto restart;
 865         }
 866         dst_use(&rt->dst, jiffies);
 867         read_unlock_bh(&table->tb6_lock);
 868         return rt;
 869
 870 }
 871
 872 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 873                                     int flags)
 874 {
 875         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 876 }
 877 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 878
 879 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 880                             const struct in6_addr *saddr, int oif, int strict)
 881 {
 882         struct flowi6 fl6 = {
 883                 .flowi6_oif = oif,
 884                 .daddr = *daddr,
 885         };
 886         struct dst_entry *dst;
 887         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 888
 889         if (saddr) {
 890                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 891                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 892         }
 893
 894         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 895         if (dst->error == 0)
 896                 return (struct rt6_info *) dst;
 897
 898         dst_release(dst);
 899
 900         return NULL;
 901 }
 902 EXPORT_SYMBOL(rt6_lookup);
 903
 904 /* ip6_ins_rt is called with FREE table->tb6_lock.
 905    It takes new route entry, the addition fails by any reason the
 906    route is freed. In any case, if caller does not hold it, it may
 907    be destroyed.
 908  */
 909
 910 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 911                         struct mx6_config *mxc)
 912 {
 913         int err;
 914         struct fib6_table *table;
 915
 916         table = rt->rt6i_table;
 917         write_lock_bh(&table->tb6_lock);
 918         err = fib6_add(&table->tb6_root, rt, info, mxc);
 919         write_unlock_bh(&table->tb6_lock);
 920
 921         return err;
 922 }
 923
 924 int ip6_ins_rt(struct rt6_info *rt)
 925 {
 926         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 927         struct mx6_config mxc = { .mx = NULL, };
 928
 929         return __ip6_ins_rt(rt, &info, &mxc);
 930 }
 931
 932 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 933                                            const struct in6_addr *daddr,
 934                                            const struct in6_addr *saddr)
 935 {
 936         struct rt6_info *rt;
 937
 938         /*
 939          *      Clone the route.
 940          */
 941
 942         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 943                 ort = (struct rt6_info *)ort->dst.from;
 944
 945         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 946
 947         if (!rt)
 948                 return NULL;
 949
 950         ip6_rt_copy_init(rt, ort);
 951         rt->rt6i_flags |= RTF_CACHE;
 952         rt->rt6i_metric = 0;
 953         rt->dst.flags |= DST_HOST;
 954         rt->rt6i_dst.addr = *daddr;
 955         rt->rt6i_dst.plen = 128;
 956
 957         if (!rt6_is_gw_or_nonexthop(ort)) {
 958                 if (ort->rt6i_dst.plen != 128 &&
 959                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 960                         rt->rt6i_flags |= RTF_ANYCAST;
 961 #ifdef CONFIG_IPV6_SUBTREES
 962                 if (rt->rt6i_src.plen && saddr) {
 963                         rt->rt6i_src.addr = *saddr;
 964                         rt->rt6i_src.plen = 128;
 965                 }
 966 #endif
 967         }
 968
 969         return rt;
 970 }
 971
 972 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 973 {
 974         struct rt6_info *pcpu_rt;
 975
 976         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 977                                   rt->dst.dev, rt->dst.flags);
 978
 979         if (!pcpu_rt)
 980                 return NULL;
 981         ip6_rt_copy_init(pcpu_rt, rt);
 982         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 983         pcpu_rt->rt6i_flags |= RTF_PCPU;
 984         return pcpu_rt;
 985 }
 986
 987 /* It should be called with read_lock_bh(&tb6_lock) acquired */
 988 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
 989 {
 990         struct rt6_info *pcpu_rt, **p;
 991
 992         p = this_cpu_ptr(rt->rt6i_pcpu);
 993         pcpu_rt = *p;
 994
 995         if (pcpu_rt) {
 996                 dst_hold(&pcpu_rt->dst);
 997                 rt6_dst_from_metrics_check(pcpu_rt);
 998         }
 999         return pcpu_rt;
1000 }
1001
1002 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1003 {
1004         struct fib6_table *table = rt->rt6i_table;
1005         struct rt6_info *pcpu_rt, *prev, **p;
1006
1007         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1008         if (!pcpu_rt) {
1009                 struct net *net = dev_net(rt->dst.dev);
1010
1011                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1012                 return net->ipv6.ip6_null_entry;
1013         }
1014
1015         read_lock_bh(&table->tb6_lock);
1016         if (rt->rt6i_pcpu) {
1017                 p = this_cpu_ptr(rt->rt6i_pcpu);
1018                 prev = cmpxchg(p, NULL, pcpu_rt);
1019                 if (prev) {
1020                         /* If someone did it before us, return prev instead */
1021                         dst_destroy(&pcpu_rt->dst);
1022                         pcpu_rt = prev;
1023                 }
1024         } else {
1025                 /* rt has been removed from the fib6 tree
1026                  * before we have a chance to acquire the read_lock.
1027                  * In this case, don't brother to create a pcpu rt
1028                  * since rt is going away anyway.  The next
1029                  * dst_check() will trigger a re-lookup.
1030                  */
1031                 dst_destroy(&pcpu_rt->dst);
1032                 pcpu_rt = rt;
1033         }
1034         dst_hold(&pcpu_rt->dst);
1035         rt6_dst_from_metrics_check(pcpu_rt);
1036         read_unlock_bh(&table->tb6_lock);
1037         return pcpu_rt;
1038 }
1039
1040 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1041                                       struct flowi6 *fl6, int flags)
1042 {
1043         struct fib6_node *fn, *saved_fn;
1044         struct rt6_info *rt;
1045         int strict = 0;
1046
1047         strict |= flags & RT6_LOOKUP_F_IFACE;
1048         if (net->ipv6.devconf_all->forwarding == 0)
1049                 strict |= RT6_LOOKUP_F_REACHABLE;
1050
1051         read_lock_bh(&table->tb6_lock);
1052
1053         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1054         saved_fn = fn;
1055
1056         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1057                 oif = 0;
1058
1059 redo_rt6_select:
1060         rt = rt6_select(fn, oif, strict);
1061         if (rt->rt6i_nsiblings)
1062                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1063         if (rt == net->ipv6.ip6_null_entry) {
1064                 fn = fib6_backtrack(fn, &fl6->saddr);
1065                 if (fn)
1066                         goto redo_rt6_select;
1067                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1068                         /* also consider unreachable route */
1069                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1070                         fn = saved_fn;
1071                         goto redo_rt6_select;
1072                 }
1073         }
1074
1075
1076         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1077                 dst_use(&rt->dst, jiffies);
1078                 read_unlock_bh(&table->tb6_lock);
1079
1080                 rt6_dst_from_metrics_check(rt);
1081                 return rt;
1082         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1083                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1084                 /* Create a RTF_CACHE clone which will not be
1085                  * owned by the fib6 tree.  It is for the special case where
1086                  * the daddr in the skb during the neighbor look-up is different
1087                  * from the fl6->daddr used to look-up route here.
1088                  */
1089
1090                 struct rt6_info *uncached_rt;
1091
1092                 dst_use(&rt->dst, jiffies);
1093                 read_unlock_bh(&table->tb6_lock);
1094
1095                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1096                 dst_release(&rt->dst);
1097
1098                 if (uncached_rt)
1099                         rt6_uncached_list_add(uncached_rt);
1100                 else
1101                         uncached_rt = net->ipv6.ip6_null_entry;
1102
1103                 dst_hold(&uncached_rt->dst);
1104                 return uncached_rt;
1105
1106         } else {
1107                 /* Get a percpu copy */
1108
1109                 struct rt6_info *pcpu_rt;
1110
1111                 rt->dst.lastuse = jiffies;
1112                 rt->dst.__use++;
1113                 pcpu_rt = rt6_get_pcpu_route(rt);
1114
1115                 if (pcpu_rt) {
1116                         read_unlock_bh(&table->tb6_lock);
1117                 } else {
1118                         /* We have to do the read_unlock first
1119                          * because rt6_make_pcpu_route() may trigger
1120                          * ip6_dst_gc() which will take the write_lock.
1121                          */
1122                         dst_hold(&rt->dst);
1123                         read_unlock_bh(&table->tb6_lock);
1124                         pcpu_rt = rt6_make_pcpu_route(rt);
1125                         dst_release(&rt->dst);
1126                 }
1127
1128                 return pcpu_rt;
1129
1130         }
1131 }
1132
1133 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1134                                             struct flowi6 *fl6, int flags)
1135 {
1136         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1137 }
1138
1139 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1140                                                 struct net_device *dev,
1141                                                 struct flowi6 *fl6, int flags)
1142 {
1143         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1144                 flags |= RT6_LOOKUP_F_IFACE;
1145
1146         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1147 }
1148
1149 void ip6_route_input(struct sk_buff *skb)
1150 {
1151         const struct ipv6hdr *iph = ipv6_hdr(skb);
1152         struct net *net = dev_net(skb->dev);
1153         int flags = RT6_LOOKUP_F_HAS_SADDR;
1154         struct ip_tunnel_info *tun_info;
1155         struct flowi6 fl6 = {
1156                 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1157                 .daddr = iph->daddr,
1158                 .saddr = iph->saddr,
1159                 .flowlabel = ip6_flowinfo(iph),
1160                 .flowi6_mark = skb->mark,
1161                 .flowi6_proto = iph->nexthdr,
1162         };
1163
1164         tun_info = skb_tunnel_info(skb);
1165         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1166                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1167         skb_dst_drop(skb);
1168         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1169 }
1170
1171 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1172                                              struct flowi6 *fl6, int flags)
1173 {
1174         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1175 }
1176
1177 struct dst_entry *ip6_route_output(struct net *net, const struct sock *sk,
1178                                     struct flowi6 *fl6)
1179 {
1180         struct dst_entry *dst;
1181         int flags = 0;
1182         bool any_src;
1183
1184         dst = l3mdev_rt6_dst_by_oif(net, fl6);
1185         if (dst)
1186                 return dst;
1187
1188         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1189
1190         any_src = ipv6_addr_any(&fl6->saddr);
1191         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1192             (fl6->flowi6_oif && any_src))
1193                 flags |= RT6_LOOKUP_F_IFACE;
1194
1195         if (!any_src)
1196                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1197         else if (sk)
1198                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1199
1200         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1201 }
1202 EXPORT_SYMBOL(ip6_route_output);
1203
1204 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1205 {
1206         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1207         struct dst_entry *new = NULL;
1208
1209         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1210         if (rt) {
1211                 rt6_info_init(rt);
1212
1213                 new = &rt->dst;
1214                 new->__use = 1;
1215                 new->input = dst_discard;
1216                 new->output = dst_discard_out;
1217
1218                 dst_copy_metrics(new, &ort->dst);
1219                 rt->rt6i_idev = ort->rt6i_idev;
1220                 if (rt->rt6i_idev)
1221                         in6_dev_hold(rt->rt6i_idev);
1222
1223                 rt->rt6i_gateway = ort->rt6i_gateway;
1224                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1225                 rt->rt6i_metric = 0;
1226
1227                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1228 #ifdef CONFIG_IPV6_SUBTREES
1229                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1230 #endif
1231
1232                 dst_free(new);
1233         }
1234
1235         dst_release(dst_orig);
1236         return new ? new : ERR_PTR(-ENOMEM);
1237 }
1238
1239 /*
1240  *      Destination cache support functions
1241  */
1242
1243 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1244 {
1245         if (rt->dst.from &&
1246             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1247                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1248 }
1249
1250 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1251 {
1252         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1253                 return NULL;
1254
1255         if (rt6_check_expired(rt))
1256                 return NULL;
1257
1258         return &rt->dst;
1259 }
1260
1261 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1262 {
1263         if (!__rt6_check_expired(rt) &&
1264             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1265             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1266                 return &rt->dst;
1267         else
1268                 return NULL;
1269 }
1270
1271 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1272 {
1273         struct rt6_info *rt;
1274
1275         rt = (struct rt6_info *) dst;
1276
1277         /* All IPV6 dsts are created with ->obsolete set to the value
1278          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1279          * into this function always.
1280          */
1281
1282         rt6_dst_from_metrics_check(rt);
1283
1284         if (rt->rt6i_flags & RTF_PCPU ||
1285             (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1286                 return rt6_dst_from_check(rt, cookie);
1287         else
1288                 return rt6_check(rt, cookie);
1289 }
1290
1291 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1292 {
1293         struct rt6_info *rt = (struct rt6_info *) dst;
1294
1295         if (rt) {
1296                 if (rt->rt6i_flags & RTF_CACHE) {
1297                         if (rt6_check_expired(rt)) {
1298                                 ip6_del_rt(rt);
1299                                 dst = NULL;
1300                         }
1301                 } else {
1302                         dst_release(dst);
1303                         dst = NULL;
1304                 }
1305         }
1306         return dst;
1307 }
1308
1309 static void ip6_link_failure(struct sk_buff *skb)
1310 {
1311         struct rt6_info *rt;
1312
1313         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1314
1315         rt = (struct rt6_info *) skb_dst(skb);
1316         if (rt) {
1317                 if (rt->rt6i_flags & RTF_CACHE) {
1318                         dst_hold(&rt->dst);
1319                         ip6_del_rt(rt);
1320                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1321                         rt->rt6i_node->fn_sernum = -1;
1322                 }
1323         }
1324 }
1325
1326 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1327 {
1328         struct net *net = dev_net(rt->dst.dev);
1329
1330         rt->rt6i_flags |= RTF_MODIFIED;
1331         rt->rt6i_pmtu = mtu;
1332         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1333 }
1334
1335 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1336 {
1337         return !(rt->rt6i_flags & RTF_CACHE) &&
1338                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1339 }
1340
1341 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1342                                  const struct ipv6hdr *iph, u32 mtu)
1343 {
1344         struct rt6_info *rt6 = (struct rt6_info *)dst;
1345
1346         if (rt6->rt6i_flags & RTF_LOCAL)
1347                 return;
1348
1349         dst_confirm(dst);
1350         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1351         if (mtu >= dst_mtu(dst))
1352                 return;
1353
1354         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1355                 rt6_do_update_pmtu(rt6, mtu);
1356         } else {
1357                 const struct in6_addr *daddr, *saddr;
1358                 struct rt6_info *nrt6;
1359
1360                 if (iph) {
1361                         daddr = &iph->daddr;
1362                         saddr = &iph->saddr;
1363                 } else if (sk) {
1364                         daddr = &sk->sk_v6_daddr;
1365                         saddr = &inet6_sk(sk)->saddr;
1366                 } else {
1367                         return;
1368                 }
1369                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1370                 if (nrt6) {
1371                         rt6_do_update_pmtu(nrt6, mtu);
1372
1373                         /* ip6_ins_rt(nrt6) will bump the
1374                          * rt6->rt6i_node->fn_sernum
1375                          * which will fail the next rt6_check() and
1376                          * invalidate the sk->sk_dst_cache.
1377                          */
1378                         ip6_ins_rt(nrt6);
1379                 }
1380         }
1381 }
1382
1383 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1384                                struct sk_buff *skb, u32 mtu)
1385 {
1386         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1387 }
1388
1389 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1390                      int oif, u32 mark)
1391 {
1392         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1393         struct dst_entry *dst;
1394         struct flowi6 fl6;
1395
1396         memset(&fl6, 0, sizeof(fl6));
1397         fl6.flowi6_oif = oif;
1398         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1399         fl6.daddr = iph->daddr;
1400         fl6.saddr = iph->saddr;
1401         fl6.flowlabel = ip6_flowinfo(iph);
1402
1403         dst = ip6_route_output(net, NULL, &fl6);
1404         if (!dst->error)
1405                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1406         dst_release(dst);
1407 }
1408 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1409
1410 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1411 {
1412         ip6_update_pmtu(skb, sock_net(sk), mtu,
1413                         sk->sk_bound_dev_if, sk->sk_mark);
1414 }
1415 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1416
1417 /* Handle redirects */
1418 struct ip6rd_flowi {
1419         struct flowi6 fl6;
1420         struct in6_addr gateway;
1421 };
1422
1423 static struct rt6_info *__ip6_route_redirect(struct net *net,
1424                                              struct fib6_table *table,
1425                                              struct flowi6 *fl6,
1426                                              int flags)
1427 {
1428         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1429         struct rt6_info *rt;
1430         struct fib6_node *fn;
1431
1432         /* Get the "current" route for this destination and
1433          * check if the redirect has come from approriate router.
1434          *
1435          * RFC 4861 specifies that redirects should only be
1436          * accepted if they come from the nexthop to the target.
1437          * Due to the way the routes are chosen, this notion
1438          * is a bit fuzzy and one might need to check all possible
1439          * routes.
1440          */
1441
1442         read_lock_bh(&table->tb6_lock);
1443         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1444 restart:
1445         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1446                 if (rt6_check_expired(rt))
1447                         continue;
1448                 if (rt->dst.error)
1449                         break;
1450                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1451                         continue;
1452                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1453                         continue;
1454                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1455                         continue;
1456                 break;
1457         }
1458
1459         if (!rt)
1460                 rt = net->ipv6.ip6_null_entry;
1461         else if (rt->dst.error) {
1462                 rt = net->ipv6.ip6_null_entry;
1463                 goto out;
1464         }
1465
1466         if (rt == net->ipv6.ip6_null_entry) {
1467                 fn = fib6_backtrack(fn, &fl6->saddr);
1468                 if (fn)
1469                         goto restart;
1470         }
1471
1472 out:
1473         dst_hold(&rt->dst);
1474
1475         read_unlock_bh(&table->tb6_lock);
1476
1477         return rt;
1478 };
1479
1480 static struct dst_entry *ip6_route_redirect(struct net *net,
1481                                         const struct flowi6 *fl6,
1482                                         const struct in6_addr *gateway)
1483 {
1484         int flags = RT6_LOOKUP_F_HAS_SADDR;
1485         struct ip6rd_flowi rdfl;
1486
1487         rdfl.fl6 = *fl6;
1488         rdfl.gateway = *gateway;
1489
1490         return fib6_rule_lookup(net, &rdfl.fl6,
1491                                 flags, __ip6_route_redirect);
1492 }
1493
1494 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1495 {
1496         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1497         struct dst_entry *dst;
1498         struct flowi6 fl6;
1499
1500         memset(&fl6, 0, sizeof(fl6));
1501         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1502         fl6.flowi6_oif = oif;
1503         fl6.flowi6_mark = mark;
1504         fl6.daddr = iph->daddr;
1505         fl6.saddr = iph->saddr;
1506         fl6.flowlabel = ip6_flowinfo(iph);
1507
1508         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1509         rt6_do_redirect(dst, NULL, skb);
1510         dst_release(dst);
1511 }
1512 EXPORT_SYMBOL_GPL(ip6_redirect);
1513
1514 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1515                             u32 mark)
1516 {
1517         const struct ipv6hdr *iph = ipv6_hdr(skb);
1518         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1519         struct dst_entry *dst;
1520         struct flowi6 fl6;
1521
1522         memset(&fl6, 0, sizeof(fl6));
1523         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1524         fl6.flowi6_oif = oif;
1525         fl6.flowi6_mark = mark;
1526         fl6.daddr = msg->dest;
1527         fl6.saddr = iph->daddr;
1528
1529         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1530         rt6_do_redirect(dst, NULL, skb);
1531         dst_release(dst);
1532 }
1533
1534 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1535 {
1536         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1537 }
1538 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1539
1540 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1541 {
1542         struct net_device *dev = dst->dev;
1543         unsigned int mtu = dst_mtu(dst);
1544         struct net *net = dev_net(dev);
1545
1546         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1547
1548         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1549                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1550
1551         /*
1552          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1553          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1554          * IPV6_MAXPLEN is also valid and means: "any MSS,
1555          * rely only on pmtu discovery"
1556          */
1557         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1558                 mtu = IPV6_MAXPLEN;
1559         return mtu;
1560 }
1561
1562 static unsigned int ip6_mtu(const struct dst_entry *dst)
1563 {
1564         const struct rt6_info *rt = (const struct rt6_info *)dst;
1565         unsigned int mtu = rt->rt6i_pmtu;
1566         struct inet6_dev *idev;
1567
1568         if (mtu)
1569                 goto out;
1570
1571         mtu = dst_metric_raw(dst, RTAX_MTU);
1572         if (mtu)
1573                 goto out;
1574
1575         mtu = IPV6_MIN_MTU;
1576
1577         rcu_read_lock();
1578         idev = __in6_dev_get(dst->dev);
1579         if (idev)
1580                 mtu = idev->cnf.mtu6;
1581         rcu_read_unlock();
1582
1583 out:
1584         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1585 }
1586
1587 static struct dst_entry *icmp6_dst_gc_list;
1588 static DEFINE_SPINLOCK(icmp6_dst_lock);
1589
1590 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1591                                   struct flowi6 *fl6)
1592 {
1593         struct dst_entry *dst;
1594         struct rt6_info *rt;
1595         struct inet6_dev *idev = in6_dev_get(dev);
1596         struct net *net = dev_net(dev);
1597
1598         if (unlikely(!idev))
1599                 return ERR_PTR(-ENODEV);
1600
1601         rt = ip6_dst_alloc(net, dev, 0);
1602         if (unlikely(!rt)) {
1603                 in6_dev_put(idev);
1604                 dst = ERR_PTR(-ENOMEM);
1605                 goto out;
1606         }
1607
1608         rt->dst.flags |= DST_HOST;
1609         rt->dst.output  = ip6_output;
1610         atomic_set(&rt->dst.__refcnt, 1);
1611         rt->rt6i_gateway  = fl6->daddr;
1612         rt->rt6i_dst.addr = fl6->daddr;
1613         rt->rt6i_dst.plen = 128;
1614         rt->rt6i_idev     = idev;
1615         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1616
1617         spin_lock_bh(&icmp6_dst_lock);
1618         rt->dst.next = icmp6_dst_gc_list;
1619         icmp6_dst_gc_list = &rt->dst;
1620         spin_unlock_bh(&icmp6_dst_lock);
1621
1622         fib6_force_start_gc(net);
1623
1624         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1625
1626 out:
1627         return dst;
1628 }
1629
1630 int icmp6_dst_gc(void)
1631 {
1632         struct dst_entry *dst, **pprev;
1633         int more = 0;
1634
1635         spin_lock_bh(&icmp6_dst_lock);
1636         pprev = &icmp6_dst_gc_list;
1637
1638         while ((dst = *pprev) != NULL) {
1639                 if (!atomic_read(&dst->__refcnt)) {
1640                         *pprev = dst->next;
1641                         dst_free(dst);
1642                 } else {
1643                         pprev = &dst->next;
1644                         ++more;
1645                 }
1646         }
1647
1648         spin_unlock_bh(&icmp6_dst_lock);
1649
1650         return more;
1651 }
1652
1653 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1654                             void *arg)
1655 {
1656         struct dst_entry *dst, **pprev;
1657
1658         spin_lock_bh(&icmp6_dst_lock);
1659         pprev = &icmp6_dst_gc_list;
1660         while ((dst = *pprev) != NULL) {
1661                 struct rt6_info *rt = (struct rt6_info *) dst;
1662                 if (func(rt, arg)) {
1663                         *pprev = dst->next;
1664                         dst_free(dst);
1665                 } else {
1666                         pprev = &dst->next;
1667                 }
1668         }
1669         spin_unlock_bh(&icmp6_dst_lock);
1670 }
1671
1672 static int ip6_dst_gc(struct dst_ops *ops)
1673 {
1674         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1675         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1676         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1677         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1678         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1679         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1680         int entries;
1681
1682         entries = dst_entries_get_fast(ops);
1683         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1684             entries <= rt_max_size)
1685                 goto out;
1686
1687         net->ipv6.ip6_rt_gc_expire++;
1688         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1689         entries = dst_entries_get_slow(ops);
1690         if (entries < ops->gc_thresh)
1691                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1692 out:
1693         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1694         return entries > rt_max_size;
1695 }
1696
1697 static int ip6_convert_metrics(struct mx6_config *mxc,
1698                                const struct fib6_config *cfg)
1699 {
1700         bool ecn_ca = false;
1701         struct nlattr *nla;
1702         int remaining;
1703         u32 *mp;
1704
1705         if (!cfg->fc_mx)
1706                 return 0;
1707
1708         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1709         if (unlikely(!mp))
1710                 return -ENOMEM;
1711
1712         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1713                 int type = nla_type(nla);
1714                 u32 val;
1715
1716                 if (!type)
1717                         continue;
1718                 if (unlikely(type > RTAX_MAX))
1719                         goto err;
1720
1721                 if (type == RTAX_CC_ALGO) {
1722                         char tmp[TCP_CA_NAME_MAX];
1723
1724                         nla_strlcpy(tmp, nla, sizeof(tmp));
1725                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1726                         if (val == TCP_CA_UNSPEC)
1727                                 goto err;
1728                 } else {
1729                         val = nla_get_u32(nla);
1730                 }
1731                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1732                         goto err;
1733
1734                 mp[type - 1] = val;
1735                 __set_bit(type - 1, mxc->mx_valid);
1736         }
1737
1738         if (ecn_ca) {
1739                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1740                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1741         }
1742
1743         mxc->mx = mp;
1744         return 0;
1745  err:
1746         kfree(mp);
1747         return -EINVAL;
1748 }
1749
1750 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1751 {
1752         struct net *net = cfg->fc_nlinfo.nl_net;
1753         struct rt6_info *rt = NULL;
1754         struct net_device *dev = NULL;
1755         struct inet6_dev *idev = NULL;
1756         struct fib6_table *table;
1757         int addr_type;
1758         int err = -EINVAL;
1759
1760         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1761                 goto out;
1762 #ifndef CONFIG_IPV6_SUBTREES
1763         if (cfg->fc_src_len)
1764                 goto out;
1765 #endif
1766         if (cfg->fc_ifindex) {
1767                 err = -ENODEV;
1768                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1769                 if (!dev)
1770                         goto out;
1771                 idev = in6_dev_get(dev);
1772                 if (!idev)
1773                         goto out;
1774         }
1775
1776         if (cfg->fc_metric == 0)
1777                 cfg->fc_metric = IP6_RT_PRIO_USER;
1778
1779         err = -ENOBUFS;
1780         if (cfg->fc_nlinfo.nlh &&
1781             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1782                 table = fib6_get_table(net, cfg->fc_table);
1783                 if (!table) {
1784                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1785                         table = fib6_new_table(net, cfg->fc_table);
1786                 }
1787         } else {
1788                 table = fib6_new_table(net, cfg->fc_table);
1789         }
1790
1791         if (!table)
1792                 goto out;
1793
1794         rt = ip6_dst_alloc(net, NULL,
1795                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1796
1797         if (!rt) {
1798                 err = -ENOMEM;
1799                 goto out;
1800         }
1801
1802         if (cfg->fc_flags & RTF_EXPIRES)
1803                 rt6_set_expires(rt, jiffies +
1804                                 clock_t_to_jiffies(cfg->fc_expires));
1805         else
1806                 rt6_clean_expires(rt);
1807
1808         if (cfg->fc_protocol == RTPROT_UNSPEC)
1809                 cfg->fc_protocol = RTPROT_BOOT;
1810         rt->rt6i_protocol = cfg->fc_protocol;
1811
1812         addr_type = ipv6_addr_type(&cfg->fc_dst);
1813
1814         if (addr_type & IPV6_ADDR_MULTICAST)
1815                 rt->dst.input = ip6_mc_input;
1816         else if (cfg->fc_flags & RTF_LOCAL)
1817                 rt->dst.input = ip6_input;
1818         else
1819                 rt->dst.input = ip6_forward;
1820
1821         rt->dst.output = ip6_output;
1822
1823         if (cfg->fc_encap) {
1824                 struct lwtunnel_state *lwtstate;
1825
1826                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1827                                            cfg->fc_encap, AF_INET6, cfg,
1828                                            &lwtstate);
1829                 if (err)
1830                         goto out;
1831                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1832                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1833                         rt->dst.lwtstate->orig_output = rt->dst.output;
1834                         rt->dst.output = lwtunnel_output;
1835                 }
1836                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1837                         rt->dst.lwtstate->orig_input = rt->dst.input;
1838                         rt->dst.input = lwtunnel_input;
1839                 }
1840         }
1841
1842         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1843         rt->rt6i_dst.plen = cfg->fc_dst_len;
1844         if (rt->rt6i_dst.plen == 128)
1845                 rt->dst.flags |= DST_HOST;
1846
1847 #ifdef CONFIG_IPV6_SUBTREES
1848         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1849         rt->rt6i_src.plen = cfg->fc_src_len;
1850 #endif
1851
1852         rt->rt6i_metric = cfg->fc_metric;
1853
1854         /* We cannot add true routes via loopback here,
1855            they would result in kernel looping; promote them to reject routes
1856          */
1857         if ((cfg->fc_flags & RTF_REJECT) ||
1858             (dev && (dev->flags & IFF_LOOPBACK) &&
1859              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1860              !(cfg->fc_flags & RTF_LOCAL))) {
1861                 /* hold loopback dev/idev if we haven't done so. */
1862                 if (dev != net->loopback_dev) {
1863                         if (dev) {
1864                                 dev_put(dev);
1865                                 in6_dev_put(idev);
1866                         }
1867                         dev = net->loopback_dev;
1868                         dev_hold(dev);
1869                         idev = in6_dev_get(dev);
1870                         if (!idev) {
1871                                 err = -ENODEV;
1872                                 goto out;
1873                         }
1874                 }
1875                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1876                 switch (cfg->fc_type) {
1877                 case RTN_BLACKHOLE:
1878                         rt->dst.error = -EINVAL;
1879                         rt->dst.output = dst_discard_out;
1880                         rt->dst.input = dst_discard;
1881                         break;
1882                 case RTN_PROHIBIT:
1883                         rt->dst.error = -EACCES;
1884                         rt->dst.output = ip6_pkt_prohibit_out;
1885                         rt->dst.input = ip6_pkt_prohibit;
1886                         break;
1887                 case RTN_THROW:
1888                 case RTN_UNREACHABLE:
1889                 default:
1890                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1891                                         : (cfg->fc_type == RTN_UNREACHABLE)
1892                                         ? -EHOSTUNREACH : -ENETUNREACH;
1893                         rt->dst.output = ip6_pkt_discard_out;
1894                         rt->dst.input = ip6_pkt_discard;
1895                         break;
1896                 }
1897                 goto install_route;
1898         }
1899
1900         if (cfg->fc_flags & RTF_GATEWAY) {
1901                 const struct in6_addr *gw_addr;
1902                 int gwa_type;
1903
1904                 gw_addr = &cfg->fc_gateway;
1905                 gwa_type = ipv6_addr_type(gw_addr);
1906
1907                 /* if gw_addr is local we will fail to detect this in case
1908                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1909                  * will return already-added prefix route via interface that
1910                  * prefix route was assigned to, which might be non-loopback.
1911                  */
1912                 err = -EINVAL;
1913                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1914                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1915                                             dev : NULL, 0, 0))
1916                         goto out;
1917
1918                 rt->rt6i_gateway = *gw_addr;
1919
1920                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1921                         struct rt6_info *grt;
1922
1923                         /* IPv6 strictly inhibits using not link-local
1924                            addresses as nexthop address.
1925                            Otherwise, router will not able to send redirects.
1926                            It is very good, but in some (rare!) circumstances
1927                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1928                            some exceptions. --ANK
1929                          */
1930                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1931                                 goto out;
1932
1933                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1934
1935                         err = -EHOSTUNREACH;
1936                         if (!grt)
1937                                 goto out;
1938                         if (dev) {
1939                                 if (dev != grt->dst.dev) {
1940                                         ip6_rt_put(grt);
1941                                         goto out;
1942                                 }
1943                         } else {
1944                                 dev = grt->dst.dev;
1945                                 idev = grt->rt6i_idev;
1946                                 dev_hold(dev);
1947                                 in6_dev_hold(grt->rt6i_idev);
1948                         }
1949                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1950                                 err = 0;
1951                         ip6_rt_put(grt);
1952
1953                         if (err)
1954                                 goto out;
1955                 }
1956                 err = -EINVAL;
1957                 if (!dev || (dev->flags & IFF_LOOPBACK))
1958                         goto out;
1959         }
1960
1961         err = -ENODEV;
1962         if (!dev)
1963                 goto out;
1964
1965         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1966                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1967                         err = -EINVAL;
1968                         goto out;
1969                 }
1970                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1971                 rt->rt6i_prefsrc.plen = 128;
1972         } else
1973                 rt->rt6i_prefsrc.plen = 0;
1974
1975         rt->rt6i_flags = cfg->fc_flags;
1976
1977 install_route:
1978         rt->dst.dev = dev;
1979         rt->rt6i_idev = idev;
1980         rt->rt6i_table = table;
1981
1982         cfg->fc_nlinfo.nl_net = dev_net(dev);
1983
1984         return rt;
1985 out:
1986         if (dev)
1987                 dev_put(dev);
1988         if (idev)
1989                 in6_dev_put(idev);
1990         if (rt)
1991                 dst_free(&rt->dst);
1992
1993         return ERR_PTR(err);
1994 }
1995
1996 int ip6_route_add(struct fib6_config *cfg)
1997 {
1998         struct mx6_config mxc = { .mx = NULL, };
1999         struct rt6_info *rt;
2000         int err;
2001
2002         rt = ip6_route_info_create(cfg);
2003         if (IS_ERR(rt)) {
2004                 err = PTR_ERR(rt);
2005                 rt = NULL;
2006                 goto out;
2007         }
2008
2009         err = ip6_convert_metrics(&mxc, cfg);
2010         if (err)
2011                 goto out;
2012
2013         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2014
2015         kfree(mxc.mx);
2016
2017         return err;
2018 out:
2019         if (rt)
2020                 dst_free(&rt->dst);
2021
2022         return err;
2023 }
2024
2025 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2026 {
2027         int err;
2028         struct fib6_table *table;
2029         struct net *net = dev_net(rt->dst.dev);
2030
2031         if (rt == net->ipv6.ip6_null_entry ||
2032             rt->dst.flags & DST_NOCACHE) {
2033                 err = -ENOENT;
2034                 goto out;
2035         }
2036
2037         table = rt->rt6i_table;
2038         write_lock_bh(&table->tb6_lock);
2039         err = fib6_del(rt, info);
2040         write_unlock_bh(&table->tb6_lock);
2041
2042 out:
2043         ip6_rt_put(rt);
2044         return err;
2045 }
2046
2047 int ip6_del_rt(struct rt6_info *rt)
2048 {
2049         struct nl_info info = {
2050                 .nl_net = dev_net(rt->dst.dev),
2051         };
2052         return __ip6_del_rt(rt, &info);
2053 }
2054
2055 static int ip6_route_del(struct fib6_config *cfg)
2056 {
2057         struct fib6_table *table;
2058         struct fib6_node *fn;
2059         struct rt6_info *rt;
2060         int err = -ESRCH;
2061
2062         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2063         if (!table)
2064                 return err;
2065
2066         read_lock_bh(&table->tb6_lock);
2067
2068         fn = fib6_locate(&table->tb6_root,
2069                          &cfg->fc_dst, cfg->fc_dst_len,
2070                          &cfg->fc_src, cfg->fc_src_len);
2071
2072         if (fn) {
2073                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2074                         if ((rt->rt6i_flags & RTF_CACHE) &&
2075                             !(cfg->fc_flags & RTF_CACHE))
2076                                 continue;
2077                         if (cfg->fc_ifindex &&
2078                             (!rt->dst.dev ||
2079                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2080                                 continue;
2081                         if (cfg->fc_flags & RTF_GATEWAY &&
2082                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2083                                 continue;
2084                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2085                                 continue;
2086                         dst_hold(&rt->dst);
2087                         read_unlock_bh(&table->tb6_lock);
2088
2089                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2090                 }
2091         }
2092         read_unlock_bh(&table->tb6_lock);
2093
2094         return err;
2095 }
2096
2097 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2098 {
2099         struct netevent_redirect netevent;
2100         struct rt6_info *rt, *nrt = NULL;
2101         struct ndisc_options ndopts;
2102         struct inet6_dev *in6_dev;
2103         struct neighbour *neigh;
2104         struct rd_msg *msg;
2105         int optlen, on_link;
2106         u8 *lladdr;
2107
2108         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2109         optlen -= sizeof(*msg);
2110
2111         if (optlen < 0) {
2112                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2113                 return;
2114         }
2115
2116         msg = (struct rd_msg *)icmp6_hdr(skb);
2117
2118         if (ipv6_addr_is_multicast(&msg->dest)) {
2119                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2120                 return;
2121         }
2122
2123         on_link = 0;
2124         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2125                 on_link = 1;
2126         } else if (ipv6_addr_type(&msg->target) !=
2127                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2128                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2129                 return;
2130         }
2131
2132         in6_dev = __in6_dev_get(skb->dev);
2133         if (!in6_dev)
2134                 return;
2135         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2136                 return;
2137
2138         /* RFC2461 8.1:
2139          *      The IP source address of the Redirect MUST be the same as the current
2140          *      first-hop router for the specified ICMP Destination Address.
2141          */
2142
2143         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2144                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2145                 return;
2146         }
2147
2148         lladdr = NULL;
2149         if (ndopts.nd_opts_tgt_lladdr) {
2150                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2151                                              skb->dev);
2152                 if (!lladdr) {
2153                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2154                         return;
2155                 }
2156         }
2157
2158         rt = (struct rt6_info *) dst;
2159         if (rt->rt6i_flags & RTF_REJECT) {
2160                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2161                 return;
2162         }
2163
2164         /* Redirect received -> path was valid.
2165          * Look, redirects are sent only in response to data packets,
2166          * so that this nexthop apparently is reachable. --ANK
2167          */
2168         dst_confirm(&rt->dst);
2169
2170         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2171         if (!neigh)
2172                 return;
2173
2174         /*
2175          *      We have finally decided to accept it.
2176          */
2177
2178         neigh_update(neigh, lladdr, NUD_STALE,
2179                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2180                      NEIGH_UPDATE_F_OVERRIDE|
2181                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2182                                      NEIGH_UPDATE_F_ISROUTER))
2183                      );
2184
2185         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2186         if (!nrt)
2187                 goto out;
2188
2189         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2190         if (on_link)
2191                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2192
2193         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2194
2195         if (ip6_ins_rt(nrt))
2196                 goto out;
2197
2198         netevent.old = &rt->dst;
2199         netevent.new = &nrt->dst;
2200         netevent.daddr = &msg->dest;
2201         netevent.neigh = neigh;
2202         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2203
2204         if (rt->rt6i_flags & RTF_CACHE) {
2205                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2206                 ip6_del_rt(rt);
2207         }
2208
2209 out:
2210         neigh_release(neigh);
2211 }
2212
2213 /*
2214  *      Misc support functions
2215  */
2216
2217 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2218 {
2219         BUG_ON(from->dst.from);
2220
2221         rt->rt6i_flags &= ~RTF_EXPIRES;
2222         dst_hold(&from->dst);
2223         rt->dst.from = &from->dst;
2224         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2225 }
2226
2227 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2228 {
2229         rt->dst.input = ort->dst.input;
2230         rt->dst.output = ort->dst.output;
2231         rt->rt6i_dst = ort->rt6i_dst;
2232         rt->dst.error = ort->dst.error;
2233         rt->rt6i_idev = ort->rt6i_idev;
2234         if (rt->rt6i_idev)
2235                 in6_dev_hold(rt->rt6i_idev);
2236         rt->dst.lastuse = jiffies;
2237         rt->rt6i_gateway = ort->rt6i_gateway;
2238         rt->rt6i_flags = ort->rt6i_flags;
2239         rt6_set_from(rt, ort);
2240         rt->rt6i_metric = ort->rt6i_metric;
2241 #ifdef CONFIG_IPV6_SUBTREES
2242         rt->rt6i_src = ort->rt6i_src;
2243 #endif
2244         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2245         rt->rt6i_table = ort->rt6i_table;
2246         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2247 }
2248
2249 #ifdef CONFIG_IPV6_ROUTE_INFO
2250 static struct rt6_info *rt6_get_route_info(struct net *net,
2251                                            const struct in6_addr *prefix, int prefixlen,
2252                                            const struct in6_addr *gwaddr, int ifindex)
2253 {
2254         struct fib6_node *fn;
2255         struct rt6_info *rt = NULL;
2256         struct fib6_table *table;
2257
2258         table = fib6_get_table(net, RT6_TABLE_INFO);
2259         if (!table)
2260                 return NULL;
2261
2262         read_lock_bh(&table->tb6_lock);
2263         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2264         if (!fn)
2265                 goto out;
2266
2267         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2268                 if (rt->dst.dev->ifindex != ifindex)
2269                         continue;
2270                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2271                         continue;
2272                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2273                         continue;
2274                 dst_hold(&rt->dst);
2275                 break;
2276         }
2277 out:
2278         read_unlock_bh(&table->tb6_lock);
2279         return rt;
2280 }
2281
2282 static struct rt6_info *rt6_add_route_info(struct net *net,
2283                                            const struct in6_addr *prefix, int prefixlen,
2284                                            const struct in6_addr *gwaddr, int ifindex,
2285                                            unsigned int pref)
2286 {
2287         struct fib6_config cfg = {
2288                 .fc_metric      = IP6_RT_PRIO_USER,
2289                 .fc_ifindex     = ifindex,
2290                 .fc_dst_len     = prefixlen,
2291                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2292                                   RTF_UP | RTF_PREF(pref),
2293                 .fc_nlinfo.portid = 0,
2294                 .fc_nlinfo.nlh = NULL,
2295                 .fc_nlinfo.nl_net = net,
2296         };
2297
2298         cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2299         cfg.fc_dst = *prefix;
2300         cfg.fc_gateway = *gwaddr;
2301
2302         /* We should treat it as a default route if prefix length is 0. */
2303         if (!prefixlen)
2304                 cfg.fc_flags |= RTF_DEFAULT;
2305
2306         ip6_route_add(&cfg);
2307
2308         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2309 }
2310 #endif
2311
2312 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2313 {
2314         struct rt6_info *rt;
2315         struct fib6_table *table;
2316
2317         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2318         if (!table)
2319                 return NULL;
2320
2321         read_lock_bh(&table->tb6_lock);
2322         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2323                 if (dev == rt->dst.dev &&
2324                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2325                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2326                         break;
2327         }
2328         if (rt)
2329                 dst_hold(&rt->dst);
2330         read_unlock_bh(&table->tb6_lock);
2331         return rt;
2332 }
2333
2334 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2335                                      struct net_device *dev,
2336                                      unsigned int pref)
2337 {
2338         struct fib6_config cfg = {
2339                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2340                 .fc_metric      = IP6_RT_PRIO_USER,
2341                 .fc_ifindex     = dev->ifindex,
2342                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2343                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2344                 .fc_nlinfo.portid = 0,
2345                 .fc_nlinfo.nlh = NULL,
2346                 .fc_nlinfo.nl_net = dev_net(dev),
2347         };
2348
2349         cfg.fc_gateway = *gwaddr;
2350
2351         ip6_route_add(&cfg);
2352
2353         return rt6_get_dflt_router(gwaddr, dev);
2354 }
2355
2356 void rt6_purge_dflt_routers(struct net *net)
2357 {
2358         struct rt6_info *rt;
2359         struct fib6_table *table;
2360
2361         /* NOTE: Keep consistent with rt6_get_dflt_router */
2362         table = fib6_get_table(net, RT6_TABLE_DFLT);
2363         if (!table)
2364                 return;
2365
2366 restart:
2367         read_lock_bh(&table->tb6_lock);
2368         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2369                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2370                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2371                         dst_hold(&rt->dst);
2372                         read_unlock_bh(&table->tb6_lock);
2373                         ip6_del_rt(rt);
2374                         goto restart;
2375                 }
2376         }
2377         read_unlock_bh(&table->tb6_lock);
2378 }
2379
2380 static void rtmsg_to_fib6_config(struct net *net,
2381                                  struct in6_rtmsg *rtmsg,
2382                                  struct fib6_config *cfg)
2383 {
2384         memset(cfg, 0, sizeof(*cfg));
2385
2386         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2387                          : RT6_TABLE_MAIN;
2388         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2389         cfg->fc_metric = rtmsg->rtmsg_metric;
2390         cfg->fc_expires = rtmsg->rtmsg_info;
2391         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2392         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2393         cfg->fc_flags = rtmsg->rtmsg_flags;
2394
2395         cfg->fc_nlinfo.nl_net = net;
2396
2397         cfg->fc_dst = rtmsg->rtmsg_dst;
2398         cfg->fc_src = rtmsg->rtmsg_src;
2399         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2400 }
2401
2402 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2403 {
2404         struct fib6_config cfg;
2405         struct in6_rtmsg rtmsg;
2406         int err;
2407
2408         switch (cmd) {
2409         case SIOCADDRT:         /* Add a route */
2410         case SIOCDELRT:         /* Delete a route */
2411                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2412                         return -EPERM;
2413                 err = copy_from_user(&rtmsg, arg,
2414                                      sizeof(struct in6_rtmsg));
2415                 if (err)
2416                         return -EFAULT;
2417
2418                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2419
2420                 rtnl_lock();
2421                 switch (cmd) {
2422                 case SIOCADDRT:
2423                         err = ip6_route_add(&cfg);
2424                         break;
2425                 case SIOCDELRT:
2426                         err = ip6_route_del(&cfg);
2427                         break;
2428                 default:
2429                         err = -EINVAL;
2430                 }
2431                 rtnl_unlock();
2432
2433                 return err;
2434         }
2435
2436         return -EINVAL;
2437 }
2438
2439 /*
2440  *      Drop the packet on the floor
2441  */
2442
2443 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2444 {
2445         int type;
2446         struct dst_entry *dst = skb_dst(skb);
2447         switch (ipstats_mib_noroutes) {
2448         case IPSTATS_MIB_INNOROUTES:
2449                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2450                 if (type == IPV6_ADDR_ANY) {
2451                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2452                                       IPSTATS_MIB_INADDRERRORS);
2453                         break;
2454                 }
2455                 /* FALLTHROUGH */
2456         case IPSTATS_MIB_OUTNOROUTES:
2457                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2458                               ipstats_mib_noroutes);
2459                 break;
2460         }
2461         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2462         kfree_skb(skb);
2463         return 0;
2464 }
2465
2466 static int ip6_pkt_discard(struct sk_buff *skb)
2467 {
2468         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2469 }
2470
2471 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2472 {
2473         skb->dev = skb_dst(skb)->dev;
2474         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2475 }
2476
2477 static int ip6_pkt_prohibit(struct sk_buff *skb)
2478 {
2479         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2480 }
2481
2482 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2483 {
2484         skb->dev = skb_dst(skb)->dev;
2485         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2486 }
2487
2488 /*
2489  *      Allocate a dst for local (unicast / anycast) address.
2490  */
2491
2492 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2493                                     const struct in6_addr *addr,
2494                                     bool anycast)
2495 {
2496         u32 tb_id;
2497         struct net *net = dev_net(idev->dev);
2498         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2499                                             DST_NOCOUNT);
2500         if (!rt)
2501                 return ERR_PTR(-ENOMEM);
2502
2503         in6_dev_hold(idev);
2504
2505         rt->dst.flags |= DST_HOST;
2506         rt->dst.input = ip6_input;
2507         rt->dst.output = ip6_output;
2508         rt->rt6i_idev = idev;
2509
2510         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2511         if (anycast)
2512                 rt->rt6i_flags |= RTF_ANYCAST;
2513         else
2514                 rt->rt6i_flags |= RTF_LOCAL;
2515
2516         rt->rt6i_gateway  = *addr;
2517         rt->rt6i_dst.addr = *addr;
2518         rt->rt6i_dst.plen = 128;
2519         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2520         rt->rt6i_table = fib6_get_table(net, tb_id);
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                 rt = ip6_route_info_create(&r_cfg);
2900                 if (IS_ERR(rt)) {
2901                         err = PTR_ERR(rt);
2902                         rt = NULL;
2903                         goto cleanup;
2904                 }
2905
2906                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2907                 if (err) {
2908                         dst_free(&rt->dst);
2909                         goto cleanup;
2910                 }
2911
2912                 rtnh = rtnh_next(rtnh, &remaining);
2913         }
2914
2915         err_nh = NULL;
2916         list_for_each_entry(nh, &rt6_nh_list, next) {
2917                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2918                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2919                 nh->rt6_info = NULL;
2920                 if (err) {
2921                         if (replace && nhn)
2922                                 ip6_print_replace_route_err(&rt6_nh_list);
2923                         err_nh = nh;
2924                         goto add_errout;
2925                 }
2926
2927                 /* Because each route is added like a single route we remove
2928                  * these flags after the first nexthop: if there is a collision,
2929                  * we have already failed to add the first nexthop:
2930                  * fib6_add_rt2node() has rejected it; when replacing, old
2931                  * nexthops have been replaced by first new, the rest should
2932                  * be added to it.
2933                  */
2934                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2935                                                      NLM_F_REPLACE);
2936                 nhn++;
2937         }
2938
2939         goto cleanup;
2940
2941 add_errout:
2942         /* Delete routes that were already added */
2943         list_for_each_entry(nh, &rt6_nh_list, next) {
2944                 if (err_nh == nh)
2945                         break;
2946                 ip6_route_del(&nh->r_cfg);
2947         }
2948
2949 cleanup:
2950         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2951                 if (nh->rt6_info)
2952                         dst_free(&nh->rt6_info->dst);
2953                 kfree(nh->mxc.mx);
2954                 list_del(&nh->next);
2955                 kfree(nh);
2956         }
2957
2958         return err;
2959 }
2960
2961 static int ip6_route_multipath_del(struct fib6_config *cfg)
2962 {
2963         struct fib6_config r_cfg;
2964         struct rtnexthop *rtnh;
2965         int remaining;
2966         int attrlen;
2967         int err = 1, last_err = 0;
2968
2969         remaining = cfg->fc_mp_len;
2970         rtnh = (struct rtnexthop *)cfg->fc_mp;
2971
2972         /* Parse a Multipath Entry */
2973         while (rtnh_ok(rtnh, remaining)) {
2974                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2975                 if (rtnh->rtnh_ifindex)
2976                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2977
2978                 attrlen = rtnh_attrlen(rtnh);
2979                 if (attrlen > 0) {
2980                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2981
2982                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2983                         if (nla) {
2984                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
2985                                 r_cfg.fc_flags |= RTF_GATEWAY;
2986                         }
2987                 }
2988                 err = ip6_route_del(&r_cfg);
2989                 if (err)
2990                         last_err = err;
2991
2992                 rtnh = rtnh_next(rtnh, &remaining);
2993         }
2994
2995         return last_err;
2996 }
2997
2998 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
2999 {
3000         struct fib6_config cfg;
3001         int err;
3002
3003         err = rtm_to_fib6_config(skb, nlh, &cfg);
3004         if (err < 0)
3005                 return err;
3006
3007         if (cfg.fc_mp)
3008                 return ip6_route_multipath_del(&cfg);
3009         else
3010                 return ip6_route_del(&cfg);
3011 }
3012
3013 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3014 {
3015         struct fib6_config cfg;
3016         int err;
3017
3018         err = rtm_to_fib6_config(skb, nlh, &cfg);
3019         if (err < 0)
3020                 return err;
3021
3022         if (cfg.fc_mp)
3023                 return ip6_route_multipath_add(&cfg);
3024         else
3025                 return ip6_route_add(&cfg);
3026 }
3027
3028 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3029 {
3030         return NLMSG_ALIGN(sizeof(struct rtmsg))
3031                + nla_total_size(16) /* RTA_SRC */
3032                + nla_total_size(16) /* RTA_DST */
3033                + nla_total_size(16) /* RTA_GATEWAY */
3034                + nla_total_size(16) /* RTA_PREFSRC */
3035                + nla_total_size(4) /* RTA_TABLE */
3036                + nla_total_size(4) /* RTA_IIF */
3037                + nla_total_size(4) /* RTA_OIF */
3038                + nla_total_size(4) /* RTA_PRIORITY */
3039                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3040                + nla_total_size(sizeof(struct rta_cacheinfo))
3041                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3042                + nla_total_size(1) /* RTA_PREF */
3043                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3044 }
3045
3046 static int rt6_fill_node(struct net *net,
3047                          struct sk_buff *skb, struct rt6_info *rt,
3048                          struct in6_addr *dst, struct in6_addr *src,
3049                          int iif, int type, u32 portid, u32 seq,
3050                          int prefix, int nowait, unsigned int flags)
3051 {
3052         u32 metrics[RTAX_MAX];
3053         struct rtmsg *rtm;
3054         struct nlmsghdr *nlh;
3055         long expires;
3056         u32 table;
3057
3058         if (prefix) {   /* user wants prefix routes only */
3059                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3060                         /* success since this is not a prefix route */
3061                         return 1;
3062                 }
3063         }
3064
3065         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3066         if (!nlh)
3067                 return -EMSGSIZE;
3068
3069         rtm = nlmsg_data(nlh);
3070         rtm->rtm_family = AF_INET6;
3071         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3072         rtm->rtm_src_len = rt->rt6i_src.plen;
3073         rtm->rtm_tos = 0;
3074         if (rt->rt6i_table)
3075                 table = rt->rt6i_table->tb6_id;
3076         else
3077                 table = RT6_TABLE_UNSPEC;
3078         rtm->rtm_table = table;
3079         if (nla_put_u32(skb, RTA_TABLE, table))
3080                 goto nla_put_failure;
3081         if (rt->rt6i_flags & RTF_REJECT) {
3082                 switch (rt->dst.error) {
3083                 case -EINVAL:
3084                         rtm->rtm_type = RTN_BLACKHOLE;
3085                         break;
3086                 case -EACCES:
3087                         rtm->rtm_type = RTN_PROHIBIT;
3088                         break;
3089                 case -EAGAIN:
3090                         rtm->rtm_type = RTN_THROW;
3091                         break;
3092                 default:
3093                         rtm->rtm_type = RTN_UNREACHABLE;
3094                         break;
3095                 }
3096         }
3097         else if (rt->rt6i_flags & RTF_LOCAL)
3098                 rtm->rtm_type = RTN_LOCAL;
3099         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3100                 rtm->rtm_type = RTN_LOCAL;
3101         else
3102                 rtm->rtm_type = RTN_UNICAST;
3103         rtm->rtm_flags = 0;
3104         if (!netif_carrier_ok(rt->dst.dev)) {
3105                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3106                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3107                         rtm->rtm_flags |= RTNH_F_DEAD;
3108         }
3109         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3110         rtm->rtm_protocol = rt->rt6i_protocol;
3111         if (rt->rt6i_flags & RTF_DYNAMIC)
3112                 rtm->rtm_protocol = RTPROT_REDIRECT;
3113         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3114                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3115                         rtm->rtm_protocol = RTPROT_RA;
3116                 else
3117                         rtm->rtm_protocol = RTPROT_KERNEL;
3118         }
3119
3120         if (rt->rt6i_flags & RTF_CACHE)
3121                 rtm->rtm_flags |= RTM_F_CLONED;
3122
3123         if (dst) {
3124                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3125                         goto nla_put_failure;
3126                 rtm->rtm_dst_len = 128;
3127         } else if (rtm->rtm_dst_len)
3128                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3129                         goto nla_put_failure;
3130 #ifdef CONFIG_IPV6_SUBTREES
3131         if (src) {
3132                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3133                         goto nla_put_failure;
3134                 rtm->rtm_src_len = 128;
3135         } else if (rtm->rtm_src_len &&
3136                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3137                 goto nla_put_failure;
3138 #endif
3139         if (iif) {
3140 #ifdef CONFIG_IPV6_MROUTE
3141                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3142                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3143                         if (err <= 0) {
3144                                 if (!nowait) {
3145                                         if (err == 0)
3146                                                 return 0;
3147                                         goto nla_put_failure;
3148                                 } else {
3149                                         if (err == -EMSGSIZE)
3150                                                 goto nla_put_failure;
3151                                 }
3152                         }
3153                 } else
3154 #endif
3155                         if (nla_put_u32(skb, RTA_IIF, iif))
3156                                 goto nla_put_failure;
3157         } else if (dst) {
3158                 struct in6_addr saddr_buf;
3159                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3160                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3161                         goto nla_put_failure;
3162         }
3163
3164         if (rt->rt6i_prefsrc.plen) {
3165                 struct in6_addr saddr_buf;
3166                 saddr_buf = rt->rt6i_prefsrc.addr;
3167                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3168                         goto nla_put_failure;
3169         }
3170
3171         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3172         if (rt->rt6i_pmtu)
3173                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3174         if (rtnetlink_put_metrics(skb, metrics) < 0)
3175                 goto nla_put_failure;
3176
3177         if (rt->rt6i_flags & RTF_GATEWAY) {
3178                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3179                         goto nla_put_failure;
3180         }
3181
3182         if (rt->dst.dev &&
3183             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3184                 goto nla_put_failure;
3185         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3186                 goto nla_put_failure;
3187
3188         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3189
3190         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3191                 goto nla_put_failure;
3192
3193         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3194                 goto nla_put_failure;
3195
3196         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3197
3198         nlmsg_end(skb, nlh);
3199         return 0;
3200
3201 nla_put_failure:
3202         nlmsg_cancel(skb, nlh);
3203         return -EMSGSIZE;
3204 }
3205
3206 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3207 {
3208         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3209         int prefix;
3210
3211         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3212                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3213                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3214         } else
3215                 prefix = 0;
3216
3217         return rt6_fill_node(arg->net,
3218                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3219                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3220                      prefix, 0, NLM_F_MULTI);
3221 }
3222
3223 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3224 {
3225         struct net *net = sock_net(in_skb->sk);
3226         struct nlattr *tb[RTA_MAX+1];
3227         struct rt6_info *rt;
3228         struct sk_buff *skb;
3229         struct rtmsg *rtm;
3230         struct flowi6 fl6;
3231         int err, iif = 0, oif = 0;
3232
3233         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3234         if (err < 0)
3235                 goto errout;
3236
3237         err = -EINVAL;
3238         memset(&fl6, 0, sizeof(fl6));
3239
3240         if (tb[RTA_SRC]) {
3241                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3242                         goto errout;
3243
3244                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3245         }
3246
3247         if (tb[RTA_DST]) {
3248                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3249                         goto errout;
3250
3251                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3252         }
3253
3254         if (tb[RTA_IIF])
3255                 iif = nla_get_u32(tb[RTA_IIF]);
3256
3257         if (tb[RTA_OIF])
3258                 oif = nla_get_u32(tb[RTA_OIF]);
3259
3260         if (tb[RTA_MARK])
3261                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3262
3263         if (iif) {
3264                 struct net_device *dev;
3265                 int flags = 0;
3266
3267                 dev = __dev_get_by_index(net, iif);
3268                 if (!dev) {
3269                         err = -ENODEV;
3270                         goto errout;
3271                 }
3272
3273                 fl6.flowi6_iif = iif;
3274
3275                 if (!ipv6_addr_any(&fl6.saddr))
3276                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3277
3278                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3279                                                                flags);
3280         } else {
3281                 fl6.flowi6_oif = oif;
3282
3283                 if (netif_index_is_l3_master(net, oif)) {
3284                         fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3285                                            FLOWI_FLAG_SKIP_NH_OIF;
3286                 }
3287
3288                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3289         }
3290
3291         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3292         if (!skb) {
3293                 ip6_rt_put(rt);
3294                 err = -ENOBUFS;
3295                 goto errout;
3296         }
3297
3298         /* Reserve room for dummy headers, this skb can pass
3299            through good chunk of routing engine.
3300          */
3301         skb_reset_mac_header(skb);
3302         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3303
3304         skb_dst_set(skb, &rt->dst);
3305
3306         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3307                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3308                             nlh->nlmsg_seq, 0, 0, 0);
3309         if (err < 0) {
3310                 kfree_skb(skb);
3311                 goto errout;
3312         }
3313
3314         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3315 errout:
3316         return err;
3317 }
3318
3319 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3320                      unsigned int nlm_flags)
3321 {
3322         struct sk_buff *skb;
3323         struct net *net = info->nl_net;
3324         u32 seq;
3325         int err;
3326
3327         err = -ENOBUFS;
3328         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3329
3330         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3331         if (!skb)
3332                 goto errout;
3333
3334         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3335                                 event, info->portid, seq, 0, 0, nlm_flags);
3336         if (err < 0) {
3337                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3338                 WARN_ON(err == -EMSGSIZE);
3339                 kfree_skb(skb);
3340                 goto errout;
3341         }
3342         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3343                     info->nlh, gfp_any());
3344         return;
3345 errout:
3346         if (err < 0)
3347                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3348 }
3349
3350 static int ip6_route_dev_notify(struct notifier_block *this,
3351                                 unsigned long event, void *ptr)
3352 {
3353         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3354         struct net *net = dev_net(dev);
3355
3356         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3357                 net->ipv6.ip6_null_entry->dst.dev = dev;
3358                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3359 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3360                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3361                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3362                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3363                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3364 #endif
3365         }
3366
3367         return NOTIFY_OK;
3368 }
3369
3370 /*
3371  *      /proc
3372  */
3373
3374 #ifdef CONFIG_PROC_FS
3375
3376 static const struct file_operations ipv6_route_proc_fops = {
3377         .owner          = THIS_MODULE,
3378         .open           = ipv6_route_open,
3379         .read           = seq_read,
3380         .llseek         = seq_lseek,
3381         .release        = seq_release_net,
3382 };
3383
3384 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3385 {
3386         struct net *net = (struct net *)seq->private;
3387         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3388                    net->ipv6.rt6_stats->fib_nodes,
3389                    net->ipv6.rt6_stats->fib_route_nodes,
3390                    net->ipv6.rt6_stats->fib_rt_alloc,
3391                    net->ipv6.rt6_stats->fib_rt_entries,
3392                    net->ipv6.rt6_stats->fib_rt_cache,
3393                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3394                    net->ipv6.rt6_stats->fib_discarded_routes);
3395
3396         return 0;
3397 }
3398
3399 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3400 {
3401         return single_open_net(inode, file, rt6_stats_seq_show);
3402 }
3403
3404 static const struct file_operations rt6_stats_seq_fops = {
3405         .owner   = THIS_MODULE,
3406         .open    = rt6_stats_seq_open,
3407         .read    = seq_read,
3408         .llseek  = seq_lseek,
3409         .release = single_release_net,
3410 };
3411 #endif  /* CONFIG_PROC_FS */
3412
3413 #ifdef CONFIG_SYSCTL
3414
3415 static
3416 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3417                               void __user *buffer, size_t *lenp, loff_t *ppos)
3418 {
3419         struct net *net;
3420         int delay;
3421         if (!write)
3422                 return -EINVAL;
3423
3424         net = (struct net *)ctl->extra1;
3425         delay = net->ipv6.sysctl.flush_delay;
3426         proc_dointvec(ctl, write, buffer, lenp, ppos);
3427         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3428         return 0;
3429 }
3430
3431 struct ctl_table ipv6_route_table_template[] = {
3432         {
3433                 .procname       =       "flush",
3434                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3435                 .maxlen         =       sizeof(int),
3436                 .mode           =       0200,
3437                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3438         },
3439         {
3440                 .procname       =       "gc_thresh",
3441                 .data           =       &ip6_dst_ops_template.gc_thresh,
3442                 .maxlen         =       sizeof(int),
3443                 .mode           =       0644,
3444                 .proc_handler   =       proc_dointvec,
3445         },
3446         {
3447                 .procname       =       "max_size",
3448                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3449                 .maxlen         =       sizeof(int),
3450                 .mode           =       0644,
3451                 .proc_handler   =       proc_dointvec,
3452         },
3453         {
3454                 .procname       =       "gc_min_interval",
3455                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3456                 .maxlen         =       sizeof(int),
3457                 .mode           =       0644,
3458                 .proc_handler   =       proc_dointvec_jiffies,
3459         },
3460         {
3461                 .procname       =       "gc_timeout",
3462                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3463                 .maxlen         =       sizeof(int),
3464                 .mode           =       0644,
3465                 .proc_handler   =       proc_dointvec_jiffies,
3466         },
3467         {
3468                 .procname       =       "gc_interval",
3469                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3470                 .maxlen         =       sizeof(int),
3471                 .mode           =       0644,
3472                 .proc_handler   =       proc_dointvec_jiffies,
3473         },
3474         {
3475                 .procname       =       "gc_elasticity",
3476                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3477                 .maxlen         =       sizeof(int),
3478                 .mode           =       0644,
3479                 .proc_handler   =       proc_dointvec,
3480         },
3481         {
3482                 .procname       =       "mtu_expires",
3483                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3484                 .maxlen         =       sizeof(int),
3485                 .mode           =       0644,
3486                 .proc_handler   =       proc_dointvec_jiffies,
3487         },
3488         {
3489                 .procname       =       "min_adv_mss",
3490                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3491                 .maxlen         =       sizeof(int),
3492                 .mode           =       0644,
3493                 .proc_handler   =       proc_dointvec,
3494         },
3495         {
3496                 .procname       =       "gc_min_interval_ms",
3497                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3498                 .maxlen         =       sizeof(int),
3499                 .mode           =       0644,
3500                 .proc_handler   =       proc_dointvec_ms_jiffies,
3501         },
3502         { }
3503 };
3504
3505 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3506 {
3507         struct ctl_table *table;
3508
3509         table = kmemdup(ipv6_route_table_template,
3510                         sizeof(ipv6_route_table_template),
3511                         GFP_KERNEL);
3512
3513         if (table) {
3514                 table[0].data = &net->ipv6.sysctl.flush_delay;
3515                 table[0].extra1 = net;
3516                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3517                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3518                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3519                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3520                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3521                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3522                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3523                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3524                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3525
3526                 /* Don't export sysctls to unprivileged users */
3527                 if (net->user_ns != &init_user_ns)
3528                         table[0].procname = NULL;
3529         }
3530
3531         return table;
3532 }
3533 #endif
3534
3535 static int __net_init ip6_route_net_init(struct net *net)
3536 {
3537         int ret = -ENOMEM;
3538
3539         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3540                sizeof(net->ipv6.ip6_dst_ops));
3541
3542         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3543                 goto out_ip6_dst_ops;
3544
3545         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3546                                            sizeof(*net->ipv6.ip6_null_entry),
3547                                            GFP_KERNEL);
3548         if (!net->ipv6.ip6_null_entry)
3549                 goto out_ip6_dst_entries;
3550         net->ipv6.ip6_null_entry->dst.path =
3551                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3552         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3553         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3554                          ip6_template_metrics, true);
3555
3556 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3557         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3558                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3559                                                GFP_KERNEL);
3560         if (!net->ipv6.ip6_prohibit_entry)
3561                 goto out_ip6_null_entry;
3562         net->ipv6.ip6_prohibit_entry->dst.path =
3563                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3564         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3565         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3566                          ip6_template_metrics, true);
3567
3568         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3569                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3570                                                GFP_KERNEL);
3571         if (!net->ipv6.ip6_blk_hole_entry)
3572                 goto out_ip6_prohibit_entry;
3573         net->ipv6.ip6_blk_hole_entry->dst.path =
3574                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3575         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3576         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3577                          ip6_template_metrics, true);
3578 #endif
3579
3580         net->ipv6.sysctl.flush_delay = 0;
3581         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3582         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3583         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3584         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3585         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3586         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3587         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3588
3589         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3590
3591         ret = 0;
3592 out:
3593         return ret;
3594
3595 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3596 out_ip6_prohibit_entry:
3597         kfree(net->ipv6.ip6_prohibit_entry);
3598 out_ip6_null_entry:
3599         kfree(net->ipv6.ip6_null_entry);
3600 #endif
3601 out_ip6_dst_entries:
3602         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3603 out_ip6_dst_ops:
3604         goto out;
3605 }
3606
3607 static void __net_exit ip6_route_net_exit(struct net *net)
3608 {
3609         kfree(net->ipv6.ip6_null_entry);
3610 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3611         kfree(net->ipv6.ip6_prohibit_entry);
3612         kfree(net->ipv6.ip6_blk_hole_entry);
3613 #endif
3614         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3615 }
3616
3617 static int __net_init ip6_route_net_init_late(struct net *net)
3618 {
3619 #ifdef CONFIG_PROC_FS
3620         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3621         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3622 #endif
3623         return 0;
3624 }
3625
3626 static void __net_exit ip6_route_net_exit_late(struct net *net)
3627 {
3628 #ifdef CONFIG_PROC_FS
3629         remove_proc_entry("ipv6_route", net->proc_net);
3630         remove_proc_entry("rt6_stats", net->proc_net);
3631 #endif
3632 }
3633
3634 static struct pernet_operations ip6_route_net_ops = {
3635         .init = ip6_route_net_init,
3636         .exit = ip6_route_net_exit,
3637 };
3638
3639 static int __net_init ipv6_inetpeer_init(struct net *net)
3640 {
3641         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3642
3643         if (!bp)
3644                 return -ENOMEM;
3645         inet_peer_base_init(bp);
3646         net->ipv6.peers = bp;
3647         return 0;
3648 }
3649
3650 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3651 {
3652         struct inet_peer_base *bp = net->ipv6.peers;
3653
3654         net->ipv6.peers = NULL;
3655         inetpeer_invalidate_tree(bp);
3656         kfree(bp);
3657 }
3658
3659 static struct pernet_operations ipv6_inetpeer_ops = {
3660         .init   =       ipv6_inetpeer_init,
3661         .exit   =       ipv6_inetpeer_exit,
3662 };
3663
3664 static struct pernet_operations ip6_route_net_late_ops = {
3665         .init = ip6_route_net_init_late,
3666         .exit = ip6_route_net_exit_late,
3667 };
3668
3669 static struct notifier_block ip6_route_dev_notifier = {
3670         .notifier_call = ip6_route_dev_notify,
3671         .priority = 0,
3672 };
3673
3674 int __init ip6_route_init(void)
3675 {
3676         int ret;
3677         int cpu;
3678
3679         ret = -ENOMEM;
3680         ip6_dst_ops_template.kmem_cachep =
3681                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3682                                   SLAB_HWCACHE_ALIGN, NULL);
3683         if (!ip6_dst_ops_template.kmem_cachep)
3684                 goto out;
3685
3686         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3687         if (ret)
3688                 goto out_kmem_cache;
3689
3690         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3691         if (ret)
3692                 goto out_dst_entries;
3693
3694         ret = register_pernet_subsys(&ip6_route_net_ops);
3695         if (ret)
3696                 goto out_register_inetpeer;
3697
3698         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3699
3700         /* Registering of the loopback is done before this portion of code,
3701          * the loopback reference in rt6_info will not be taken, do it
3702          * manually for init_net */
3703         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3704         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3705   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3706         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3707         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3708         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3709         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3710   #endif
3711         ret = fib6_init();
3712         if (ret)
3713                 goto out_register_subsys;
3714
3715         ret = xfrm6_init();
3716         if (ret)
3717                 goto out_fib6_init;
3718
3719         ret = fib6_rules_init();
3720         if (ret)
3721                 goto xfrm6_init;
3722
3723         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3724         if (ret)
3725                 goto fib6_rules_init;
3726
3727         ret = -ENOBUFS;
3728         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3729             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3730             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3731                 goto out_register_late_subsys;
3732
3733         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3734         if (ret)
3735                 goto out_register_late_subsys;
3736
3737         for_each_possible_cpu(cpu) {
3738                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3739
3740                 INIT_LIST_HEAD(&ul->head);
3741                 spin_lock_init(&ul->lock);
3742         }
3743
3744 out:
3745         return ret;
3746
3747 out_register_late_subsys:
3748         unregister_pernet_subsys(&ip6_route_net_late_ops);
3749 fib6_rules_init:
3750         fib6_rules_cleanup();
3751 xfrm6_init:
3752         xfrm6_fini();
3753 out_fib6_init:
3754         fib6_gc_cleanup();
3755 out_register_subsys:
3756         unregister_pernet_subsys(&ip6_route_net_ops);
3757 out_register_inetpeer:
3758         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3759 out_dst_entries:
3760         dst_entries_destroy(&ip6_dst_blackhole_ops);
3761 out_kmem_cache:
3762         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3763         goto out;
3764 }
3765
3766 void ip6_route_cleanup(void)
3767 {
3768         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3769         unregister_pernet_subsys(&ip6_route_net_late_ops);
3770         fib6_rules_cleanup();
3771         xfrm6_fini();
3772         fib6_gc_cleanup();
3773         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3774         unregister_pernet_subsys(&ip6_route_net_ops);
3775         dst_entries_destroy(&ip6_dst_blackhole_ops);
3776         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3777 }