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 #include <trace/events/fib6.h>
  66
  67 #include <asm/uaccess.h>
  68
  69 #ifdef CONFIG_SYSCTL
  70 #include <linux/sysctl.h>
  71 #endif
  72
  73 enum rt6_nud_state {
  74         RT6_NUD_FAIL_HARD = -3,
  75         RT6_NUD_FAIL_PROBE = -2,
  76         RT6_NUD_FAIL_DO_RR = -1,
  77         RT6_NUD_SUCCEED = 1
  78 };
  79
  80 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
  81 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  82 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  83 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  84 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  85 static void             ip6_dst_destroy(struct dst_entry *);
  86 static void             ip6_dst_ifdown(struct dst_entry *,
  87                                        struct net_device *dev, int how);
  88 static int               ip6_dst_gc(struct dst_ops *ops);
  89
  90 static int              ip6_pkt_discard(struct sk_buff *skb);
  91 static int              ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  92 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  93 static int              ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  94 static void             ip6_link_failure(struct sk_buff *skb);
  95 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  96                                            struct sk_buff *skb, u32 mtu);
  97 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  98                                         struct sk_buff *skb);
  99 static void             rt6_dst_from_metrics_check(struct rt6_info *rt);
 100 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
 101
 102 #ifdef CONFIG_IPV6_ROUTE_INFO
 103 static struct rt6_info *rt6_add_route_info(struct net *net,
 104                                            const struct in6_addr *prefix, int prefixlen,
 105                                            const struct in6_addr *gwaddr, int ifindex,
 106                                            unsigned int pref);
 107 static struct rt6_info *rt6_get_route_info(struct net *net,
 108                                            const struct in6_addr *prefix, int prefixlen,
 109                                            const struct in6_addr *gwaddr, int ifindex);
 110 #endif
 111
 112 struct uncached_list {
 113         spinlock_t              lock;
 114         struct list_head        head;
 115 };
 116
 117 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 118
 119 static void rt6_uncached_list_add(struct rt6_info *rt)
 120 {
 121         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 122
 123         rt->dst.flags |= DST_NOCACHE;
 124         rt->rt6i_uncached_list = ul;
 125
 126         spin_lock_bh(&ul->lock);
 127         list_add_tail(&rt->rt6i_uncached, &ul->head);
 128         spin_unlock_bh(&ul->lock);
 129 }
 130
 131 static void rt6_uncached_list_del(struct rt6_info *rt)
 132 {
 133         if (!list_empty(&rt->rt6i_uncached)) {
 134                 struct uncached_list *ul = rt->rt6i_uncached_list;
 135
 136                 spin_lock_bh(&ul->lock);
 137                 list_del(&rt->rt6i_uncached);
 138                 spin_unlock_bh(&ul->lock);
 139         }
 140 }
 141
 142 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
 143 {
 144         struct net_device *loopback_dev = net->loopback_dev;
 145         int cpu;
 146
 147         if (dev == loopback_dev)
 148                 return;
 149
 150         for_each_possible_cpu(cpu) {
 151                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 152                 struct rt6_info *rt;
 153
 154                 spin_lock_bh(&ul->lock);
 155                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 156                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 157                         struct net_device *rt_dev = rt->dst.dev;
 158
 159                         if (rt_idev->dev == dev) {
 160                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 161                                 in6_dev_put(rt_idev);
 162                         }
 163
 164                         if (rt_dev == dev) {
 165                                 rt->dst.dev = loopback_dev;
 166                                 dev_hold(rt->dst.dev);
 167                                 dev_put(rt_dev);
 168                         }
 169                 }
 170                 spin_unlock_bh(&ul->lock);
 171         }
 172 }
 173
 174 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 175 {
 176         return dst_metrics_write_ptr(rt->dst.from);
 177 }
 178
 179 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 180 {
 181         struct rt6_info *rt = (struct rt6_info *)dst;
 182
 183         if (rt->rt6i_flags & RTF_PCPU)
 184                 return rt6_pcpu_cow_metrics(rt);
 185         else if (rt->rt6i_flags & RTF_CACHE)
 186                 return NULL;
 187         else
 188                 return dst_cow_metrics_generic(dst, old);
 189 }
 190
 191 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 192                                              struct sk_buff *skb,
 193                                              const void *daddr)
 194 {
 195         struct in6_addr *p = &rt->rt6i_gateway;
 196
 197         if (!ipv6_addr_any(p))
 198                 return (const void *) p;
 199         else if (skb)
 200                 return &ipv6_hdr(skb)->daddr;
 201         return daddr;
 202 }
 203
 204 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 205                                           struct sk_buff *skb,
 206                                           const void *daddr)
 207 {
 208         struct rt6_info *rt = (struct rt6_info *) dst;
 209         struct neighbour *n;
 210
 211         daddr = choose_neigh_daddr(rt, skb, daddr);
 212         n = __ipv6_neigh_lookup(dst->dev, daddr);
 213         if (n)
 214                 return n;
 215         return neigh_create(&nd_tbl, daddr, dst->dev);
 216 }
 217
 218 static struct dst_ops ip6_dst_ops_template = {
 219         .family                 =       AF_INET6,
 220         .gc                     =       ip6_dst_gc,
 221         .gc_thresh              =       1024,
 222         .check                  =       ip6_dst_check,
 223         .default_advmss         =       ip6_default_advmss,
 224         .mtu                    =       ip6_mtu,
 225         .cow_metrics            =       ipv6_cow_metrics,
 226         .destroy                =       ip6_dst_destroy,
 227         .ifdown                 =       ip6_dst_ifdown,
 228         .negative_advice        =       ip6_negative_advice,
 229         .link_failure           =       ip6_link_failure,
 230         .update_pmtu            =       ip6_rt_update_pmtu,
 231         .redirect               =       rt6_do_redirect,
 232         .local_out              =       __ip6_local_out,
 233         .neigh_lookup           =       ip6_neigh_lookup,
 234 };
 235
 236 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 237 {
 238         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 239
 240         return mtu ? : dst->dev->mtu;
 241 }
 242
 243 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 244                                          struct sk_buff *skb, u32 mtu)
 245 {
 246 }
 247
 248 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 249                                       struct sk_buff *skb)
 250 {
 251 }
 252
 253 static struct dst_ops ip6_dst_blackhole_ops = {
 254         .family                 =       AF_INET6,
 255         .destroy                =       ip6_dst_destroy,
 256         .check                  =       ip6_dst_check,
 257         .mtu                    =       ip6_blackhole_mtu,
 258         .default_advmss         =       ip6_default_advmss,
 259         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 260         .redirect               =       ip6_rt_blackhole_redirect,
 261         .cow_metrics            =       dst_cow_metrics_generic,
 262         .neigh_lookup           =       ip6_neigh_lookup,
 263 };
 264
 265 static const u32 ip6_template_metrics[RTAX_MAX] = {
 266         [RTAX_HOPLIMIT - 1] = 0,
 267 };
 268
 269 static const struct rt6_info ip6_null_entry_template = {
 270         .dst = {
 271                 .__refcnt       = ATOMIC_INIT(1),
 272                 .__use          = 1,
 273                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 274                 .error          = -ENETUNREACH,
 275                 .input          = ip6_pkt_discard,
 276                 .output         = ip6_pkt_discard_out,
 277         },
 278         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 279         .rt6i_protocol  = RTPROT_KERNEL,
 280         .rt6i_metric    = ~(u32) 0,
 281         .rt6i_ref       = ATOMIC_INIT(1),
 282 };
 283
 284 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 285
 286 static const struct rt6_info ip6_prohibit_entry_template = {
 287         .dst = {
 288                 .__refcnt       = ATOMIC_INIT(1),
 289                 .__use          = 1,
 290                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 291                 .error          = -EACCES,
 292                 .input          = ip6_pkt_prohibit,
 293                 .output         = ip6_pkt_prohibit_out,
 294         },
 295         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 296         .rt6i_protocol  = RTPROT_KERNEL,
 297         .rt6i_metric    = ~(u32) 0,
 298         .rt6i_ref       = ATOMIC_INIT(1),
 299 };
 300
 301 static const struct rt6_info ip6_blk_hole_entry_template = {
 302         .dst = {
 303                 .__refcnt       = ATOMIC_INIT(1),
 304                 .__use          = 1,
 305                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 306                 .error          = -EINVAL,
 307                 .input          = dst_discard,
 308                 .output         = dst_discard_out,
 309         },
 310         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 311         .rt6i_protocol  = RTPROT_KERNEL,
 312         .rt6i_metric    = ~(u32) 0,
 313         .rt6i_ref       = ATOMIC_INIT(1),
 314 };
 315
 316 #endif
 317
 318 static void rt6_info_init(struct rt6_info *rt)
 319 {
 320         struct dst_entry *dst = &rt->dst;
 321
 322         memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 323         INIT_LIST_HEAD(&rt->rt6i_siblings);
 324         INIT_LIST_HEAD(&rt->rt6i_uncached);
 325 }
 326
 327 /* allocate dst with ip6_dst_ops */
 328 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 329                                         struct net_device *dev,
 330                                         int flags)
 331 {
 332         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 333                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 334
 335         if (rt)
 336                 rt6_info_init(rt);
 337
 338         return rt;
 339 }
 340
 341 struct rt6_info *ip6_dst_alloc(struct net *net,
 342                                struct net_device *dev,
 343                                int flags)
 344 {
 345         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 346
 347         if (rt) {
 348                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 349                 if (rt->rt6i_pcpu) {
 350                         int cpu;
 351
 352                         for_each_possible_cpu(cpu) {
 353                                 struct rt6_info **p;
 354
 355                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 356                                 /* no one shares rt */
 357                                 *p =  NULL;
 358                         }
 359                 } else {
 360                         dst_destroy((struct dst_entry *)rt);
 361                         return NULL;
 362                 }
 363         }
 364
 365         return rt;
 366 }
 367 EXPORT_SYMBOL(ip6_dst_alloc);
 368
 369 static void ip6_dst_destroy(struct dst_entry *dst)
 370 {
 371         struct rt6_info *rt = (struct rt6_info *)dst;
 372         struct dst_entry *from = dst->from;
 373         struct inet6_dev *idev;
 374
 375         dst_destroy_metrics_generic(dst);
 376         free_percpu(rt->rt6i_pcpu);
 377         rt6_uncached_list_del(rt);
 378
 379         idev = rt->rt6i_idev;
 380         if (idev) {
 381                 rt->rt6i_idev = NULL;
 382                 in6_dev_put(idev);
 383         }
 384
 385         dst->from = NULL;
 386         dst_release(from);
 387 }
 388
 389 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 390                            int how)
 391 {
 392         struct rt6_info *rt = (struct rt6_info *)dst;
 393         struct inet6_dev *idev = rt->rt6i_idev;
 394         struct net_device *loopback_dev =
 395                 dev_net(dev)->loopback_dev;
 396
 397         if (dev != loopback_dev) {
 398                 if (idev && idev->dev == dev) {
 399                         struct inet6_dev *loopback_idev =
 400                                 in6_dev_get(loopback_dev);
 401                         if (loopback_idev) {
 402                                 rt->rt6i_idev = loopback_idev;
 403                                 in6_dev_put(idev);
 404                         }
 405                 }
 406         }
 407 }
 408
 409 static bool __rt6_check_expired(const struct rt6_info *rt)
 410 {
 411         if (rt->rt6i_flags & RTF_EXPIRES)
 412                 return time_after(jiffies, rt->dst.expires);
 413         else
 414                 return false;
 415 }
 416
 417 static bool rt6_check_expired(const struct rt6_info *rt)
 418 {
 419         if (rt->rt6i_flags & RTF_EXPIRES) {
 420                 if (time_after(jiffies, rt->dst.expires))
 421                         return true;
 422         } else if (rt->dst.from) {
 423                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 424         }
 425         return false;
 426 }
 427
 428 /* Multipath route selection:
 429  *   Hash based function using packet header and flowlabel.
 430  * Adapted from fib_info_hashfn()
 431  */
 432 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 433                                const struct flowi6 *fl6)
 434 {
 435         return get_hash_from_flowi6(fl6) % candidate_count;
 436 }
 437
 438 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 439                                              struct flowi6 *fl6, int oif,
 440                                              int strict)
 441 {
 442         struct rt6_info *sibling, *next_sibling;
 443         int route_choosen;
 444
 445         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 446         /* Don't change the route, if route_choosen == 0
 447          * (siblings does not include ourself)
 448          */
 449         if (route_choosen)
 450                 list_for_each_entry_safe(sibling, next_sibling,
 451                                 &match->rt6i_siblings, rt6i_siblings) {
 452                         route_choosen--;
 453                         if (route_choosen == 0) {
 454                                 if (rt6_score_route(sibling, oif, strict) < 0)
 455                                         break;
 456                                 match = sibling;
 457                                 break;
 458                         }
 459                 }
 460         return match;
 461 }
 462
 463 /*
 464  *      Route lookup. Any table->tb6_lock is implied.
 465  */
 466
 467 static inline struct rt6_info *rt6_device_match(struct net *net,
 468                                                     struct rt6_info *rt,
 469                                                     const struct in6_addr *saddr,
 470                                                     int oif,
 471                                                     int flags)
 472 {
 473         struct rt6_info *local = NULL;
 474         struct rt6_info *sprt;
 475
 476         if (!oif && ipv6_addr_any(saddr))
 477                 goto out;
 478
 479         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 480                 struct net_device *dev = sprt->dst.dev;
 481
 482                 if (oif) {
 483                         if (dev->ifindex == oif)
 484                                 return sprt;
 485                         if (dev->flags & IFF_LOOPBACK) {
 486                                 if (!sprt->rt6i_idev ||
 487                                     sprt->rt6i_idev->dev->ifindex != oif) {
 488                                         if (flags & RT6_LOOKUP_F_IFACE)
 489                                                 continue;
 490                                         if (local &&
 491                                             local->rt6i_idev->dev->ifindex == oif)
 492                                                 continue;
 493                                 }
 494                                 local = sprt;
 495                         }
 496                 } else {
 497                         if (ipv6_chk_addr(net, saddr, dev,
 498                                           flags & RT6_LOOKUP_F_IFACE))
 499                                 return sprt;
 500                 }
 501         }
 502
 503         if (oif) {
 504                 if (local)
 505                         return local;
 506
 507                 if (flags & RT6_LOOKUP_F_IFACE)
 508                         return net->ipv6.ip6_null_entry;
 509         }
 510 out:
 511         return rt;
 512 }
 513
 514 #ifdef CONFIG_IPV6_ROUTER_PREF
 515 struct __rt6_probe_work {
 516         struct work_struct work;
 517         struct in6_addr target;
 518         struct net_device *dev;
 519 };
 520
 521 static void rt6_probe_deferred(struct work_struct *w)
 522 {
 523         struct in6_addr mcaddr;
 524         struct __rt6_probe_work *work =
 525                 container_of(w, struct __rt6_probe_work, work);
 526
 527         addrconf_addr_solict_mult(&work->target, &mcaddr);
 528         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
 529         dev_put(work->dev);
 530         kfree(work);
 531 }
 532
 533 static void rt6_probe(struct rt6_info *rt)
 534 {
 535         struct __rt6_probe_work *work;
 536         struct neighbour *neigh;
 537         /*
 538          * Okay, this does not seem to be appropriate
 539          * for now, however, we need to check if it
 540          * is really so; aka Router Reachability Probing.
 541          *
 542          * Router Reachability Probe MUST be rate-limited
 543          * to no more than one per minute.
 544          */
 545         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 546                 return;
 547         rcu_read_lock_bh();
 548         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 549         if (neigh) {
 550                 if (neigh->nud_state & NUD_VALID)
 551                         goto out;
 552
 553                 work = NULL;
 554                 write_lock(&neigh->lock);
 555                 if (!(neigh->nud_state & NUD_VALID) &&
 556                     time_after(jiffies,
 557                                neigh->updated +
 558                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 559                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 560                         if (work)
 561                                 __neigh_set_probe_once(neigh);
 562                 }
 563                 write_unlock(&neigh->lock);
 564         } else {
 565                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 566         }
 567
 568         if (work) {
 569                 INIT_WORK(&work->work, rt6_probe_deferred);
 570                 work->target = rt->rt6i_gateway;
 571                 dev_hold(rt->dst.dev);
 572                 work->dev = rt->dst.dev;
 573                 schedule_work(&work->work);
 574         }
 575
 576 out:
 577         rcu_read_unlock_bh();
 578 }
 579 #else
 580 static inline void rt6_probe(struct rt6_info *rt)
 581 {
 582 }
 583 #endif
 584
 585 /*
 586  * Default Router Selection (RFC 2461 6.3.6)
 587  */
 588 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 589 {
 590         struct net_device *dev = rt->dst.dev;
 591         if (!oif || dev->ifindex == oif)
 592                 return 2;
 593         if ((dev->flags & IFF_LOOPBACK) &&
 594             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 595                 return 1;
 596         return 0;
 597 }
 598
 599 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 600 {
 601         struct neighbour *neigh;
 602         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 603
 604         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 605             !(rt->rt6i_flags & RTF_GATEWAY))
 606                 return RT6_NUD_SUCCEED;
 607
 608         rcu_read_lock_bh();
 609         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 610         if (neigh) {
 611                 read_lock(&neigh->lock);
 612                 if (neigh->nud_state & NUD_VALID)
 613                         ret = RT6_NUD_SUCCEED;
 614 #ifdef CONFIG_IPV6_ROUTER_PREF
 615                 else if (!(neigh->nud_state & NUD_FAILED))
 616                         ret = RT6_NUD_SUCCEED;
 617                 else
 618                         ret = RT6_NUD_FAIL_PROBE;
 619 #endif
 620                 read_unlock(&neigh->lock);
 621         } else {
 622                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 623                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 624         }
 625         rcu_read_unlock_bh();
 626
 627         return ret;
 628 }
 629
 630 static int rt6_score_route(struct rt6_info *rt, int oif,
 631                            int strict)
 632 {
 633         int m;
 634
 635         m = rt6_check_dev(rt, oif);
 636         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 637                 return RT6_NUD_FAIL_HARD;
 638 #ifdef CONFIG_IPV6_ROUTER_PREF
 639         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 640 #endif
 641         if (strict & RT6_LOOKUP_F_REACHABLE) {
 642                 int n = rt6_check_neigh(rt);
 643                 if (n < 0)
 644                         return n;
 645         }
 646         return m;
 647 }
 648
 649 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 650                                    int *mpri, struct rt6_info *match,
 651                                    bool *do_rr)
 652 {
 653         int m;
 654         bool match_do_rr = false;
 655         struct inet6_dev *idev = rt->rt6i_idev;
 656         struct net_device *dev = rt->dst.dev;
 657
 658         if (dev && !netif_carrier_ok(dev) &&
 659             idev->cnf.ignore_routes_with_linkdown)
 660                 goto out;
 661
 662         if (rt6_check_expired(rt))
 663                 goto out;
 664
 665         m = rt6_score_route(rt, oif, strict);
 666         if (m == RT6_NUD_FAIL_DO_RR) {
 667                 match_do_rr = true;
 668                 m = 0; /* lowest valid score */
 669         } else if (m == RT6_NUD_FAIL_HARD) {
 670                 goto out;
 671         }
 672
 673         if (strict & RT6_LOOKUP_F_REACHABLE)
 674                 rt6_probe(rt);
 675
 676         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 677         if (m > *mpri) {
 678                 *do_rr = match_do_rr;
 679                 *mpri = m;
 680                 match = rt;
 681         }
 682 out:
 683         return match;
 684 }
 685
 686 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 687                                      struct rt6_info *rr_head,
 688                                      u32 metric, int oif, int strict,
 689                                      bool *do_rr)
 690 {
 691         struct rt6_info *rt, *match, *cont;
 692         int mpri = -1;
 693
 694         match = NULL;
 695         cont = NULL;
 696         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 697                 if (rt->rt6i_metric != metric) {
 698                         cont = rt;
 699                         break;
 700                 }
 701
 702                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 703         }
 704
 705         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 706                 if (rt->rt6i_metric != metric) {
 707                         cont = rt;
 708                         break;
 709                 }
 710
 711                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 712         }
 713
 714         if (match || !cont)
 715                 return match;
 716
 717         for (rt = cont; rt; rt = rt->dst.rt6_next)
 718                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 719
 720         return match;
 721 }
 722
 723 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 724 {
 725         struct rt6_info *match, *rt0;
 726         struct net *net;
 727         bool do_rr = false;
 728
 729         rt0 = fn->rr_ptr;
 730         if (!rt0)
 731                 fn->rr_ptr = rt0 = fn->leaf;
 732
 733         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 734                              &do_rr);
 735
 736         if (do_rr) {
 737                 struct rt6_info *next = rt0->dst.rt6_next;
 738
 739                 /* no entries matched; do round-robin */
 740                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 741                         next = fn->leaf;
 742
 743                 if (next != rt0)
 744                         fn->rr_ptr = next;
 745         }
 746
 747         net = dev_net(rt0->dst.dev);
 748         return match ? match : net->ipv6.ip6_null_entry;
 749 }
 750
 751 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 752 {
 753         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 754 }
 755
 756 #ifdef CONFIG_IPV6_ROUTE_INFO
 757 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 758                   const struct in6_addr *gwaddr)
 759 {
 760         struct net *net = dev_net(dev);
 761         struct route_info *rinfo = (struct route_info *) opt;
 762         struct in6_addr prefix_buf, *prefix;
 763         unsigned int pref;
 764         unsigned long lifetime;
 765         struct rt6_info *rt;
 766
 767         if (len < sizeof(struct route_info)) {
 768                 return -EINVAL;
 769         }
 770
 771         /* Sanity check for prefix_len and length */
 772         if (rinfo->length > 3) {
 773                 return -EINVAL;
 774         } else if (rinfo->prefix_len > 128) {
 775                 return -EINVAL;
 776         } else if (rinfo->prefix_len > 64) {
 777                 if (rinfo->length < 2) {
 778                         return -EINVAL;
 779                 }
 780         } else if (rinfo->prefix_len > 0) {
 781                 if (rinfo->length < 1) {
 782                         return -EINVAL;
 783                 }
 784         }
 785
 786         pref = rinfo->route_pref;
 787         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 788                 return -EINVAL;
 789
 790         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 791
 792         if (rinfo->length == 3)
 793                 prefix = (struct in6_addr *)rinfo->prefix;
 794         else {
 795                 /* this function is safe */
 796                 ipv6_addr_prefix(&prefix_buf,
 797                                  (struct in6_addr *)rinfo->prefix,
 798                                  rinfo->prefix_len);
 799                 prefix = &prefix_buf;
 800         }
 801
 802         if (rinfo->prefix_len == 0)
 803                 rt = rt6_get_dflt_router(gwaddr, dev);
 804         else
 805                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 806                                         gwaddr, dev->ifindex);
 807
 808         if (rt && !lifetime) {
 809                 ip6_del_rt(rt);
 810                 rt = NULL;
 811         }
 812
 813         if (!rt && lifetime)
 814                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 815                                         pref);
 816         else if (rt)
 817                 rt->rt6i_flags = RTF_ROUTEINFO |
 818                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 819
 820         if (rt) {
 821                 if (!addrconf_finite_timeout(lifetime))
 822                         rt6_clean_expires(rt);
 823                 else
 824                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 825
 826                 ip6_rt_put(rt);
 827         }
 828         return 0;
 829 }
 830 #endif
 831
 832 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 833                                         struct in6_addr *saddr)
 834 {
 835         struct fib6_node *pn;
 836         while (1) {
 837                 if (fn->fn_flags & RTN_TL_ROOT)
 838                         return NULL;
 839                 pn = fn->parent;
 840                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 841                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 842                 else
 843                         fn = pn;
 844                 if (fn->fn_flags & RTN_RTINFO)
 845                         return fn;
 846         }
 847 }
 848
 849 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 850                                              struct fib6_table *table,
 851                                              struct flowi6 *fl6, int flags)
 852 {
 853         struct fib6_node *fn;
 854         struct rt6_info *rt;
 855
 856         read_lock_bh(&table->tb6_lock);
 857         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 858 restart:
 859         rt = fn->leaf;
 860         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 861         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 862                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 863         if (rt == net->ipv6.ip6_null_entry) {
 864                 fn = fib6_backtrack(fn, &fl6->saddr);
 865                 if (fn)
 866                         goto restart;
 867         }
 868         dst_use(&rt->dst, jiffies);
 869         read_unlock_bh(&table->tb6_lock);
 870
 871         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
 872
 873         return rt;
 874
 875 }
 876
 877 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 878                                     int flags)
 879 {
 880         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 881 }
 882 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 883
 884 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 885                             const struct in6_addr *saddr, int oif, int strict)
 886 {
 887         struct flowi6 fl6 = {
 888                 .flowi6_oif = oif,
 889                 .daddr = *daddr,
 890         };
 891         struct dst_entry *dst;
 892         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 893
 894         if (saddr) {
 895                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 896                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 897         }
 898
 899         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 900         if (dst->error == 0)
 901                 return (struct rt6_info *) dst;
 902
 903         dst_release(dst);
 904
 905         return NULL;
 906 }
 907 EXPORT_SYMBOL(rt6_lookup);
 908
 909 /* ip6_ins_rt is called with FREE table->tb6_lock.
 910    It takes new route entry, the addition fails by any reason the
 911    route is freed. In any case, if caller does not hold it, it may
 912    be destroyed.
 913  */
 914
 915 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 916                         struct mx6_config *mxc)
 917 {
 918         int err;
 919         struct fib6_table *table;
 920
 921         table = rt->rt6i_table;
 922         write_lock_bh(&table->tb6_lock);
 923         err = fib6_add(&table->tb6_root, rt, info, mxc);
 924         write_unlock_bh(&table->tb6_lock);
 925
 926         return err;
 927 }
 928
 929 int ip6_ins_rt(struct rt6_info *rt)
 930 {
 931         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 932         struct mx6_config mxc = { .mx = NULL, };
 933
 934         return __ip6_ins_rt(rt, &info, &mxc);
 935 }
 936
 937 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 938                                            const struct in6_addr *daddr,
 939                                            const struct in6_addr *saddr)
 940 {
 941         struct rt6_info *rt;
 942
 943         /*
 944          *      Clone the route.
 945          */
 946
 947         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 948                 ort = (struct rt6_info *)ort->dst.from;
 949
 950         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 951
 952         if (!rt)
 953                 return NULL;
 954
 955         ip6_rt_copy_init(rt, ort);
 956         rt->rt6i_flags |= RTF_CACHE;
 957         rt->rt6i_metric = 0;
 958         rt->dst.flags |= DST_HOST;
 959         rt->rt6i_dst.addr = *daddr;
 960         rt->rt6i_dst.plen = 128;
 961
 962         if (!rt6_is_gw_or_nonexthop(ort)) {
 963                 if (ort->rt6i_dst.plen != 128 &&
 964                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 965                         rt->rt6i_flags |= RTF_ANYCAST;
 966 #ifdef CONFIG_IPV6_SUBTREES
 967                 if (rt->rt6i_src.plen && saddr) {
 968                         rt->rt6i_src.addr = *saddr;
 969                         rt->rt6i_src.plen = 128;
 970                 }
 971 #endif
 972         }
 973
 974         return rt;
 975 }
 976
 977 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 978 {
 979         struct rt6_info *pcpu_rt;
 980
 981         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 982                                   rt->dst.dev, rt->dst.flags);
 983
 984         if (!pcpu_rt)
 985                 return NULL;
 986         ip6_rt_copy_init(pcpu_rt, rt);
 987         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 988         pcpu_rt->rt6i_flags |= RTF_PCPU;
 989         return pcpu_rt;
 990 }
 991
 992 /* It should be called with read_lock_bh(&tb6_lock) acquired */
 993 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
 994 {
 995         struct rt6_info *pcpu_rt, **p;
 996
 997         p = this_cpu_ptr(rt->rt6i_pcpu);
 998         pcpu_rt = *p;
 999
1000         if (pcpu_rt) {
1001                 dst_hold(&pcpu_rt->dst);
1002                 rt6_dst_from_metrics_check(pcpu_rt);
1003         }
1004         return pcpu_rt;
1005 }
1006
1007 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1008 {
1009         struct fib6_table *table = rt->rt6i_table;
1010         struct rt6_info *pcpu_rt, *prev, **p;
1011
1012         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1013         if (!pcpu_rt) {
1014                 struct net *net = dev_net(rt->dst.dev);
1015
1016                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1017                 return net->ipv6.ip6_null_entry;
1018         }
1019
1020         read_lock_bh(&table->tb6_lock);
1021         if (rt->rt6i_pcpu) {
1022                 p = this_cpu_ptr(rt->rt6i_pcpu);
1023                 prev = cmpxchg(p, NULL, pcpu_rt);
1024                 if (prev) {
1025                         /* If someone did it before us, return prev instead */
1026                         dst_destroy(&pcpu_rt->dst);
1027                         pcpu_rt = prev;
1028                 }
1029         } else {
1030                 /* rt has been removed from the fib6 tree
1031                  * before we have a chance to acquire the read_lock.
1032                  * In this case, don't brother to create a pcpu rt
1033                  * since rt is going away anyway.  The next
1034                  * dst_check() will trigger a re-lookup.
1035                  */
1036                 dst_destroy(&pcpu_rt->dst);
1037                 pcpu_rt = rt;
1038         }
1039         dst_hold(&pcpu_rt->dst);
1040         rt6_dst_from_metrics_check(pcpu_rt);
1041         read_unlock_bh(&table->tb6_lock);
1042         return pcpu_rt;
1043 }
1044
1045 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1046                                       struct flowi6 *fl6, int flags)
1047 {
1048         struct fib6_node *fn, *saved_fn;
1049         struct rt6_info *rt;
1050         int strict = 0;
1051
1052         strict |= flags & RT6_LOOKUP_F_IFACE;
1053         if (net->ipv6.devconf_all->forwarding == 0)
1054                 strict |= RT6_LOOKUP_F_REACHABLE;
1055
1056         read_lock_bh(&table->tb6_lock);
1057
1058         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1059         saved_fn = fn;
1060
1061         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1062                 oif = 0;
1063
1064 redo_rt6_select:
1065         rt = rt6_select(fn, oif, strict);
1066         if (rt->rt6i_nsiblings)
1067                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1068         if (rt == net->ipv6.ip6_null_entry) {
1069                 fn = fib6_backtrack(fn, &fl6->saddr);
1070                 if (fn)
1071                         goto redo_rt6_select;
1072                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1073                         /* also consider unreachable route */
1074                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1075                         fn = saved_fn;
1076                         goto redo_rt6_select;
1077                 }
1078         }
1079
1080
1081         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1082                 dst_use(&rt->dst, jiffies);
1083                 read_unlock_bh(&table->tb6_lock);
1084
1085                 rt6_dst_from_metrics_check(rt);
1086
1087                 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1088                 return rt;
1089         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1090                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1091                 /* Create a RTF_CACHE clone which will not be
1092                  * owned by the fib6 tree.  It is for the special case where
1093                  * the daddr in the skb during the neighbor look-up is different
1094                  * from the fl6->daddr used to look-up route here.
1095                  */
1096
1097                 struct rt6_info *uncached_rt;
1098
1099                 dst_use(&rt->dst, jiffies);
1100                 read_unlock_bh(&table->tb6_lock);
1101
1102                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1103                 dst_release(&rt->dst);
1104
1105                 if (uncached_rt)
1106                         rt6_uncached_list_add(uncached_rt);
1107                 else
1108                         uncached_rt = net->ipv6.ip6_null_entry;
1109
1110                 dst_hold(&uncached_rt->dst);
1111
1112                 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1113                 return uncached_rt;
1114
1115         } else {
1116                 /* Get a percpu copy */
1117
1118                 struct rt6_info *pcpu_rt;
1119
1120                 rt->dst.lastuse = jiffies;
1121                 rt->dst.__use++;
1122                 pcpu_rt = rt6_get_pcpu_route(rt);
1123
1124                 if (pcpu_rt) {
1125                         read_unlock_bh(&table->tb6_lock);
1126                 } else {
1127                         /* We have to do the read_unlock first
1128                          * because rt6_make_pcpu_route() may trigger
1129                          * ip6_dst_gc() which will take the write_lock.
1130                          */
1131                         dst_hold(&rt->dst);
1132                         read_unlock_bh(&table->tb6_lock);
1133                         pcpu_rt = rt6_make_pcpu_route(rt);
1134                         dst_release(&rt->dst);
1135                 }
1136
1137                 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1138                 return pcpu_rt;
1139
1140         }
1141 }
1142
1143 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1144                                             struct flowi6 *fl6, int flags)
1145 {
1146         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1147 }
1148
1149 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1150                                                 struct net_device *dev,
1151                                                 struct flowi6 *fl6, int flags)
1152 {
1153         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1154                 flags |= RT6_LOOKUP_F_IFACE;
1155
1156         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1157 }
1158
1159 void ip6_route_input(struct sk_buff *skb)
1160 {
1161         const struct ipv6hdr *iph = ipv6_hdr(skb);
1162         struct net *net = dev_net(skb->dev);
1163         int flags = RT6_LOOKUP_F_HAS_SADDR;
1164         struct ip_tunnel_info *tun_info;
1165         struct flowi6 fl6 = {
1166                 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1167                 .daddr = iph->daddr,
1168                 .saddr = iph->saddr,
1169                 .flowlabel = ip6_flowinfo(iph),
1170                 .flowi6_mark = skb->mark,
1171                 .flowi6_proto = iph->nexthdr,
1172         };
1173
1174         tun_info = skb_tunnel_info(skb);
1175         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1176                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1177         skb_dst_drop(skb);
1178         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1179 }
1180
1181 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1182                                              struct flowi6 *fl6, int flags)
1183 {
1184         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1185 }
1186
1187 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1188                                          struct flowi6 *fl6, int flags)
1189 {
1190         struct dst_entry *dst;
1191         bool any_src;
1192
1193         dst = l3mdev_get_rt6_dst(net, fl6);
1194         if (dst)
1195                 return dst;
1196
1197         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1198
1199         any_src = ipv6_addr_any(&fl6->saddr);
1200         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1201             (fl6->flowi6_oif && any_src))
1202                 flags |= RT6_LOOKUP_F_IFACE;
1203
1204         if (!any_src)
1205                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1206         else if (sk)
1207                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1208
1209         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1210 }
1211 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1212
1213 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1214 {
1215         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1216         struct dst_entry *new = NULL;
1217
1218         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1219         if (rt) {
1220                 rt6_info_init(rt);
1221
1222                 new = &rt->dst;
1223                 new->__use = 1;
1224                 new->input = dst_discard;
1225                 new->output = dst_discard_out;
1226
1227                 dst_copy_metrics(new, &ort->dst);
1228                 rt->rt6i_idev = ort->rt6i_idev;
1229                 if (rt->rt6i_idev)
1230                         in6_dev_hold(rt->rt6i_idev);
1231
1232                 rt->rt6i_gateway = ort->rt6i_gateway;
1233                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1234                 rt->rt6i_metric = 0;
1235
1236                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1237 #ifdef CONFIG_IPV6_SUBTREES
1238                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1239 #endif
1240
1241                 dst_free(new);
1242         }
1243
1244         dst_release(dst_orig);
1245         return new ? new : ERR_PTR(-ENOMEM);
1246 }
1247
1248 /*
1249  *      Destination cache support functions
1250  */
1251
1252 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1253 {
1254         if (rt->dst.from &&
1255             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1256                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1257 }
1258
1259 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1260 {
1261         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1262                 return NULL;
1263
1264         if (rt6_check_expired(rt))
1265                 return NULL;
1266
1267         return &rt->dst;
1268 }
1269
1270 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1271 {
1272         if (!__rt6_check_expired(rt) &&
1273             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1274             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1275                 return &rt->dst;
1276         else
1277                 return NULL;
1278 }
1279
1280 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1281 {
1282         struct rt6_info *rt;
1283
1284         rt = (struct rt6_info *) dst;
1285
1286         /* All IPV6 dsts are created with ->obsolete set to the value
1287          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1288          * into this function always.
1289          */
1290
1291         rt6_dst_from_metrics_check(rt);
1292
1293         if (rt->rt6i_flags & RTF_PCPU ||
1294             (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1295                 return rt6_dst_from_check(rt, cookie);
1296         else
1297                 return rt6_check(rt, cookie);
1298 }
1299
1300 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1301 {
1302         struct rt6_info *rt = (struct rt6_info *) dst;
1303
1304         if (rt) {
1305                 if (rt->rt6i_flags & RTF_CACHE) {
1306                         if (rt6_check_expired(rt)) {
1307                                 ip6_del_rt(rt);
1308                                 dst = NULL;
1309                         }
1310                 } else {
1311                         dst_release(dst);
1312                         dst = NULL;
1313                 }
1314         }
1315         return dst;
1316 }
1317
1318 static void ip6_link_failure(struct sk_buff *skb)
1319 {
1320         struct rt6_info *rt;
1321
1322         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1323
1324         rt = (struct rt6_info *) skb_dst(skb);
1325         if (rt) {
1326                 if (rt->rt6i_flags & RTF_CACHE) {
1327                         dst_hold(&rt->dst);
1328                         ip6_del_rt(rt);
1329                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1330                         rt->rt6i_node->fn_sernum = -1;
1331                 }
1332         }
1333 }
1334
1335 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1336 {
1337         struct net *net = dev_net(rt->dst.dev);
1338
1339         rt->rt6i_flags |= RTF_MODIFIED;
1340         rt->rt6i_pmtu = mtu;
1341         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1342 }
1343
1344 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1345 {
1346         return !(rt->rt6i_flags & RTF_CACHE) &&
1347                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1348 }
1349
1350 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1351                                  const struct ipv6hdr *iph, u32 mtu)
1352 {
1353         struct rt6_info *rt6 = (struct rt6_info *)dst;
1354
1355         if (rt6->rt6i_flags & RTF_LOCAL)
1356                 return;
1357
1358         dst_confirm(dst);
1359         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1360         if (mtu >= dst_mtu(dst))
1361                 return;
1362
1363         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1364                 rt6_do_update_pmtu(rt6, mtu);
1365         } else {
1366                 const struct in6_addr *daddr, *saddr;
1367                 struct rt6_info *nrt6;
1368
1369                 if (iph) {
1370                         daddr = &iph->daddr;
1371                         saddr = &iph->saddr;
1372                 } else if (sk) {
1373                         daddr = &sk->sk_v6_daddr;
1374                         saddr = &inet6_sk(sk)->saddr;
1375                 } else {
1376                         return;
1377                 }
1378                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1379                 if (nrt6) {
1380                         rt6_do_update_pmtu(nrt6, mtu);
1381
1382                         /* ip6_ins_rt(nrt6) will bump the
1383                          * rt6->rt6i_node->fn_sernum
1384                          * which will fail the next rt6_check() and
1385                          * invalidate the sk->sk_dst_cache.
1386                          */
1387                         ip6_ins_rt(nrt6);
1388                 }
1389         }
1390 }
1391
1392 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1393                                struct sk_buff *skb, u32 mtu)
1394 {
1395         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1396 }
1397
1398 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1399                      int oif, u32 mark)
1400 {
1401         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1402         struct dst_entry *dst;
1403         struct flowi6 fl6;
1404
1405         memset(&fl6, 0, sizeof(fl6));
1406         fl6.flowi6_oif = oif;
1407         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1408         fl6.daddr = iph->daddr;
1409         fl6.saddr = iph->saddr;
1410         fl6.flowlabel = ip6_flowinfo(iph);
1411
1412         dst = ip6_route_output(net, NULL, &fl6);
1413         if (!dst->error)
1414                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1415         dst_release(dst);
1416 }
1417 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1418
1419 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1420 {
1421         struct dst_entry *dst;
1422
1423         ip6_update_pmtu(skb, sock_net(sk), mtu,
1424                         sk->sk_bound_dev_if, sk->sk_mark);
1425
1426         dst = __sk_dst_get(sk);
1427         if (!dst || !dst->obsolete ||
1428             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1429                 return;
1430
1431         bh_lock_sock(sk);
1432         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1433                 ip6_datagram_dst_update(sk, false);
1434         bh_unlock_sock(sk);
1435 }
1436 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1437
1438 /* Handle redirects */
1439 struct ip6rd_flowi {
1440         struct flowi6 fl6;
1441         struct in6_addr gateway;
1442 };
1443
1444 static struct rt6_info *__ip6_route_redirect(struct net *net,
1445                                              struct fib6_table *table,
1446                                              struct flowi6 *fl6,
1447                                              int flags)
1448 {
1449         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1450         struct rt6_info *rt;
1451         struct fib6_node *fn;
1452
1453         /* Get the "current" route for this destination and
1454          * check if the redirect has come from approriate router.
1455          *
1456          * RFC 4861 specifies that redirects should only be
1457          * accepted if they come from the nexthop to the target.
1458          * Due to the way the routes are chosen, this notion
1459          * is a bit fuzzy and one might need to check all possible
1460          * routes.
1461          */
1462
1463         read_lock_bh(&table->tb6_lock);
1464         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1465 restart:
1466         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1467                 if (rt6_check_expired(rt))
1468                         continue;
1469                 if (rt->dst.error)
1470                         break;
1471                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1472                         continue;
1473                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1474                         continue;
1475                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1476                         continue;
1477                 break;
1478         }
1479
1480         if (!rt)
1481                 rt = net->ipv6.ip6_null_entry;
1482         else if (rt->dst.error) {
1483                 rt = net->ipv6.ip6_null_entry;
1484                 goto out;
1485         }
1486
1487         if (rt == net->ipv6.ip6_null_entry) {
1488                 fn = fib6_backtrack(fn, &fl6->saddr);
1489                 if (fn)
1490                         goto restart;
1491         }
1492
1493 out:
1494         dst_hold(&rt->dst);
1495
1496         read_unlock_bh(&table->tb6_lock);
1497
1498         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1499         return rt;
1500 };
1501
1502 static struct dst_entry *ip6_route_redirect(struct net *net,
1503                                         const struct flowi6 *fl6,
1504                                         const struct in6_addr *gateway)
1505 {
1506         int flags = RT6_LOOKUP_F_HAS_SADDR;
1507         struct ip6rd_flowi rdfl;
1508
1509         rdfl.fl6 = *fl6;
1510         rdfl.gateway = *gateway;
1511
1512         return fib6_rule_lookup(net, &rdfl.fl6,
1513                                 flags, __ip6_route_redirect);
1514 }
1515
1516 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1517 {
1518         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
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 = iph->daddr;
1527         fl6.saddr = iph->saddr;
1528         fl6.flowlabel = ip6_flowinfo(iph);
1529
1530         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1531         rt6_do_redirect(dst, NULL, skb);
1532         dst_release(dst);
1533 }
1534 EXPORT_SYMBOL_GPL(ip6_redirect);
1535
1536 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1537                             u32 mark)
1538 {
1539         const struct ipv6hdr *iph = ipv6_hdr(skb);
1540         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1541         struct dst_entry *dst;
1542         struct flowi6 fl6;
1543
1544         memset(&fl6, 0, sizeof(fl6));
1545         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1546         fl6.flowi6_oif = oif;
1547         fl6.flowi6_mark = mark;
1548         fl6.daddr = msg->dest;
1549         fl6.saddr = iph->daddr;
1550
1551         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1552         rt6_do_redirect(dst, NULL, skb);
1553         dst_release(dst);
1554 }
1555
1556 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1557 {
1558         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1559 }
1560 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1561
1562 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1563 {
1564         struct net_device *dev = dst->dev;
1565         unsigned int mtu = dst_mtu(dst);
1566         struct net *net = dev_net(dev);
1567
1568         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1569
1570         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1571                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1572
1573         /*
1574          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1575          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1576          * IPV6_MAXPLEN is also valid and means: "any MSS,
1577          * rely only on pmtu discovery"
1578          */
1579         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1580                 mtu = IPV6_MAXPLEN;
1581         return mtu;
1582 }
1583
1584 static unsigned int ip6_mtu(const struct dst_entry *dst)
1585 {
1586         const struct rt6_info *rt = (const struct rt6_info *)dst;
1587         unsigned int mtu = rt->rt6i_pmtu;
1588         struct inet6_dev *idev;
1589
1590         if (mtu)
1591                 goto out;
1592
1593         mtu = dst_metric_raw(dst, RTAX_MTU);
1594         if (mtu)
1595                 goto out;
1596
1597         mtu = IPV6_MIN_MTU;
1598
1599         rcu_read_lock();
1600         idev = __in6_dev_get(dst->dev);
1601         if (idev)
1602                 mtu = idev->cnf.mtu6;
1603         rcu_read_unlock();
1604
1605 out:
1606         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1607 }
1608
1609 static struct dst_entry *icmp6_dst_gc_list;
1610 static DEFINE_SPINLOCK(icmp6_dst_lock);
1611
1612 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1613                                   struct flowi6 *fl6)
1614 {
1615         struct dst_entry *dst;
1616         struct rt6_info *rt;
1617         struct inet6_dev *idev = in6_dev_get(dev);
1618         struct net *net = dev_net(dev);
1619
1620         if (unlikely(!idev))
1621                 return ERR_PTR(-ENODEV);
1622
1623         rt = ip6_dst_alloc(net, dev, 0);
1624         if (unlikely(!rt)) {
1625                 in6_dev_put(idev);
1626                 dst = ERR_PTR(-ENOMEM);
1627                 goto out;
1628         }
1629
1630         rt->dst.flags |= DST_HOST;
1631         rt->dst.output  = ip6_output;
1632         atomic_set(&rt->dst.__refcnt, 1);
1633         rt->rt6i_gateway  = fl6->daddr;
1634         rt->rt6i_dst.addr = fl6->daddr;
1635         rt->rt6i_dst.plen = 128;
1636         rt->rt6i_idev     = idev;
1637         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1638
1639         spin_lock_bh(&icmp6_dst_lock);
1640         rt->dst.next = icmp6_dst_gc_list;
1641         icmp6_dst_gc_list = &rt->dst;
1642         spin_unlock_bh(&icmp6_dst_lock);
1643
1644         fib6_force_start_gc(net);
1645
1646         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1647
1648 out:
1649         return dst;
1650 }
1651
1652 int icmp6_dst_gc(void)
1653 {
1654         struct dst_entry *dst, **pprev;
1655         int more = 0;
1656
1657         spin_lock_bh(&icmp6_dst_lock);
1658         pprev = &icmp6_dst_gc_list;
1659
1660         while ((dst = *pprev) != NULL) {
1661                 if (!atomic_read(&dst->__refcnt)) {
1662                         *pprev = dst->next;
1663                         dst_free(dst);
1664                 } else {
1665                         pprev = &dst->next;
1666                         ++more;
1667                 }
1668         }
1669
1670         spin_unlock_bh(&icmp6_dst_lock);
1671
1672         return more;
1673 }
1674
1675 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1676                             void *arg)
1677 {
1678         struct dst_entry *dst, **pprev;
1679
1680         spin_lock_bh(&icmp6_dst_lock);
1681         pprev = &icmp6_dst_gc_list;
1682         while ((dst = *pprev) != NULL) {
1683                 struct rt6_info *rt = (struct rt6_info *) dst;
1684                 if (func(rt, arg)) {
1685                         *pprev = dst->next;
1686                         dst_free(dst);
1687                 } else {
1688                         pprev = &dst->next;
1689                 }
1690         }
1691         spin_unlock_bh(&icmp6_dst_lock);
1692 }
1693
1694 static int ip6_dst_gc(struct dst_ops *ops)
1695 {
1696         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1697         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1698         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1699         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1700         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1701         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1702         int entries;
1703
1704         entries = dst_entries_get_fast(ops);
1705         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1706             entries <= rt_max_size)
1707                 goto out;
1708
1709         net->ipv6.ip6_rt_gc_expire++;
1710         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1711         entries = dst_entries_get_slow(ops);
1712         if (entries < ops->gc_thresh)
1713                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1714 out:
1715         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1716         return entries > rt_max_size;
1717 }
1718
1719 static int ip6_convert_metrics(struct mx6_config *mxc,
1720                                const struct fib6_config *cfg)
1721 {
1722         bool ecn_ca = false;
1723         struct nlattr *nla;
1724         int remaining;
1725         u32 *mp;
1726
1727         if (!cfg->fc_mx)
1728                 return 0;
1729
1730         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1731         if (unlikely(!mp))
1732                 return -ENOMEM;
1733
1734         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1735                 int type = nla_type(nla);
1736                 u32 val;
1737
1738                 if (!type)
1739                         continue;
1740                 if (unlikely(type > RTAX_MAX))
1741                         goto err;
1742
1743                 if (type == RTAX_CC_ALGO) {
1744                         char tmp[TCP_CA_NAME_MAX];
1745
1746                         nla_strlcpy(tmp, nla, sizeof(tmp));
1747                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1748                         if (val == TCP_CA_UNSPEC)
1749                                 goto err;
1750                 } else {
1751                         val = nla_get_u32(nla);
1752                 }
1753                 if (type == RTAX_HOPLIMIT && val > 255)
1754                         val = 255;
1755                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1756                         goto err;
1757
1758                 mp[type - 1] = val;
1759                 __set_bit(type - 1, mxc->mx_valid);
1760         }
1761
1762         if (ecn_ca) {
1763                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1764                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1765         }
1766
1767         mxc->mx = mp;
1768         return 0;
1769  err:
1770         kfree(mp);
1771         return -EINVAL;
1772 }
1773
1774 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1775                                             struct fib6_config *cfg,
1776                                             const struct in6_addr *gw_addr)
1777 {
1778         struct flowi6 fl6 = {
1779                 .flowi6_oif = cfg->fc_ifindex,
1780                 .daddr = *gw_addr,
1781                 .saddr = cfg->fc_prefsrc,
1782         };
1783         struct fib6_table *table;
1784         struct rt6_info *rt;
1785         int flags = 0;
1786
1787         table = fib6_get_table(net, cfg->fc_table);
1788         if (!table)
1789                 return NULL;
1790
1791         if (!ipv6_addr_any(&cfg->fc_prefsrc))
1792                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1793
1794         rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1795
1796         /* if table lookup failed, fall back to full lookup */
1797         if (rt == net->ipv6.ip6_null_entry) {
1798                 ip6_rt_put(rt);
1799                 rt = NULL;
1800         }
1801
1802         return rt;
1803 }
1804
1805 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1806 {
1807         struct net *net = cfg->fc_nlinfo.nl_net;
1808         struct rt6_info *rt = NULL;
1809         struct net_device *dev = NULL;
1810         struct inet6_dev *idev = NULL;
1811         struct fib6_table *table;
1812         int addr_type;
1813         int err = -EINVAL;
1814
1815         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1816                 goto out;
1817 #ifndef CONFIG_IPV6_SUBTREES
1818         if (cfg->fc_src_len)
1819                 goto out;
1820 #endif
1821         if (cfg->fc_ifindex) {
1822                 err = -ENODEV;
1823                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1824                 if (!dev)
1825                         goto out;
1826                 idev = in6_dev_get(dev);
1827                 if (!idev)
1828                         goto out;
1829         }
1830
1831         if (cfg->fc_metric == 0)
1832                 cfg->fc_metric = IP6_RT_PRIO_USER;
1833
1834         err = -ENOBUFS;
1835         if (cfg->fc_nlinfo.nlh &&
1836             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1837                 table = fib6_get_table(net, cfg->fc_table);
1838                 if (!table) {
1839                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1840                         table = fib6_new_table(net, cfg->fc_table);
1841                 }
1842         } else {
1843                 table = fib6_new_table(net, cfg->fc_table);
1844         }
1845
1846         if (!table)
1847                 goto out;
1848
1849         rt = ip6_dst_alloc(net, NULL,
1850                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1851
1852         if (!rt) {
1853                 err = -ENOMEM;
1854                 goto out;
1855         }
1856
1857         if (cfg->fc_flags & RTF_EXPIRES)
1858                 rt6_set_expires(rt, jiffies +
1859                                 clock_t_to_jiffies(cfg->fc_expires));
1860         else
1861                 rt6_clean_expires(rt);
1862
1863         if (cfg->fc_protocol == RTPROT_UNSPEC)
1864                 cfg->fc_protocol = RTPROT_BOOT;
1865         rt->rt6i_protocol = cfg->fc_protocol;
1866
1867         addr_type = ipv6_addr_type(&cfg->fc_dst);
1868
1869         if (addr_type & IPV6_ADDR_MULTICAST)
1870                 rt->dst.input = ip6_mc_input;
1871         else if (cfg->fc_flags & RTF_LOCAL)
1872                 rt->dst.input = ip6_input;
1873         else
1874                 rt->dst.input = ip6_forward;
1875
1876         rt->dst.output = ip6_output;
1877
1878         if (cfg->fc_encap) {
1879                 struct lwtunnel_state *lwtstate;
1880
1881                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1882                                            cfg->fc_encap, AF_INET6, cfg,
1883                                            &lwtstate);
1884                 if (err)
1885                         goto out;
1886                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1887                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1888                         rt->dst.lwtstate->orig_output = rt->dst.output;
1889                         rt->dst.output = lwtunnel_output;
1890                 }
1891                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1892                         rt->dst.lwtstate->orig_input = rt->dst.input;
1893                         rt->dst.input = lwtunnel_input;
1894                 }
1895         }
1896
1897         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1898         rt->rt6i_dst.plen = cfg->fc_dst_len;
1899         if (rt->rt6i_dst.plen == 128)
1900                 rt->dst.flags |= DST_HOST;
1901
1902 #ifdef CONFIG_IPV6_SUBTREES
1903         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1904         rt->rt6i_src.plen = cfg->fc_src_len;
1905 #endif
1906
1907         rt->rt6i_metric = cfg->fc_metric;
1908
1909         /* We cannot add true routes via loopback here,
1910            they would result in kernel looping; promote them to reject routes
1911          */
1912         if ((cfg->fc_flags & RTF_REJECT) ||
1913             (dev && (dev->flags & IFF_LOOPBACK) &&
1914              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1915              !(cfg->fc_flags & RTF_LOCAL))) {
1916                 /* hold loopback dev/idev if we haven't done so. */
1917                 if (dev != net->loopback_dev) {
1918                         if (dev) {
1919                                 dev_put(dev);
1920                                 in6_dev_put(idev);
1921                         }
1922                         dev = net->loopback_dev;
1923                         dev_hold(dev);
1924                         idev = in6_dev_get(dev);
1925                         if (!idev) {
1926                                 err = -ENODEV;
1927                                 goto out;
1928                         }
1929                 }
1930                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1931                 switch (cfg->fc_type) {
1932                 case RTN_BLACKHOLE:
1933                         rt->dst.error = -EINVAL;
1934                         rt->dst.output = dst_discard_out;
1935                         rt->dst.input = dst_discard;
1936                         break;
1937                 case RTN_PROHIBIT:
1938                         rt->dst.error = -EACCES;
1939                         rt->dst.output = ip6_pkt_prohibit_out;
1940                         rt->dst.input = ip6_pkt_prohibit;
1941                         break;
1942                 case RTN_THROW:
1943                 case RTN_UNREACHABLE:
1944                 default:
1945                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1946                                         : (cfg->fc_type == RTN_UNREACHABLE)
1947                                         ? -EHOSTUNREACH : -ENETUNREACH;
1948                         rt->dst.output = ip6_pkt_discard_out;
1949                         rt->dst.input = ip6_pkt_discard;
1950                         break;
1951                 }
1952                 goto install_route;
1953         }
1954
1955         if (cfg->fc_flags & RTF_GATEWAY) {
1956                 const struct in6_addr *gw_addr;
1957                 int gwa_type;
1958
1959                 gw_addr = &cfg->fc_gateway;
1960                 gwa_type = ipv6_addr_type(gw_addr);
1961
1962                 /* if gw_addr is local we will fail to detect this in case
1963                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1964                  * will return already-added prefix route via interface that
1965                  * prefix route was assigned to, which might be non-loopback.
1966                  */
1967                 err = -EINVAL;
1968                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1969                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1970                                             dev : NULL, 0, 0))
1971                         goto out;
1972
1973                 rt->rt6i_gateway = *gw_addr;
1974
1975                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1976                         struct rt6_info *grt = NULL;
1977
1978                         /* IPv6 strictly inhibits using not link-local
1979                            addresses as nexthop address.
1980                            Otherwise, router will not able to send redirects.
1981                            It is very good, but in some (rare!) circumstances
1982                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1983                            some exceptions. --ANK
1984                          */
1985                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1986                                 goto out;
1987
1988                         if (cfg->fc_table)
1989                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
1990
1991                         if (!grt)
1992                                 grt = rt6_lookup(net, gw_addr, NULL,
1993                                                  cfg->fc_ifindex, 1);
1994
1995                         err = -EHOSTUNREACH;
1996                         if (!grt)
1997                                 goto out;
1998                         if (dev) {
1999                                 if (dev != grt->dst.dev) {
2000                                         ip6_rt_put(grt);
2001                                         goto out;
2002                                 }
2003                         } else {
2004                                 dev = grt->dst.dev;
2005                                 idev = grt->rt6i_idev;
2006                                 dev_hold(dev);
2007                                 in6_dev_hold(grt->rt6i_idev);
2008                         }
2009                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2010                                 err = 0;
2011                         ip6_rt_put(grt);
2012
2013                         if (err)
2014                                 goto out;
2015                 }
2016                 err = -EINVAL;
2017                 if (!dev || (dev->flags & IFF_LOOPBACK))
2018                         goto out;
2019         }
2020
2021         err = -ENODEV;
2022         if (!dev)
2023                 goto out;
2024
2025         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2026                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2027                         err = -EINVAL;
2028                         goto out;
2029                 }
2030                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2031                 rt->rt6i_prefsrc.plen = 128;
2032         } else
2033                 rt->rt6i_prefsrc.plen = 0;
2034
2035         rt->rt6i_flags = cfg->fc_flags;
2036
2037 install_route:
2038         rt->dst.dev = dev;
2039         rt->rt6i_idev = idev;
2040         rt->rt6i_table = table;
2041
2042         cfg->fc_nlinfo.nl_net = dev_net(dev);
2043
2044         return rt;
2045 out:
2046         if (dev)
2047                 dev_put(dev);
2048         if (idev)
2049                 in6_dev_put(idev);
2050         if (rt)
2051                 dst_free(&rt->dst);
2052
2053         return ERR_PTR(err);
2054 }
2055
2056 int ip6_route_add(struct fib6_config *cfg)
2057 {
2058         struct mx6_config mxc = { .mx = NULL, };
2059         struct rt6_info *rt;
2060         int err;
2061
2062         rt = ip6_route_info_create(cfg);
2063         if (IS_ERR(rt)) {
2064                 err = PTR_ERR(rt);
2065                 rt = NULL;
2066                 goto out;
2067         }
2068
2069         err = ip6_convert_metrics(&mxc, cfg);
2070         if (err)
2071                 goto out;
2072
2073         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2074
2075         kfree(mxc.mx);
2076
2077         return err;
2078 out:
2079         if (rt)
2080                 dst_free(&rt->dst);
2081
2082         return err;
2083 }
2084
2085 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2086 {
2087         int err;
2088         struct fib6_table *table;
2089         struct net *net = dev_net(rt->dst.dev);
2090
2091         if (rt == net->ipv6.ip6_null_entry ||
2092             rt->dst.flags & DST_NOCACHE) {
2093                 err = -ENOENT;
2094                 goto out;
2095         }
2096
2097         table = rt->rt6i_table;
2098         write_lock_bh(&table->tb6_lock);
2099         err = fib6_del(rt, info);
2100         write_unlock_bh(&table->tb6_lock);
2101
2102 out:
2103         ip6_rt_put(rt);
2104         return err;
2105 }
2106
2107 int ip6_del_rt(struct rt6_info *rt)
2108 {
2109         struct nl_info info = {
2110                 .nl_net = dev_net(rt->dst.dev),
2111         };
2112         return __ip6_del_rt(rt, &info);
2113 }
2114
2115 static int ip6_route_del(struct fib6_config *cfg)
2116 {
2117         struct fib6_table *table;
2118         struct fib6_node *fn;
2119         struct rt6_info *rt;
2120         int err = -ESRCH;
2121
2122         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2123         if (!table)
2124                 return err;
2125
2126         read_lock_bh(&table->tb6_lock);
2127
2128         fn = fib6_locate(&table->tb6_root,
2129                          &cfg->fc_dst, cfg->fc_dst_len,
2130                          &cfg->fc_src, cfg->fc_src_len);
2131
2132         if (fn) {
2133                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2134                         if ((rt->rt6i_flags & RTF_CACHE) &&
2135                             !(cfg->fc_flags & RTF_CACHE))
2136                                 continue;
2137                         if (cfg->fc_ifindex &&
2138                             (!rt->dst.dev ||
2139                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2140                                 continue;
2141                         if (cfg->fc_flags & RTF_GATEWAY &&
2142                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2143                                 continue;
2144                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2145                                 continue;
2146                         dst_hold(&rt->dst);
2147                         read_unlock_bh(&table->tb6_lock);
2148
2149                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2150                 }
2151         }
2152         read_unlock_bh(&table->tb6_lock);
2153
2154         return err;
2155 }
2156
2157 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2158 {
2159         struct netevent_redirect netevent;
2160         struct rt6_info *rt, *nrt = NULL;
2161         struct ndisc_options ndopts;
2162         struct inet6_dev *in6_dev;
2163         struct neighbour *neigh;
2164         struct rd_msg *msg;
2165         int optlen, on_link;
2166         u8 *lladdr;
2167
2168         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2169         optlen -= sizeof(*msg);
2170
2171         if (optlen < 0) {
2172                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2173                 return;
2174         }
2175
2176         msg = (struct rd_msg *)icmp6_hdr(skb);
2177
2178         if (ipv6_addr_is_multicast(&msg->dest)) {
2179                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2180                 return;
2181         }
2182
2183         on_link = 0;
2184         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2185                 on_link = 1;
2186         } else if (ipv6_addr_type(&msg->target) !=
2187                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2188                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2189                 return;
2190         }
2191
2192         in6_dev = __in6_dev_get(skb->dev);
2193         if (!in6_dev)
2194                 return;
2195         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2196                 return;
2197
2198         /* RFC2461 8.1:
2199          *      The IP source address of the Redirect MUST be the same as the current
2200          *      first-hop router for the specified ICMP Destination Address.
2201          */
2202
2203         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2204                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2205                 return;
2206         }
2207
2208         lladdr = NULL;
2209         if (ndopts.nd_opts_tgt_lladdr) {
2210                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2211                                              skb->dev);
2212                 if (!lladdr) {
2213                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2214                         return;
2215                 }
2216         }
2217
2218         rt = (struct rt6_info *) dst;
2219         if (rt->rt6i_flags & RTF_REJECT) {
2220                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2221                 return;
2222         }
2223
2224         /* Redirect received -> path was valid.
2225          * Look, redirects are sent only in response to data packets,
2226          * so that this nexthop apparently is reachable. --ANK
2227          */
2228         dst_confirm(&rt->dst);
2229
2230         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2231         if (!neigh)
2232                 return;
2233
2234         /*
2235          *      We have finally decided to accept it.
2236          */
2237
2238         neigh_update(neigh, lladdr, NUD_STALE,
2239                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2240                      NEIGH_UPDATE_F_OVERRIDE|
2241                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2242                                      NEIGH_UPDATE_F_ISROUTER))
2243                      );
2244
2245         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2246         if (!nrt)
2247                 goto out;
2248
2249         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2250         if (on_link)
2251                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2252
2253         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2254
2255         if (ip6_ins_rt(nrt))
2256                 goto out;
2257
2258         netevent.old = &rt->dst;
2259         netevent.new = &nrt->dst;
2260         netevent.daddr = &msg->dest;
2261         netevent.neigh = neigh;
2262         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2263
2264         if (rt->rt6i_flags & RTF_CACHE) {
2265                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2266                 ip6_del_rt(rt);
2267         }
2268
2269 out:
2270         neigh_release(neigh);
2271 }
2272
2273 /*
2274  *      Misc support functions
2275  */
2276
2277 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2278 {
2279         BUG_ON(from->dst.from);
2280
2281         rt->rt6i_flags &= ~RTF_EXPIRES;
2282         dst_hold(&from->dst);
2283         rt->dst.from = &from->dst;
2284         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2285 }
2286
2287 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2288 {
2289         rt->dst.input = ort->dst.input;
2290         rt->dst.output = ort->dst.output;
2291         rt->rt6i_dst = ort->rt6i_dst;
2292         rt->dst.error = ort->dst.error;
2293         rt->rt6i_idev = ort->rt6i_idev;
2294         if (rt->rt6i_idev)
2295                 in6_dev_hold(rt->rt6i_idev);
2296         rt->dst.lastuse = jiffies;
2297         rt->rt6i_gateway = ort->rt6i_gateway;
2298         rt->rt6i_flags = ort->rt6i_flags;
2299         rt6_set_from(rt, ort);
2300         rt->rt6i_metric = ort->rt6i_metric;
2301 #ifdef CONFIG_IPV6_SUBTREES
2302         rt->rt6i_src = ort->rt6i_src;
2303 #endif
2304         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2305         rt->rt6i_table = ort->rt6i_table;
2306         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2307 }
2308
2309 #ifdef CONFIG_IPV6_ROUTE_INFO
2310 static struct rt6_info *rt6_get_route_info(struct net *net,
2311                                            const struct in6_addr *prefix, int prefixlen,
2312                                            const struct in6_addr *gwaddr, int ifindex)
2313 {
2314         struct fib6_node *fn;
2315         struct rt6_info *rt = NULL;
2316         struct fib6_table *table;
2317
2318         table = fib6_get_table(net, RT6_TABLE_INFO);
2319         if (!table)
2320                 return NULL;
2321
2322         read_lock_bh(&table->tb6_lock);
2323         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2324         if (!fn)
2325                 goto out;
2326
2327         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2328                 if (rt->dst.dev->ifindex != ifindex)
2329                         continue;
2330                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2331                         continue;
2332                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2333                         continue;
2334                 dst_hold(&rt->dst);
2335                 break;
2336         }
2337 out:
2338         read_unlock_bh(&table->tb6_lock);
2339         return rt;
2340 }
2341
2342 static struct rt6_info *rt6_add_route_info(struct net *net,
2343                                            const struct in6_addr *prefix, int prefixlen,
2344                                            const struct in6_addr *gwaddr, int ifindex,
2345                                            unsigned int pref)
2346 {
2347         struct fib6_config cfg = {
2348                 .fc_metric      = IP6_RT_PRIO_USER,
2349                 .fc_ifindex     = ifindex,
2350                 .fc_dst_len     = prefixlen,
2351                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2352                                   RTF_UP | RTF_PREF(pref),
2353                 .fc_nlinfo.portid = 0,
2354                 .fc_nlinfo.nlh = NULL,
2355                 .fc_nlinfo.nl_net = net,
2356         };
2357
2358         cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2359         cfg.fc_dst = *prefix;
2360         cfg.fc_gateway = *gwaddr;
2361
2362         /* We should treat it as a default route if prefix length is 0. */
2363         if (!prefixlen)
2364                 cfg.fc_flags |= RTF_DEFAULT;
2365
2366         ip6_route_add(&cfg);
2367
2368         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2369 }
2370 #endif
2371
2372 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2373 {
2374         struct rt6_info *rt;
2375         struct fib6_table *table;
2376
2377         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2378         if (!table)
2379                 return NULL;
2380
2381         read_lock_bh(&table->tb6_lock);
2382         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2383                 if (dev == rt->dst.dev &&
2384                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2385                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2386                         break;
2387         }
2388         if (rt)
2389                 dst_hold(&rt->dst);
2390         read_unlock_bh(&table->tb6_lock);
2391         return rt;
2392 }
2393
2394 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2395                                      struct net_device *dev,
2396                                      unsigned int pref)
2397 {
2398         struct fib6_config cfg = {
2399                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2400                 .fc_metric      = IP6_RT_PRIO_USER,
2401                 .fc_ifindex     = dev->ifindex,
2402                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2403                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2404                 .fc_nlinfo.portid = 0,
2405                 .fc_nlinfo.nlh = NULL,
2406                 .fc_nlinfo.nl_net = dev_net(dev),
2407         };
2408
2409         cfg.fc_gateway = *gwaddr;
2410
2411         ip6_route_add(&cfg);
2412
2413         return rt6_get_dflt_router(gwaddr, dev);
2414 }
2415
2416 void rt6_purge_dflt_routers(struct net *net)
2417 {
2418         struct rt6_info *rt;
2419         struct fib6_table *table;
2420
2421         /* NOTE: Keep consistent with rt6_get_dflt_router */
2422         table = fib6_get_table(net, RT6_TABLE_DFLT);
2423         if (!table)
2424                 return;
2425
2426 restart:
2427         read_lock_bh(&table->tb6_lock);
2428         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2429                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2430                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2431                         dst_hold(&rt->dst);
2432                         read_unlock_bh(&table->tb6_lock);
2433                         ip6_del_rt(rt);
2434                         goto restart;
2435                 }
2436         }
2437         read_unlock_bh(&table->tb6_lock);
2438 }
2439
2440 static void rtmsg_to_fib6_config(struct net *net,
2441                                  struct in6_rtmsg *rtmsg,
2442                                  struct fib6_config *cfg)
2443 {
2444         memset(cfg, 0, sizeof(*cfg));
2445
2446         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2447                          : RT6_TABLE_MAIN;
2448         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2449         cfg->fc_metric = rtmsg->rtmsg_metric;
2450         cfg->fc_expires = rtmsg->rtmsg_info;
2451         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2452         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2453         cfg->fc_flags = rtmsg->rtmsg_flags;
2454
2455         cfg->fc_nlinfo.nl_net = net;
2456
2457         cfg->fc_dst = rtmsg->rtmsg_dst;
2458         cfg->fc_src = rtmsg->rtmsg_src;
2459         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2460 }
2461
2462 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2463 {
2464         struct fib6_config cfg;
2465         struct in6_rtmsg rtmsg;
2466         int err;
2467
2468         switch (cmd) {
2469         case SIOCADDRT:         /* Add a route */
2470         case SIOCDELRT:         /* Delete a route */
2471                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2472                         return -EPERM;
2473                 err = copy_from_user(&rtmsg, arg,
2474                                      sizeof(struct in6_rtmsg));
2475                 if (err)
2476                         return -EFAULT;
2477
2478                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2479
2480                 rtnl_lock();
2481                 switch (cmd) {
2482                 case SIOCADDRT:
2483                         err = ip6_route_add(&cfg);
2484                         break;
2485                 case SIOCDELRT:
2486                         err = ip6_route_del(&cfg);
2487                         break;
2488                 default:
2489                         err = -EINVAL;
2490                 }
2491                 rtnl_unlock();
2492
2493                 return err;
2494         }
2495
2496         return -EINVAL;
2497 }
2498
2499 /*
2500  *      Drop the packet on the floor
2501  */
2502
2503 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2504 {
2505         int type;
2506         struct dst_entry *dst = skb_dst(skb);
2507         switch (ipstats_mib_noroutes) {
2508         case IPSTATS_MIB_INNOROUTES:
2509                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2510                 if (type == IPV6_ADDR_ANY) {
2511                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2512                                       IPSTATS_MIB_INADDRERRORS);
2513                         break;
2514                 }
2515                 /* FALLTHROUGH */
2516         case IPSTATS_MIB_OUTNOROUTES:
2517                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2518                               ipstats_mib_noroutes);
2519                 break;
2520         }
2521         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2522         kfree_skb(skb);
2523         return 0;
2524 }
2525
2526 static int ip6_pkt_discard(struct sk_buff *skb)
2527 {
2528         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2529 }
2530
2531 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2532 {
2533         skb->dev = skb_dst(skb)->dev;
2534         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2535 }
2536
2537 static int ip6_pkt_prohibit(struct sk_buff *skb)
2538 {
2539         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2540 }
2541
2542 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2543 {
2544         skb->dev = skb_dst(skb)->dev;
2545         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2546 }
2547
2548 /*
2549  *      Allocate a dst for local (unicast / anycast) address.
2550  */
2551
2552 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2553                                     const struct in6_addr *addr,
2554                                     bool anycast)
2555 {
2556         u32 tb_id;
2557         struct net *net = dev_net(idev->dev);
2558         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2559                                             DST_NOCOUNT);
2560         if (!rt)
2561                 return ERR_PTR(-ENOMEM);
2562
2563         in6_dev_hold(idev);
2564
2565         rt->dst.flags |= DST_HOST;
2566         rt->dst.input = ip6_input;
2567         rt->dst.output = ip6_output;
2568         rt->rt6i_idev = idev;
2569
2570         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2571         if (anycast)
2572                 rt->rt6i_flags |= RTF_ANYCAST;
2573         else
2574                 rt->rt6i_flags |= RTF_LOCAL;
2575
2576         rt->rt6i_gateway  = *addr;
2577         rt->rt6i_dst.addr = *addr;
2578         rt->rt6i_dst.plen = 128;
2579         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2580         rt->rt6i_table = fib6_get_table(net, tb_id);
2581         rt->dst.flags |= DST_NOCACHE;
2582
2583         atomic_set(&rt->dst.__refcnt, 1);
2584
2585         return rt;
2586 }
2587
2588 int ip6_route_get_saddr(struct net *net,
2589                         struct rt6_info *rt,
2590                         const struct in6_addr *daddr,
2591                         unsigned int prefs,
2592                         struct in6_addr *saddr)
2593 {
2594         struct inet6_dev *idev =
2595                 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2596         int err = 0;
2597         if (rt && rt->rt6i_prefsrc.plen)
2598                 *saddr = rt->rt6i_prefsrc.addr;
2599         else
2600                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2601                                          daddr, prefs, saddr);
2602         return err;
2603 }
2604
2605 /* remove deleted ip from prefsrc entries */
2606 struct arg_dev_net_ip {
2607         struct net_device *dev;
2608         struct net *net;
2609         struct in6_addr *addr;
2610 };
2611
2612 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2613 {
2614         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2615         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2616         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2617
2618         if (((void *)rt->dst.dev == dev || !dev) &&
2619             rt != net->ipv6.ip6_null_entry &&
2620             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2621                 /* remove prefsrc entry */
2622                 rt->rt6i_prefsrc.plen = 0;
2623         }
2624         return 0;
2625 }
2626
2627 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2628 {
2629         struct net *net = dev_net(ifp->idev->dev);
2630         struct arg_dev_net_ip adni = {
2631                 .dev = ifp->idev->dev,
2632                 .net = net,
2633                 .addr = &ifp->addr,
2634         };
2635         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2636 }
2637
2638 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2639 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2640
2641 /* Remove routers and update dst entries when gateway turn into host. */
2642 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2643 {
2644         struct in6_addr *gateway = (struct in6_addr *)arg;
2645
2646         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2647              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2648              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2649                 return -1;
2650         }
2651         return 0;
2652 }
2653
2654 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2655 {
2656         fib6_clean_all(net, fib6_clean_tohost, gateway);
2657 }
2658
2659 struct arg_dev_net {
2660         struct net_device *dev;
2661         struct net *net;
2662 };
2663
2664 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2665 {
2666         const struct arg_dev_net *adn = arg;
2667         const struct net_device *dev = adn->dev;
2668
2669         if ((rt->dst.dev == dev || !dev) &&
2670             rt != adn->net->ipv6.ip6_null_entry)
2671                 return -1;
2672
2673         return 0;
2674 }
2675
2676 void rt6_ifdown(struct net *net, struct net_device *dev)
2677 {
2678         struct arg_dev_net adn = {
2679                 .dev = dev,
2680                 .net = net,
2681         };
2682
2683         fib6_clean_all(net, fib6_ifdown, &adn);
2684         icmp6_clean_all(fib6_ifdown, &adn);
2685         if (dev)
2686                 rt6_uncached_list_flush_dev(net, dev);
2687 }
2688
2689 struct rt6_mtu_change_arg {
2690         struct net_device *dev;
2691         unsigned int mtu;
2692 };
2693
2694 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2695 {
2696         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2697         struct inet6_dev *idev;
2698
2699         /* In IPv6 pmtu discovery is not optional,
2700            so that RTAX_MTU lock cannot disable it.
2701            We still use this lock to block changes
2702            caused by addrconf/ndisc.
2703         */
2704
2705         idev = __in6_dev_get(arg->dev);
2706         if (!idev)
2707                 return 0;
2708
2709         /* For administrative MTU increase, there is no way to discover
2710            IPv6 PMTU increase, so PMTU increase should be updated here.
2711            Since RFC 1981 doesn't include administrative MTU increase
2712            update PMTU increase is a MUST. (i.e. jumbo frame)
2713          */
2714         /*
2715            If new MTU is less than route PMTU, this new MTU will be the
2716            lowest MTU in the path, update the route PMTU to reflect PMTU
2717            decreases; if new MTU is greater than route PMTU, and the
2718            old MTU is the lowest MTU in the path, update the route PMTU
2719            to reflect the increase. In this case if the other nodes' MTU
2720            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2721            PMTU discouvery.
2722          */
2723         if (rt->dst.dev == arg->dev &&
2724             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2725                 if (rt->rt6i_flags & RTF_CACHE) {
2726                         /* For RTF_CACHE with rt6i_pmtu == 0
2727                          * (i.e. a redirected route),
2728                          * the metrics of its rt->dst.from has already
2729                          * been updated.
2730                          */
2731                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2732                                 rt->rt6i_pmtu = arg->mtu;
2733                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2734                            (dst_mtu(&rt->dst) < arg->mtu &&
2735                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2736                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2737                 }
2738         }
2739         return 0;
2740 }
2741
2742 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2743 {
2744         struct rt6_mtu_change_arg arg = {
2745                 .dev = dev,
2746                 .mtu = mtu,
2747         };
2748
2749         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2750 }
2751
2752 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2753         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2754         [RTA_OIF]               = { .type = NLA_U32 },
2755         [RTA_IIF]               = { .type = NLA_U32 },
2756         [RTA_PRIORITY]          = { .type = NLA_U32 },
2757         [RTA_METRICS]           = { .type = NLA_NESTED },
2758         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2759         [RTA_PREF]              = { .type = NLA_U8 },
2760         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2761         [RTA_ENCAP]             = { .type = NLA_NESTED },
2762         [RTA_EXPIRES]           = { .type = NLA_U32 },
2763 };
2764
2765 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2766                               struct fib6_config *cfg)
2767 {
2768         struct rtmsg *rtm;
2769         struct nlattr *tb[RTA_MAX+1];
2770         unsigned int pref;
2771         int err;
2772
2773         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2774         if (err < 0)
2775                 goto errout;
2776
2777         err = -EINVAL;
2778         rtm = nlmsg_data(nlh);
2779         memset(cfg, 0, sizeof(*cfg));
2780
2781         cfg->fc_table = rtm->rtm_table;
2782         cfg->fc_dst_len = rtm->rtm_dst_len;
2783         cfg->fc_src_len = rtm->rtm_src_len;
2784         cfg->fc_flags = RTF_UP;
2785         cfg->fc_protocol = rtm->rtm_protocol;
2786         cfg->fc_type = rtm->rtm_type;
2787
2788         if (rtm->rtm_type == RTN_UNREACHABLE ||
2789             rtm->rtm_type == RTN_BLACKHOLE ||
2790             rtm->rtm_type == RTN_PROHIBIT ||
2791             rtm->rtm_type == RTN_THROW)
2792                 cfg->fc_flags |= RTF_REJECT;
2793
2794         if (rtm->rtm_type == RTN_LOCAL)
2795                 cfg->fc_flags |= RTF_LOCAL;
2796
2797         if (rtm->rtm_flags & RTM_F_CLONED)
2798                 cfg->fc_flags |= RTF_CACHE;
2799
2800         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2801         cfg->fc_nlinfo.nlh = nlh;
2802         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2803
2804         if (tb[RTA_GATEWAY]) {
2805                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2806                 cfg->fc_flags |= RTF_GATEWAY;
2807         }
2808
2809         if (tb[RTA_DST]) {
2810                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2811
2812                 if (nla_len(tb[RTA_DST]) < plen)
2813                         goto errout;
2814
2815                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2816         }
2817
2818         if (tb[RTA_SRC]) {
2819                 int plen = (rtm->rtm_src_len + 7) >> 3;
2820
2821                 if (nla_len(tb[RTA_SRC]) < plen)
2822                         goto errout;
2823
2824                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2825         }
2826
2827         if (tb[RTA_PREFSRC])
2828                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2829
2830         if (tb[RTA_OIF])
2831                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2832
2833         if (tb[RTA_PRIORITY])
2834                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2835
2836         if (tb[RTA_METRICS]) {
2837                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2838                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2839         }
2840
2841         if (tb[RTA_TABLE])
2842                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2843
2844         if (tb[RTA_MULTIPATH]) {
2845                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2846                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2847         }
2848
2849         if (tb[RTA_PREF]) {
2850                 pref = nla_get_u8(tb[RTA_PREF]);
2851                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2852                     pref != ICMPV6_ROUTER_PREF_HIGH)
2853                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2854                 cfg->fc_flags |= RTF_PREF(pref);
2855         }
2856
2857         if (tb[RTA_ENCAP])
2858                 cfg->fc_encap = tb[RTA_ENCAP];
2859
2860         if (tb[RTA_ENCAP_TYPE])
2861                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2862
2863         if (tb[RTA_EXPIRES]) {
2864                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2865
2866                 if (addrconf_finite_timeout(timeout)) {
2867                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2868                         cfg->fc_flags |= RTF_EXPIRES;
2869                 }
2870         }
2871
2872         err = 0;
2873 errout:
2874         return err;
2875 }
2876
2877 struct rt6_nh {
2878         struct rt6_info *rt6_info;
2879         struct fib6_config r_cfg;
2880         struct mx6_config mxc;
2881         struct list_head next;
2882 };
2883
2884 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2885 {
2886         struct rt6_nh *nh;
2887
2888         list_for_each_entry(nh, rt6_nh_list, next) {
2889                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2890                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2891                         nh->r_cfg.fc_ifindex);
2892         }
2893 }
2894
2895 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2896                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2897 {
2898         struct rt6_nh *nh;
2899         struct rt6_info *rtnh;
2900         int err = -EEXIST;
2901
2902         list_for_each_entry(nh, rt6_nh_list, next) {
2903                 /* check if rt6_info already exists */
2904                 rtnh = nh->rt6_info;
2905
2906                 if (rtnh->dst.dev == rt->dst.dev &&
2907                     rtnh->rt6i_idev == rt->rt6i_idev &&
2908                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2909                                     &rt->rt6i_gateway))
2910                         return err;
2911         }
2912
2913         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2914         if (!nh)
2915                 return -ENOMEM;
2916         nh->rt6_info = rt;
2917         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2918         if (err) {
2919                 kfree(nh);
2920                 return err;
2921         }
2922         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2923         list_add_tail(&nh->next, rt6_nh_list);
2924
2925         return 0;
2926 }
2927
2928 static int ip6_route_multipath_add(struct fib6_config *cfg)
2929 {
2930         struct fib6_config r_cfg;
2931         struct rtnexthop *rtnh;
2932         struct rt6_info *rt;
2933         struct rt6_nh *err_nh;
2934         struct rt6_nh *nh, *nh_safe;
2935         int remaining;
2936         int attrlen;
2937         int err = 1;
2938         int nhn = 0;
2939         int replace = (cfg->fc_nlinfo.nlh &&
2940                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2941         LIST_HEAD(rt6_nh_list);
2942
2943         remaining = cfg->fc_mp_len;
2944         rtnh = (struct rtnexthop *)cfg->fc_mp;
2945
2946         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2947          * rt6_info structs per nexthop
2948          */
2949         while (rtnh_ok(rtnh, remaining)) {
2950                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2951                 if (rtnh->rtnh_ifindex)
2952                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2953
2954                 attrlen = rtnh_attrlen(rtnh);
2955                 if (attrlen > 0) {
2956                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2957
2958                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2959                         if (nla) {
2960                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2961                                 r_cfg.fc_flags |= RTF_GATEWAY;
2962                         }
2963                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2964                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2965                         if (nla)
2966                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2967                 }
2968
2969                 rt = ip6_route_info_create(&r_cfg);
2970                 if (IS_ERR(rt)) {
2971                         err = PTR_ERR(rt);
2972                         rt = NULL;
2973                         goto cleanup;
2974                 }
2975
2976                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2977                 if (err) {
2978                         dst_free(&rt->dst);
2979                         goto cleanup;
2980                 }
2981
2982                 rtnh = rtnh_next(rtnh, &remaining);
2983         }
2984
2985         err_nh = NULL;
2986         list_for_each_entry(nh, &rt6_nh_list, next) {
2987                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2988                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2989                 nh->rt6_info = NULL;
2990                 if (err) {
2991                         if (replace && nhn)
2992                                 ip6_print_replace_route_err(&rt6_nh_list);
2993                         err_nh = nh;
2994                         goto add_errout;
2995                 }
2996
2997                 /* Because each route is added like a single route we remove
2998                  * these flags after the first nexthop: if there is a collision,
2999                  * we have already failed to add the first nexthop:
3000                  * fib6_add_rt2node() has rejected it; when replacing, old
3001                  * nexthops have been replaced by first new, the rest should
3002                  * be added to it.
3003                  */
3004                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3005                                                      NLM_F_REPLACE);
3006                 nhn++;
3007         }
3008
3009         goto cleanup;
3010
3011 add_errout:
3012         /* Delete routes that were already added */
3013         list_for_each_entry(nh, &rt6_nh_list, next) {
3014                 if (err_nh == nh)
3015                         break;
3016                 ip6_route_del(&nh->r_cfg);
3017         }
3018
3019 cleanup:
3020         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3021                 if (nh->rt6_info)
3022                         dst_free(&nh->rt6_info->dst);
3023                 kfree(nh->mxc.mx);
3024                 list_del(&nh->next);
3025                 kfree(nh);
3026         }
3027
3028         return err;
3029 }
3030
3031 static int ip6_route_multipath_del(struct fib6_config *cfg)
3032 {
3033         struct fib6_config r_cfg;
3034         struct rtnexthop *rtnh;
3035         int remaining;
3036         int attrlen;
3037         int err = 1, last_err = 0;
3038
3039         remaining = cfg->fc_mp_len;
3040         rtnh = (struct rtnexthop *)cfg->fc_mp;
3041
3042         /* Parse a Multipath Entry */
3043         while (rtnh_ok(rtnh, remaining)) {
3044                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3045                 if (rtnh->rtnh_ifindex)
3046                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3047
3048                 attrlen = rtnh_attrlen(rtnh);
3049                 if (attrlen > 0) {
3050                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3051
3052                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3053                         if (nla) {
3054                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3055                                 r_cfg.fc_flags |= RTF_GATEWAY;
3056                         }
3057                 }
3058                 err = ip6_route_del(&r_cfg);
3059                 if (err)
3060                         last_err = err;
3061
3062                 rtnh = rtnh_next(rtnh, &remaining);
3063         }
3064
3065         return last_err;
3066 }
3067
3068 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3069 {
3070         struct fib6_config cfg;
3071         int err;
3072
3073         err = rtm_to_fib6_config(skb, nlh, &cfg);
3074         if (err < 0)
3075                 return err;
3076
3077         if (cfg.fc_mp)
3078                 return ip6_route_multipath_del(&cfg);
3079         else
3080                 return ip6_route_del(&cfg);
3081 }
3082
3083 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3084 {
3085         struct fib6_config cfg;
3086         int err;
3087
3088         err = rtm_to_fib6_config(skb, nlh, &cfg);
3089         if (err < 0)
3090                 return err;
3091
3092         if (cfg.fc_mp)
3093                 return ip6_route_multipath_add(&cfg);
3094         else
3095                 return ip6_route_add(&cfg);
3096 }
3097
3098 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3099 {
3100         return NLMSG_ALIGN(sizeof(struct rtmsg))
3101                + nla_total_size(16) /* RTA_SRC */
3102                + nla_total_size(16) /* RTA_DST */
3103                + nla_total_size(16) /* RTA_GATEWAY */
3104                + nla_total_size(16) /* RTA_PREFSRC */
3105                + nla_total_size(4) /* RTA_TABLE */
3106                + nla_total_size(4) /* RTA_IIF */
3107                + nla_total_size(4) /* RTA_OIF */
3108                + nla_total_size(4) /* RTA_PRIORITY */
3109                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3110                + nla_total_size(sizeof(struct rta_cacheinfo))
3111                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3112                + nla_total_size(1) /* RTA_PREF */
3113                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3114 }
3115
3116 static int rt6_fill_node(struct net *net,
3117                          struct sk_buff *skb, struct rt6_info *rt,
3118                          struct in6_addr *dst, struct in6_addr *src,
3119                          int iif, int type, u32 portid, u32 seq,
3120                          int prefix, int nowait, unsigned int flags)
3121 {
3122         u32 metrics[RTAX_MAX];
3123         struct rtmsg *rtm;
3124         struct nlmsghdr *nlh;
3125         long expires;
3126         u32 table;
3127
3128         if (prefix) {   /* user wants prefix routes only */
3129                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3130                         /* success since this is not a prefix route */
3131                         return 1;
3132                 }
3133         }
3134
3135         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3136         if (!nlh)
3137                 return -EMSGSIZE;
3138
3139         rtm = nlmsg_data(nlh);
3140         rtm->rtm_family = AF_INET6;
3141         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3142         rtm->rtm_src_len = rt->rt6i_src.plen;
3143         rtm->rtm_tos = 0;
3144         if (rt->rt6i_table)
3145                 table = rt->rt6i_table->tb6_id;
3146         else
3147                 table = RT6_TABLE_UNSPEC;
3148         rtm->rtm_table = table;
3149         if (nla_put_u32(skb, RTA_TABLE, table))
3150                 goto nla_put_failure;
3151         if (rt->rt6i_flags & RTF_REJECT) {
3152                 switch (rt->dst.error) {
3153                 case -EINVAL:
3154                         rtm->rtm_type = RTN_BLACKHOLE;
3155                         break;
3156                 case -EACCES:
3157                         rtm->rtm_type = RTN_PROHIBIT;
3158                         break;
3159                 case -EAGAIN:
3160                         rtm->rtm_type = RTN_THROW;
3161                         break;
3162                 default:
3163                         rtm->rtm_type = RTN_UNREACHABLE;
3164                         break;
3165                 }
3166         }
3167         else if (rt->rt6i_flags & RTF_LOCAL)
3168                 rtm->rtm_type = RTN_LOCAL;
3169         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3170                 rtm->rtm_type = RTN_LOCAL;
3171         else
3172                 rtm->rtm_type = RTN_UNICAST;
3173         rtm->rtm_flags = 0;
3174         if (!netif_carrier_ok(rt->dst.dev)) {
3175                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3176                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3177                         rtm->rtm_flags |= RTNH_F_DEAD;
3178         }
3179         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3180         rtm->rtm_protocol = rt->rt6i_protocol;
3181         if (rt->rt6i_flags & RTF_DYNAMIC)
3182                 rtm->rtm_protocol = RTPROT_REDIRECT;
3183         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3184                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3185                         rtm->rtm_protocol = RTPROT_RA;
3186                 else
3187                         rtm->rtm_protocol = RTPROT_KERNEL;
3188         }
3189
3190         if (rt->rt6i_flags & RTF_CACHE)
3191                 rtm->rtm_flags |= RTM_F_CLONED;
3192
3193         if (dst) {
3194                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3195                         goto nla_put_failure;
3196                 rtm->rtm_dst_len = 128;
3197         } else if (rtm->rtm_dst_len)
3198                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3199                         goto nla_put_failure;
3200 #ifdef CONFIG_IPV6_SUBTREES
3201         if (src) {
3202                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3203                         goto nla_put_failure;
3204                 rtm->rtm_src_len = 128;
3205         } else if (rtm->rtm_src_len &&
3206                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3207                 goto nla_put_failure;
3208 #endif
3209         if (iif) {
3210 #ifdef CONFIG_IPV6_MROUTE
3211                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3212                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3213                         if (err <= 0) {
3214                                 if (!nowait) {
3215                                         if (err == 0)
3216                                                 return 0;
3217                                         goto nla_put_failure;
3218                                 } else {
3219                                         if (err == -EMSGSIZE)
3220                                                 goto nla_put_failure;
3221                                 }
3222                         }
3223                 } else
3224 #endif
3225                         if (nla_put_u32(skb, RTA_IIF, iif))
3226                                 goto nla_put_failure;
3227         } else if (dst) {
3228                 struct in6_addr saddr_buf;
3229                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3230                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3231                         goto nla_put_failure;
3232         }
3233
3234         if (rt->rt6i_prefsrc.plen) {
3235                 struct in6_addr saddr_buf;
3236                 saddr_buf = rt->rt6i_prefsrc.addr;
3237                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3238                         goto nla_put_failure;
3239         }
3240
3241         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3242         if (rt->rt6i_pmtu)
3243                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3244         if (rtnetlink_put_metrics(skb, metrics) < 0)
3245                 goto nla_put_failure;
3246
3247         if (rt->rt6i_flags & RTF_GATEWAY) {
3248                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3249                         goto nla_put_failure;
3250         }
3251
3252         if (rt->dst.dev &&
3253             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3254                 goto nla_put_failure;
3255         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3256                 goto nla_put_failure;
3257
3258         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3259
3260         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3261                 goto nla_put_failure;
3262
3263         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3264                 goto nla_put_failure;
3265
3266         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3267
3268         nlmsg_end(skb, nlh);
3269         return 0;
3270
3271 nla_put_failure:
3272         nlmsg_cancel(skb, nlh);
3273         return -EMSGSIZE;
3274 }
3275
3276 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3277 {
3278         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3279         int prefix;
3280
3281         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3282                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3283                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3284         } else
3285                 prefix = 0;
3286
3287         return rt6_fill_node(arg->net,
3288                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3289                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3290                      prefix, 0, NLM_F_MULTI);
3291 }
3292
3293 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3294 {
3295         struct net *net = sock_net(in_skb->sk);
3296         struct nlattr *tb[RTA_MAX+1];
3297         struct rt6_info *rt;
3298         struct sk_buff *skb;
3299         struct rtmsg *rtm;
3300         struct flowi6 fl6;
3301         int err, iif = 0, oif = 0;
3302
3303         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3304         if (err < 0)
3305                 goto errout;
3306
3307         err = -EINVAL;
3308         memset(&fl6, 0, sizeof(fl6));
3309
3310         if (tb[RTA_SRC]) {
3311                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3312                         goto errout;
3313
3314                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3315         }
3316
3317         if (tb[RTA_DST]) {
3318                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3319                         goto errout;
3320
3321                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3322         }
3323
3324         if (tb[RTA_IIF])
3325                 iif = nla_get_u32(tb[RTA_IIF]);
3326
3327         if (tb[RTA_OIF])
3328                 oif = nla_get_u32(tb[RTA_OIF]);
3329
3330         if (tb[RTA_MARK])
3331                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3332
3333         if (iif) {
3334                 struct net_device *dev;
3335                 int flags = 0;
3336
3337                 dev = __dev_get_by_index(net, iif);
3338                 if (!dev) {
3339                         err = -ENODEV;
3340                         goto errout;
3341                 }
3342
3343                 fl6.flowi6_iif = iif;
3344
3345                 if (!ipv6_addr_any(&fl6.saddr))
3346                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3347
3348                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3349                                                                flags);
3350         } else {
3351                 fl6.flowi6_oif = oif;
3352
3353                 if (netif_index_is_l3_master(net, oif)) {
3354                         fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3355                                            FLOWI_FLAG_SKIP_NH_OIF;
3356                 }
3357
3358                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3359         }
3360
3361         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3362         if (!skb) {
3363                 ip6_rt_put(rt);
3364                 err = -ENOBUFS;
3365                 goto errout;
3366         }
3367
3368         /* Reserve room for dummy headers, this skb can pass
3369            through good chunk of routing engine.
3370          */
3371         skb_reset_mac_header(skb);
3372         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3373
3374         skb_dst_set(skb, &rt->dst);
3375
3376         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3377                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3378                             nlh->nlmsg_seq, 0, 0, 0);
3379         if (err < 0) {
3380                 kfree_skb(skb);
3381                 goto errout;
3382         }
3383
3384         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3385 errout:
3386         return err;
3387 }
3388
3389 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3390                      unsigned int nlm_flags)
3391 {
3392         struct sk_buff *skb;
3393         struct net *net = info->nl_net;
3394         u32 seq;
3395         int err;
3396
3397         err = -ENOBUFS;
3398         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3399
3400         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3401         if (!skb)
3402                 goto errout;
3403
3404         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3405                                 event, info->portid, seq, 0, 0, nlm_flags);
3406         if (err < 0) {
3407                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3408                 WARN_ON(err == -EMSGSIZE);
3409                 kfree_skb(skb);
3410                 goto errout;
3411         }
3412         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3413                     info->nlh, gfp_any());
3414         return;
3415 errout:
3416         if (err < 0)
3417                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3418 }
3419
3420 static int ip6_route_dev_notify(struct notifier_block *this,
3421                                 unsigned long event, void *ptr)
3422 {
3423         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3424         struct net *net = dev_net(dev);
3425
3426         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3427                 net->ipv6.ip6_null_entry->dst.dev = dev;
3428                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3429 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3430                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3431                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3432                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3433                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3434 #endif
3435         }
3436
3437         return NOTIFY_OK;
3438 }
3439
3440 /*
3441  *      /proc
3442  */
3443
3444 #ifdef CONFIG_PROC_FS
3445
3446 static const struct file_operations ipv6_route_proc_fops = {
3447         .owner          = THIS_MODULE,
3448         .open           = ipv6_route_open,
3449         .read           = seq_read,
3450         .llseek         = seq_lseek,
3451         .release        = seq_release_net,
3452 };
3453
3454 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3455 {
3456         struct net *net = (struct net *)seq->private;
3457         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3458                    net->ipv6.rt6_stats->fib_nodes,
3459                    net->ipv6.rt6_stats->fib_route_nodes,
3460                    net->ipv6.rt6_stats->fib_rt_alloc,
3461                    net->ipv6.rt6_stats->fib_rt_entries,
3462                    net->ipv6.rt6_stats->fib_rt_cache,
3463                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3464                    net->ipv6.rt6_stats->fib_discarded_routes);
3465
3466         return 0;
3467 }
3468
3469 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3470 {
3471         return single_open_net(inode, file, rt6_stats_seq_show);
3472 }
3473
3474 static const struct file_operations rt6_stats_seq_fops = {
3475         .owner   = THIS_MODULE,
3476         .open    = rt6_stats_seq_open,
3477         .read    = seq_read,
3478         .llseek  = seq_lseek,
3479         .release = single_release_net,
3480 };
3481 #endif  /* CONFIG_PROC_FS */
3482
3483 #ifdef CONFIG_SYSCTL
3484
3485 static
3486 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3487                               void __user *buffer, size_t *lenp, loff_t *ppos)
3488 {
3489         struct net *net;
3490         int delay;
3491         if (!write)
3492                 return -EINVAL;
3493
3494         net = (struct net *)ctl->extra1;
3495         delay = net->ipv6.sysctl.flush_delay;
3496         proc_dointvec(ctl, write, buffer, lenp, ppos);
3497         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3498         return 0;
3499 }
3500
3501 struct ctl_table ipv6_route_table_template[] = {
3502         {
3503                 .procname       =       "flush",
3504                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3505                 .maxlen         =       sizeof(int),
3506                 .mode           =       0200,
3507                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3508         },
3509         {
3510                 .procname       =       "gc_thresh",
3511                 .data           =       &ip6_dst_ops_template.gc_thresh,
3512                 .maxlen         =       sizeof(int),
3513                 .mode           =       0644,
3514                 .proc_handler   =       proc_dointvec,
3515         },
3516         {
3517                 .procname       =       "max_size",
3518                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3519                 .maxlen         =       sizeof(int),
3520                 .mode           =       0644,
3521                 .proc_handler   =       proc_dointvec,
3522         },
3523         {
3524                 .procname       =       "gc_min_interval",
3525                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3526                 .maxlen         =       sizeof(int),
3527                 .mode           =       0644,
3528                 .proc_handler   =       proc_dointvec_jiffies,
3529         },
3530         {
3531                 .procname       =       "gc_timeout",
3532                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3533                 .maxlen         =       sizeof(int),
3534                 .mode           =       0644,
3535                 .proc_handler   =       proc_dointvec_jiffies,
3536         },
3537         {
3538                 .procname       =       "gc_interval",
3539                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3540                 .maxlen         =       sizeof(int),
3541                 .mode           =       0644,
3542                 .proc_handler   =       proc_dointvec_jiffies,
3543         },
3544         {
3545                 .procname       =       "gc_elasticity",
3546                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3547                 .maxlen         =       sizeof(int),
3548                 .mode           =       0644,
3549                 .proc_handler   =       proc_dointvec,
3550         },
3551         {
3552                 .procname       =       "mtu_expires",
3553                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3554                 .maxlen         =       sizeof(int),
3555                 .mode           =       0644,
3556                 .proc_handler   =       proc_dointvec_jiffies,
3557         },
3558         {
3559                 .procname       =       "min_adv_mss",
3560                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3561                 .maxlen         =       sizeof(int),
3562                 .mode           =       0644,
3563                 .proc_handler   =       proc_dointvec,
3564         },
3565         {
3566                 .procname       =       "gc_min_interval_ms",
3567                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3568                 .maxlen         =       sizeof(int),
3569                 .mode           =       0644,
3570                 .proc_handler   =       proc_dointvec_ms_jiffies,
3571         },
3572         { }
3573 };
3574
3575 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3576 {
3577         struct ctl_table *table;
3578
3579         table = kmemdup(ipv6_route_table_template,
3580                         sizeof(ipv6_route_table_template),
3581                         GFP_KERNEL);
3582
3583         if (table) {
3584                 table[0].data = &net->ipv6.sysctl.flush_delay;
3585                 table[0].extra1 = net;
3586                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3587                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3588                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3589                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3590                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3591                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3592                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3593                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3594                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3595
3596                 /* Don't export sysctls to unprivileged users */
3597                 if (net->user_ns != &init_user_ns)
3598                         table[0].procname = NULL;
3599         }
3600
3601         return table;
3602 }
3603 #endif
3604
3605 static int __net_init ip6_route_net_init(struct net *net)
3606 {
3607         int ret = -ENOMEM;
3608
3609         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3610                sizeof(net->ipv6.ip6_dst_ops));
3611
3612         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3613                 goto out_ip6_dst_ops;
3614
3615         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3616                                            sizeof(*net->ipv6.ip6_null_entry),
3617                                            GFP_KERNEL);
3618         if (!net->ipv6.ip6_null_entry)
3619                 goto out_ip6_dst_entries;
3620         net->ipv6.ip6_null_entry->dst.path =
3621                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3622         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3623         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3624                          ip6_template_metrics, true);
3625
3626 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3627         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3628                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3629                                                GFP_KERNEL);
3630         if (!net->ipv6.ip6_prohibit_entry)
3631                 goto out_ip6_null_entry;
3632         net->ipv6.ip6_prohibit_entry->dst.path =
3633                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3634         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3635         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3636                          ip6_template_metrics, true);
3637
3638         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3639                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3640                                                GFP_KERNEL);
3641         if (!net->ipv6.ip6_blk_hole_entry)
3642                 goto out_ip6_prohibit_entry;
3643         net->ipv6.ip6_blk_hole_entry->dst.path =
3644                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3645         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3646         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3647                          ip6_template_metrics, true);
3648 #endif
3649
3650         net->ipv6.sysctl.flush_delay = 0;
3651         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3652         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3653         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3654         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3655         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3656         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3657         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3658
3659         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3660
3661         ret = 0;
3662 out:
3663         return ret;
3664
3665 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3666 out_ip6_prohibit_entry:
3667         kfree(net->ipv6.ip6_prohibit_entry);
3668 out_ip6_null_entry:
3669         kfree(net->ipv6.ip6_null_entry);
3670 #endif
3671 out_ip6_dst_entries:
3672         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3673 out_ip6_dst_ops:
3674         goto out;
3675 }
3676
3677 static void __net_exit ip6_route_net_exit(struct net *net)
3678 {
3679         kfree(net->ipv6.ip6_null_entry);
3680 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3681         kfree(net->ipv6.ip6_prohibit_entry);
3682         kfree(net->ipv6.ip6_blk_hole_entry);
3683 #endif
3684         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3685 }
3686
3687 static int __net_init ip6_route_net_init_late(struct net *net)
3688 {
3689 #ifdef CONFIG_PROC_FS
3690         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3691         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3692 #endif
3693         return 0;
3694 }
3695
3696 static void __net_exit ip6_route_net_exit_late(struct net *net)
3697 {
3698 #ifdef CONFIG_PROC_FS
3699         remove_proc_entry("ipv6_route", net->proc_net);
3700         remove_proc_entry("rt6_stats", net->proc_net);
3701 #endif
3702 }
3703
3704 static struct pernet_operations ip6_route_net_ops = {
3705         .init = ip6_route_net_init,
3706         .exit = ip6_route_net_exit,
3707 };
3708
3709 static int __net_init ipv6_inetpeer_init(struct net *net)
3710 {
3711         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3712
3713         if (!bp)
3714                 return -ENOMEM;
3715         inet_peer_base_init(bp);
3716         net->ipv6.peers = bp;
3717         return 0;
3718 }
3719
3720 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3721 {
3722         struct inet_peer_base *bp = net->ipv6.peers;
3723
3724         net->ipv6.peers = NULL;
3725         inetpeer_invalidate_tree(bp);
3726         kfree(bp);
3727 }
3728
3729 static struct pernet_operations ipv6_inetpeer_ops = {
3730         .init   =       ipv6_inetpeer_init,
3731         .exit   =       ipv6_inetpeer_exit,
3732 };
3733
3734 static struct pernet_operations ip6_route_net_late_ops = {
3735         .init = ip6_route_net_init_late,
3736         .exit = ip6_route_net_exit_late,
3737 };
3738
3739 static struct notifier_block ip6_route_dev_notifier = {
3740         .notifier_call = ip6_route_dev_notify,
3741         .priority = 0,
3742 };
3743
3744 int __init ip6_route_init(void)
3745 {
3746         int ret;
3747         int cpu;
3748
3749         ret = -ENOMEM;
3750         ip6_dst_ops_template.kmem_cachep =
3751                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3752                                   SLAB_HWCACHE_ALIGN, NULL);
3753         if (!ip6_dst_ops_template.kmem_cachep)
3754                 goto out;
3755
3756         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3757         if (ret)
3758                 goto out_kmem_cache;
3759
3760         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3761         if (ret)
3762                 goto out_dst_entries;
3763
3764         ret = register_pernet_subsys(&ip6_route_net_ops);
3765         if (ret)
3766                 goto out_register_inetpeer;
3767
3768         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3769
3770         /* Registering of the loopback is done before this portion of code,
3771          * the loopback reference in rt6_info will not be taken, do it
3772          * manually for init_net */
3773         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3774         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3775   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3776         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3777         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3778         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3779         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3780   #endif
3781         ret = fib6_init();
3782         if (ret)
3783                 goto out_register_subsys;
3784
3785         ret = xfrm6_init();
3786         if (ret)
3787                 goto out_fib6_init;
3788
3789         ret = fib6_rules_init();
3790         if (ret)
3791                 goto xfrm6_init;
3792
3793         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3794         if (ret)
3795                 goto fib6_rules_init;
3796
3797         ret = -ENOBUFS;
3798         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3799             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3800             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3801                 goto out_register_late_subsys;
3802
3803         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3804         if (ret)
3805                 goto out_register_late_subsys;
3806
3807         for_each_possible_cpu(cpu) {
3808                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3809
3810                 INIT_LIST_HEAD(&ul->head);
3811                 spin_lock_init(&ul->lock);
3812         }
3813
3814 out:
3815         return ret;
3816
3817 out_register_late_subsys:
3818         unregister_pernet_subsys(&ip6_route_net_late_ops);
3819 fib6_rules_init:
3820         fib6_rules_cleanup();
3821 xfrm6_init:
3822         xfrm6_fini();
3823 out_fib6_init:
3824         fib6_gc_cleanup();
3825 out_register_subsys:
3826         unregister_pernet_subsys(&ip6_route_net_ops);
3827 out_register_inetpeer:
3828         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3829 out_dst_entries:
3830         dst_entries_destroy(&ip6_dst_blackhole_ops);
3831 out_kmem_cache:
3832         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3833         goto out;
3834 }
3835
3836 void ip6_route_cleanup(void)
3837 {
3838         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3839         unregister_pernet_subsys(&ip6_route_net_late_ops);
3840         fib6_rules_cleanup();
3841         xfrm6_fini();
3842         fib6_gc_cleanup();
3843         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3844         unregister_pernet_subsys(&ip6_route_net_ops);
3845         dst_entries_destroy(&ip6_dst_blackhole_ops);
3846         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3847 }