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 static 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
 368 static void ip6_dst_destroy(struct dst_entry *dst)
 369 {
 370         struct rt6_info *rt = (struct rt6_info *)dst;
 371         struct dst_entry *from = dst->from;
 372         struct inet6_dev *idev;
 373
 374         dst_destroy_metrics_generic(dst);
 375         free_percpu(rt->rt6i_pcpu);
 376         rt6_uncached_list_del(rt);
 377
 378         idev = rt->rt6i_idev;
 379         if (idev) {
 380                 rt->rt6i_idev = NULL;
 381                 in6_dev_put(idev);
 382         }
 383
 384         dst->from = NULL;
 385         dst_release(from);
 386 }
 387
 388 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 389                            int how)
 390 {
 391         struct rt6_info *rt = (struct rt6_info *)dst;
 392         struct inet6_dev *idev = rt->rt6i_idev;
 393         struct net_device *loopback_dev =
 394                 dev_net(dev)->loopback_dev;
 395
 396         if (dev != loopback_dev) {
 397                 if (idev && idev->dev == dev) {
 398                         struct inet6_dev *loopback_idev =
 399                                 in6_dev_get(loopback_dev);
 400                         if (loopback_idev) {
 401                                 rt->rt6i_idev = loopback_idev;
 402                                 in6_dev_put(idev);
 403                         }
 404                 }
 405         }
 406 }
 407
 408 static bool __rt6_check_expired(const struct rt6_info *rt)
 409 {
 410         if (rt->rt6i_flags & RTF_EXPIRES)
 411                 return time_after(jiffies, rt->dst.expires);
 412         else
 413                 return false;
 414 }
 415
 416 static bool rt6_check_expired(const struct rt6_info *rt)
 417 {
 418         if (rt->rt6i_flags & RTF_EXPIRES) {
 419                 if (time_after(jiffies, rt->dst.expires))
 420                         return true;
 421         } else if (rt->dst.from) {
 422                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 423         }
 424         return false;
 425 }
 426
 427 /* Multipath route selection:
 428  *   Hash based function using packet header and flowlabel.
 429  * Adapted from fib_info_hashfn()
 430  */
 431 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 432                                const struct flowi6 *fl6)
 433 {
 434         return get_hash_from_flowi6(fl6) % candidate_count;
 435 }
 436
 437 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 438                                              struct flowi6 *fl6, int oif,
 439                                              int strict)
 440 {
 441         struct rt6_info *sibling, *next_sibling;
 442         int route_choosen;
 443
 444         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 445         /* Don't change the route, if route_choosen == 0
 446          * (siblings does not include ourself)
 447          */
 448         if (route_choosen)
 449                 list_for_each_entry_safe(sibling, next_sibling,
 450                                 &match->rt6i_siblings, rt6i_siblings) {
 451                         route_choosen--;
 452                         if (route_choosen == 0) {
 453                                 if (rt6_score_route(sibling, oif, strict) < 0)
 454                                         break;
 455                                 match = sibling;
 456                                 break;
 457                         }
 458                 }
 459         return match;
 460 }
 461
 462 /*
 463  *      Route lookup. Any table->tb6_lock is implied.
 464  */
 465
 466 static inline struct rt6_info *rt6_device_match(struct net *net,
 467                                                     struct rt6_info *rt,
 468                                                     const struct in6_addr *saddr,
 469                                                     int oif,
 470                                                     int flags)
 471 {
 472         struct rt6_info *local = NULL;
 473         struct rt6_info *sprt;
 474
 475         if (!oif && ipv6_addr_any(saddr))
 476                 goto out;
 477
 478         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 479                 struct net_device *dev = sprt->dst.dev;
 480
 481                 if (oif) {
 482                         if (dev->ifindex == oif)
 483                                 return sprt;
 484                         if (dev->flags & IFF_LOOPBACK) {
 485                                 if (!sprt->rt6i_idev ||
 486                                     sprt->rt6i_idev->dev->ifindex != oif) {
 487                                         if (flags & RT6_LOOKUP_F_IFACE)
 488                                                 continue;
 489                                         if (local &&
 490                                             local->rt6i_idev->dev->ifindex == oif)
 491                                                 continue;
 492                                 }
 493                                 local = sprt;
 494                         }
 495                 } else {
 496                         if (ipv6_chk_addr(net, saddr, dev,
 497                                           flags & RT6_LOOKUP_F_IFACE))
 498                                 return sprt;
 499                 }
 500         }
 501
 502         if (oif) {
 503                 if (local)
 504                         return local;
 505
 506                 if (flags & RT6_LOOKUP_F_IFACE)
 507                         return net->ipv6.ip6_null_entry;
 508         }
 509 out:
 510         return rt;
 511 }
 512
 513 #ifdef CONFIG_IPV6_ROUTER_PREF
 514 struct __rt6_probe_work {
 515         struct work_struct work;
 516         struct in6_addr target;
 517         struct net_device *dev;
 518 };
 519
 520 static void rt6_probe_deferred(struct work_struct *w)
 521 {
 522         struct in6_addr mcaddr;
 523         struct __rt6_probe_work *work =
 524                 container_of(w, struct __rt6_probe_work, work);
 525
 526         addrconf_addr_solict_mult(&work->target, &mcaddr);
 527         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL);
 528         dev_put(work->dev);
 529         kfree(work);
 530 }
 531
 532 static void rt6_probe(struct rt6_info *rt)
 533 {
 534         struct __rt6_probe_work *work;
 535         struct neighbour *neigh;
 536         /*
 537          * Okay, this does not seem to be appropriate
 538          * for now, however, we need to check if it
 539          * is really so; aka Router Reachability Probing.
 540          *
 541          * Router Reachability Probe MUST be rate-limited
 542          * to no more than one per minute.
 543          */
 544         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 545                 return;
 546         rcu_read_lock_bh();
 547         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 548         if (neigh) {
 549                 if (neigh->nud_state & NUD_VALID)
 550                         goto out;
 551
 552                 work = NULL;
 553                 write_lock(&neigh->lock);
 554                 if (!(neigh->nud_state & NUD_VALID) &&
 555                     time_after(jiffies,
 556                                neigh->updated +
 557                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 558                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 559                         if (work)
 560                                 __neigh_set_probe_once(neigh);
 561                 }
 562                 write_unlock(&neigh->lock);
 563         } else {
 564                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 565         }
 566
 567         if (work) {
 568                 INIT_WORK(&work->work, rt6_probe_deferred);
 569                 work->target = rt->rt6i_gateway;
 570                 dev_hold(rt->dst.dev);
 571                 work->dev = rt->dst.dev;
 572                 schedule_work(&work->work);
 573         }
 574
 575 out:
 576         rcu_read_unlock_bh();
 577 }
 578 #else
 579 static inline void rt6_probe(struct rt6_info *rt)
 580 {
 581 }
 582 #endif
 583
 584 /*
 585  * Default Router Selection (RFC 2461 6.3.6)
 586  */
 587 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 588 {
 589         struct net_device *dev = rt->dst.dev;
 590         if (!oif || dev->ifindex == oif)
 591                 return 2;
 592         if ((dev->flags & IFF_LOOPBACK) &&
 593             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 594                 return 1;
 595         return 0;
 596 }
 597
 598 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 599 {
 600         struct neighbour *neigh;
 601         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 602
 603         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 604             !(rt->rt6i_flags & RTF_GATEWAY))
 605                 return RT6_NUD_SUCCEED;
 606
 607         rcu_read_lock_bh();
 608         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 609         if (neigh) {
 610                 read_lock(&neigh->lock);
 611                 if (neigh->nud_state & NUD_VALID)
 612                         ret = RT6_NUD_SUCCEED;
 613 #ifdef CONFIG_IPV6_ROUTER_PREF
 614                 else if (!(neigh->nud_state & NUD_FAILED))
 615                         ret = RT6_NUD_SUCCEED;
 616                 else
 617                         ret = RT6_NUD_FAIL_PROBE;
 618 #endif
 619                 read_unlock(&neigh->lock);
 620         } else {
 621                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 622                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 623         }
 624         rcu_read_unlock_bh();
 625
 626         return ret;
 627 }
 628
 629 static int rt6_score_route(struct rt6_info *rt, int oif,
 630                            int strict)
 631 {
 632         int m;
 633
 634         m = rt6_check_dev(rt, oif);
 635         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 636                 return RT6_NUD_FAIL_HARD;
 637 #ifdef CONFIG_IPV6_ROUTER_PREF
 638         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 639 #endif
 640         if (strict & RT6_LOOKUP_F_REACHABLE) {
 641                 int n = rt6_check_neigh(rt);
 642                 if (n < 0)
 643                         return n;
 644         }
 645         return m;
 646 }
 647
 648 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 649                                    int *mpri, struct rt6_info *match,
 650                                    bool *do_rr)
 651 {
 652         int m;
 653         bool match_do_rr = false;
 654         struct inet6_dev *idev = rt->rt6i_idev;
 655         struct net_device *dev = rt->dst.dev;
 656
 657         if (dev && !netif_carrier_ok(dev) &&
 658             idev->cnf.ignore_routes_with_linkdown)
 659                 goto out;
 660
 661         if (rt6_check_expired(rt))
 662                 goto out;
 663
 664         m = rt6_score_route(rt, oif, strict);
 665         if (m == RT6_NUD_FAIL_DO_RR) {
 666                 match_do_rr = true;
 667                 m = 0; /* lowest valid score */
 668         } else if (m == RT6_NUD_FAIL_HARD) {
 669                 goto out;
 670         }
 671
 672         if (strict & RT6_LOOKUP_F_REACHABLE)
 673                 rt6_probe(rt);
 674
 675         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 676         if (m > *mpri) {
 677                 *do_rr = match_do_rr;
 678                 *mpri = m;
 679                 match = rt;
 680         }
 681 out:
 682         return match;
 683 }
 684
 685 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 686                                      struct rt6_info *rr_head,
 687                                      u32 metric, int oif, int strict,
 688                                      bool *do_rr)
 689 {
 690         struct rt6_info *rt, *match, *cont;
 691         int mpri = -1;
 692
 693         match = NULL;
 694         cont = NULL;
 695         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 696                 if (rt->rt6i_metric != metric) {
 697                         cont = rt;
 698                         break;
 699                 }
 700
 701                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 702         }
 703
 704         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 705                 if (rt->rt6i_metric != metric) {
 706                         cont = rt;
 707                         break;
 708                 }
 709
 710                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 711         }
 712
 713         if (match || !cont)
 714                 return match;
 715
 716         for (rt = cont; rt; rt = rt->dst.rt6_next)
 717                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 718
 719         return match;
 720 }
 721
 722 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 723 {
 724         struct rt6_info *match, *rt0;
 725         struct net *net;
 726         bool do_rr = false;
 727
 728         rt0 = fn->rr_ptr;
 729         if (!rt0)
 730                 fn->rr_ptr = rt0 = fn->leaf;
 731
 732         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 733                              &do_rr);
 734
 735         if (do_rr) {
 736                 struct rt6_info *next = rt0->dst.rt6_next;
 737
 738                 /* no entries matched; do round-robin */
 739                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 740                         next = fn->leaf;
 741
 742                 if (next != rt0)
 743                         fn->rr_ptr = next;
 744         }
 745
 746         net = dev_net(rt0->dst.dev);
 747         return match ? match : net->ipv6.ip6_null_entry;
 748 }
 749
 750 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 751 {
 752         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 753 }
 754
 755 #ifdef CONFIG_IPV6_ROUTE_INFO
 756 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 757                   const struct in6_addr *gwaddr)
 758 {
 759         struct net *net = dev_net(dev);
 760         struct route_info *rinfo = (struct route_info *) opt;
 761         struct in6_addr prefix_buf, *prefix;
 762         unsigned int pref;
 763         unsigned long lifetime;
 764         struct rt6_info *rt;
 765
 766         if (len < sizeof(struct route_info)) {
 767                 return -EINVAL;
 768         }
 769
 770         /* Sanity check for prefix_len and length */
 771         if (rinfo->length > 3) {
 772                 return -EINVAL;
 773         } else if (rinfo->prefix_len > 128) {
 774                 return -EINVAL;
 775         } else if (rinfo->prefix_len > 64) {
 776                 if (rinfo->length < 2) {
 777                         return -EINVAL;
 778                 }
 779         } else if (rinfo->prefix_len > 0) {
 780                 if (rinfo->length < 1) {
 781                         return -EINVAL;
 782                 }
 783         }
 784
 785         pref = rinfo->route_pref;
 786         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 787                 return -EINVAL;
 788
 789         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 790
 791         if (rinfo->length == 3)
 792                 prefix = (struct in6_addr *)rinfo->prefix;
 793         else {
 794                 /* this function is safe */
 795                 ipv6_addr_prefix(&prefix_buf,
 796                                  (struct in6_addr *)rinfo->prefix,
 797                                  rinfo->prefix_len);
 798                 prefix = &prefix_buf;
 799         }
 800
 801         if (rinfo->prefix_len == 0)
 802                 rt = rt6_get_dflt_router(gwaddr, dev);
 803         else
 804                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 805                                         gwaddr, dev->ifindex);
 806
 807         if (rt && !lifetime) {
 808                 ip6_del_rt(rt);
 809                 rt = NULL;
 810         }
 811
 812         if (!rt && lifetime)
 813                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr, dev->ifindex,
 814                                         pref);
 815         else if (rt)
 816                 rt->rt6i_flags = RTF_ROUTEINFO |
 817                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 818
 819         if (rt) {
 820                 if (!addrconf_finite_timeout(lifetime))
 821                         rt6_clean_expires(rt);
 822                 else
 823                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 824
 825                 ip6_rt_put(rt);
 826         }
 827         return 0;
 828 }
 829 #endif
 830
 831 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 832                                         struct in6_addr *saddr)
 833 {
 834         struct fib6_node *pn;
 835         while (1) {
 836                 if (fn->fn_flags & RTN_TL_ROOT)
 837                         return NULL;
 838                 pn = fn->parent;
 839                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 840                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 841                 else
 842                         fn = pn;
 843                 if (fn->fn_flags & RTN_RTINFO)
 844                         return fn;
 845         }
 846 }
 847
 848 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 849                                              struct fib6_table *table,
 850                                              struct flowi6 *fl6, int flags)
 851 {
 852         struct fib6_node *fn;
 853         struct rt6_info *rt;
 854
 855         read_lock_bh(&table->tb6_lock);
 856         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 857 restart:
 858         rt = fn->leaf;
 859         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 860         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 861                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 862         if (rt == net->ipv6.ip6_null_entry) {
 863                 fn = fib6_backtrack(fn, &fl6->saddr);
 864                 if (fn)
 865                         goto restart;
 866         }
 867         dst_use(&rt->dst, jiffies);
 868         read_unlock_bh(&table->tb6_lock);
 869
 870         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
 871
 872         return rt;
 873
 874 }
 875
 876 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 877                                     int flags)
 878 {
 879         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 880 }
 881 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 882
 883 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 884                             const struct in6_addr *saddr, int oif, int strict)
 885 {
 886         struct flowi6 fl6 = {
 887                 .flowi6_oif = oif,
 888                 .daddr = *daddr,
 889         };
 890         struct dst_entry *dst;
 891         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 892
 893         if (saddr) {
 894                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 895                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 896         }
 897
 898         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 899         if (dst->error == 0)
 900                 return (struct rt6_info *) dst;
 901
 902         dst_release(dst);
 903
 904         return NULL;
 905 }
 906 EXPORT_SYMBOL(rt6_lookup);
 907
 908 /* ip6_ins_rt is called with FREE table->tb6_lock.
 909    It takes new route entry, the addition fails by any reason the
 910    route is freed. In any case, if caller does not hold it, it may
 911    be destroyed.
 912  */
 913
 914 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 915                         struct mx6_config *mxc)
 916 {
 917         int err;
 918         struct fib6_table *table;
 919
 920         table = rt->rt6i_table;
 921         write_lock_bh(&table->tb6_lock);
 922         err = fib6_add(&table->tb6_root, rt, info, mxc);
 923         write_unlock_bh(&table->tb6_lock);
 924
 925         return err;
 926 }
 927
 928 int ip6_ins_rt(struct rt6_info *rt)
 929 {
 930         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 931         struct mx6_config mxc = { .mx = NULL, };
 932
 933         return __ip6_ins_rt(rt, &info, &mxc);
 934 }
 935
 936 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 937                                            const struct in6_addr *daddr,
 938                                            const struct in6_addr *saddr)
 939 {
 940         struct rt6_info *rt;
 941
 942         /*
 943          *      Clone the route.
 944          */
 945
 946         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 947                 ort = (struct rt6_info *)ort->dst.from;
 948
 949         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 950
 951         if (!rt)
 952                 return NULL;
 953
 954         ip6_rt_copy_init(rt, ort);
 955         rt->rt6i_flags |= RTF_CACHE;
 956         rt->rt6i_metric = 0;
 957         rt->dst.flags |= DST_HOST;
 958         rt->rt6i_dst.addr = *daddr;
 959         rt->rt6i_dst.plen = 128;
 960
 961         if (!rt6_is_gw_or_nonexthop(ort)) {
 962                 if (ort->rt6i_dst.plen != 128 &&
 963                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 964                         rt->rt6i_flags |= RTF_ANYCAST;
 965 #ifdef CONFIG_IPV6_SUBTREES
 966                 if (rt->rt6i_src.plen && saddr) {
 967                         rt->rt6i_src.addr = *saddr;
 968                         rt->rt6i_src.plen = 128;
 969                 }
 970 #endif
 971         }
 972
 973         return rt;
 974 }
 975
 976 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
 977 {
 978         struct rt6_info *pcpu_rt;
 979
 980         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
 981                                   rt->dst.dev, rt->dst.flags);
 982
 983         if (!pcpu_rt)
 984                 return NULL;
 985         ip6_rt_copy_init(pcpu_rt, rt);
 986         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
 987         pcpu_rt->rt6i_flags |= RTF_PCPU;
 988         return pcpu_rt;
 989 }
 990
 991 /* It should be called with read_lock_bh(&tb6_lock) acquired */
 992 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
 993 {
 994         struct rt6_info *pcpu_rt, **p;
 995
 996         p = this_cpu_ptr(rt->rt6i_pcpu);
 997         pcpu_rt = *p;
 998
 999         if (pcpu_rt) {
1000                 dst_hold(&pcpu_rt->dst);
1001                 rt6_dst_from_metrics_check(pcpu_rt);
1002         }
1003         return pcpu_rt;
1004 }
1005
1006 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1007 {
1008         struct fib6_table *table = rt->rt6i_table;
1009         struct rt6_info *pcpu_rt, *prev, **p;
1010
1011         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1012         if (!pcpu_rt) {
1013                 struct net *net = dev_net(rt->dst.dev);
1014
1015                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1016                 return net->ipv6.ip6_null_entry;
1017         }
1018
1019         read_lock_bh(&table->tb6_lock);
1020         if (rt->rt6i_pcpu) {
1021                 p = this_cpu_ptr(rt->rt6i_pcpu);
1022                 prev = cmpxchg(p, NULL, pcpu_rt);
1023                 if (prev) {
1024                         /* If someone did it before us, return prev instead */
1025                         dst_destroy(&pcpu_rt->dst);
1026                         pcpu_rt = prev;
1027                 }
1028         } else {
1029                 /* rt has been removed from the fib6 tree
1030                  * before we have a chance to acquire the read_lock.
1031                  * In this case, don't brother to create a pcpu rt
1032                  * since rt is going away anyway.  The next
1033                  * dst_check() will trigger a re-lookup.
1034                  */
1035                 dst_destroy(&pcpu_rt->dst);
1036                 pcpu_rt = rt;
1037         }
1038         dst_hold(&pcpu_rt->dst);
1039         rt6_dst_from_metrics_check(pcpu_rt);
1040         read_unlock_bh(&table->tb6_lock);
1041         return pcpu_rt;
1042 }
1043
1044 static struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table, int oif,
1045                                       struct flowi6 *fl6, int flags)
1046 {
1047         struct fib6_node *fn, *saved_fn;
1048         struct rt6_info *rt;
1049         int strict = 0;
1050
1051         strict |= flags & RT6_LOOKUP_F_IFACE;
1052         if (net->ipv6.devconf_all->forwarding == 0)
1053                 strict |= RT6_LOOKUP_F_REACHABLE;
1054
1055         read_lock_bh(&table->tb6_lock);
1056
1057         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1058         saved_fn = fn;
1059
1060         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1061                 oif = 0;
1062
1063 redo_rt6_select:
1064         rt = rt6_select(fn, oif, strict);
1065         if (rt->rt6i_nsiblings)
1066                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1067         if (rt == net->ipv6.ip6_null_entry) {
1068                 fn = fib6_backtrack(fn, &fl6->saddr);
1069                 if (fn)
1070                         goto redo_rt6_select;
1071                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1072                         /* also consider unreachable route */
1073                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1074                         fn = saved_fn;
1075                         goto redo_rt6_select;
1076                 }
1077         }
1078
1079
1080         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1081                 dst_use(&rt->dst, jiffies);
1082                 read_unlock_bh(&table->tb6_lock);
1083
1084                 rt6_dst_from_metrics_check(rt);
1085
1086                 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1087                 return rt;
1088         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1089                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1090                 /* Create a RTF_CACHE clone which will not be
1091                  * owned by the fib6 tree.  It is for the special case where
1092                  * the daddr in the skb during the neighbor look-up is different
1093                  * from the fl6->daddr used to look-up route here.
1094                  */
1095
1096                 struct rt6_info *uncached_rt;
1097
1098                 dst_use(&rt->dst, jiffies);
1099                 read_unlock_bh(&table->tb6_lock);
1100
1101                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1102                 dst_release(&rt->dst);
1103
1104                 if (uncached_rt)
1105                         rt6_uncached_list_add(uncached_rt);
1106                 else
1107                         uncached_rt = net->ipv6.ip6_null_entry;
1108
1109                 dst_hold(&uncached_rt->dst);
1110
1111                 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1112                 return uncached_rt;
1113
1114         } else {
1115                 /* Get a percpu copy */
1116
1117                 struct rt6_info *pcpu_rt;
1118
1119                 rt->dst.lastuse = jiffies;
1120                 rt->dst.__use++;
1121                 pcpu_rt = rt6_get_pcpu_route(rt);
1122
1123                 if (pcpu_rt) {
1124                         read_unlock_bh(&table->tb6_lock);
1125                 } else {
1126                         /* We have to do the read_unlock first
1127                          * because rt6_make_pcpu_route() may trigger
1128                          * ip6_dst_gc() which will take the write_lock.
1129                          */
1130                         dst_hold(&rt->dst);
1131                         read_unlock_bh(&table->tb6_lock);
1132                         pcpu_rt = rt6_make_pcpu_route(rt);
1133                         dst_release(&rt->dst);
1134                 }
1135
1136                 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1137                 return pcpu_rt;
1138
1139         }
1140 }
1141
1142 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1143                                             struct flowi6 *fl6, int flags)
1144 {
1145         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1146 }
1147
1148 static struct dst_entry *ip6_route_input_lookup(struct net *net,
1149                                                 struct net_device *dev,
1150                                                 struct flowi6 *fl6, int flags)
1151 {
1152         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1153                 flags |= RT6_LOOKUP_F_IFACE;
1154
1155         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1156 }
1157
1158 void ip6_route_input(struct sk_buff *skb)
1159 {
1160         const struct ipv6hdr *iph = ipv6_hdr(skb);
1161         struct net *net = dev_net(skb->dev);
1162         int flags = RT6_LOOKUP_F_HAS_SADDR;
1163         struct ip_tunnel_info *tun_info;
1164         struct flowi6 fl6 = {
1165                 .flowi6_iif = l3mdev_fib_oif(skb->dev),
1166                 .daddr = iph->daddr,
1167                 .saddr = iph->saddr,
1168                 .flowlabel = ip6_flowinfo(iph),
1169                 .flowi6_mark = skb->mark,
1170                 .flowi6_proto = iph->nexthdr,
1171         };
1172
1173         tun_info = skb_tunnel_info(skb);
1174         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1175                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1176         skb_dst_drop(skb);
1177         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1178 }
1179
1180 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1181                                              struct flowi6 *fl6, int flags)
1182 {
1183         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1184 }
1185
1186 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1187                                          struct flowi6 *fl6, int flags)
1188 {
1189         struct dst_entry *dst;
1190         bool any_src;
1191
1192         dst = l3mdev_rt6_dst_by_oif(net, fl6);
1193         if (dst)
1194                 return dst;
1195
1196         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1197
1198         any_src = ipv6_addr_any(&fl6->saddr);
1199         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1200             (fl6->flowi6_oif && any_src))
1201                 flags |= RT6_LOOKUP_F_IFACE;
1202
1203         if (!any_src)
1204                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1205         else if (sk)
1206                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1207
1208         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1209 }
1210 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1211
1212 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1213 {
1214         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1215         struct dst_entry *new = NULL;
1216
1217         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1218         if (rt) {
1219                 rt6_info_init(rt);
1220
1221                 new = &rt->dst;
1222                 new->__use = 1;
1223                 new->input = dst_discard;
1224                 new->output = dst_discard_out;
1225
1226                 dst_copy_metrics(new, &ort->dst);
1227                 rt->rt6i_idev = ort->rt6i_idev;
1228                 if (rt->rt6i_idev)
1229                         in6_dev_hold(rt->rt6i_idev);
1230
1231                 rt->rt6i_gateway = ort->rt6i_gateway;
1232                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1233                 rt->rt6i_metric = 0;
1234
1235                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1236 #ifdef CONFIG_IPV6_SUBTREES
1237                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1238 #endif
1239
1240                 dst_free(new);
1241         }
1242
1243         dst_release(dst_orig);
1244         return new ? new : ERR_PTR(-ENOMEM);
1245 }
1246
1247 /*
1248  *      Destination cache support functions
1249  */
1250
1251 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1252 {
1253         if (rt->dst.from &&
1254             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1255                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1256 }
1257
1258 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1259 {
1260         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1261                 return NULL;
1262
1263         if (rt6_check_expired(rt))
1264                 return NULL;
1265
1266         return &rt->dst;
1267 }
1268
1269 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1270 {
1271         if (!__rt6_check_expired(rt) &&
1272             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1273             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1274                 return &rt->dst;
1275         else
1276                 return NULL;
1277 }
1278
1279 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1280 {
1281         struct rt6_info *rt;
1282
1283         rt = (struct rt6_info *) dst;
1284
1285         /* All IPV6 dsts are created with ->obsolete set to the value
1286          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1287          * into this function always.
1288          */
1289
1290         rt6_dst_from_metrics_check(rt);
1291
1292         if (rt->rt6i_flags & RTF_PCPU ||
1293             (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1294                 return rt6_dst_from_check(rt, cookie);
1295         else
1296                 return rt6_check(rt, cookie);
1297 }
1298
1299 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1300 {
1301         struct rt6_info *rt = (struct rt6_info *) dst;
1302
1303         if (rt) {
1304                 if (rt->rt6i_flags & RTF_CACHE) {
1305                         if (rt6_check_expired(rt)) {
1306                                 ip6_del_rt(rt);
1307                                 dst = NULL;
1308                         }
1309                 } else {
1310                         dst_release(dst);
1311                         dst = NULL;
1312                 }
1313         }
1314         return dst;
1315 }
1316
1317 static void ip6_link_failure(struct sk_buff *skb)
1318 {
1319         struct rt6_info *rt;
1320
1321         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1322
1323         rt = (struct rt6_info *) skb_dst(skb);
1324         if (rt) {
1325                 if (rt->rt6i_flags & RTF_CACHE) {
1326                         dst_hold(&rt->dst);
1327                         ip6_del_rt(rt);
1328                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1329                         rt->rt6i_node->fn_sernum = -1;
1330                 }
1331         }
1332 }
1333
1334 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1335 {
1336         struct net *net = dev_net(rt->dst.dev);
1337
1338         rt->rt6i_flags |= RTF_MODIFIED;
1339         rt->rt6i_pmtu = mtu;
1340         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1341 }
1342
1343 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1344 {
1345         return !(rt->rt6i_flags & RTF_CACHE) &&
1346                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1347 }
1348
1349 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1350                                  const struct ipv6hdr *iph, u32 mtu)
1351 {
1352         struct rt6_info *rt6 = (struct rt6_info *)dst;
1353
1354         if (rt6->rt6i_flags & RTF_LOCAL)
1355                 return;
1356
1357         dst_confirm(dst);
1358         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1359         if (mtu >= dst_mtu(dst))
1360                 return;
1361
1362         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1363                 rt6_do_update_pmtu(rt6, mtu);
1364         } else {
1365                 const struct in6_addr *daddr, *saddr;
1366                 struct rt6_info *nrt6;
1367
1368                 if (iph) {
1369                         daddr = &iph->daddr;
1370                         saddr = &iph->saddr;
1371                 } else if (sk) {
1372                         daddr = &sk->sk_v6_daddr;
1373                         saddr = &inet6_sk(sk)->saddr;
1374                 } else {
1375                         return;
1376                 }
1377                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1378                 if (nrt6) {
1379                         rt6_do_update_pmtu(nrt6, mtu);
1380
1381                         /* ip6_ins_rt(nrt6) will bump the
1382                          * rt6->rt6i_node->fn_sernum
1383                          * which will fail the next rt6_check() and
1384                          * invalidate the sk->sk_dst_cache.
1385                          */
1386                         ip6_ins_rt(nrt6);
1387                 }
1388         }
1389 }
1390
1391 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1392                                struct sk_buff *skb, u32 mtu)
1393 {
1394         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1395 }
1396
1397 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1398                      int oif, u32 mark)
1399 {
1400         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1401         struct dst_entry *dst;
1402         struct flowi6 fl6;
1403
1404         memset(&fl6, 0, sizeof(fl6));
1405         fl6.flowi6_oif = oif;
1406         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1407         fl6.daddr = iph->daddr;
1408         fl6.saddr = iph->saddr;
1409         fl6.flowlabel = ip6_flowinfo(iph);
1410
1411         dst = ip6_route_output(net, NULL, &fl6);
1412         if (!dst->error)
1413                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1414         dst_release(dst);
1415 }
1416 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1417
1418 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1419 {
1420         ip6_update_pmtu(skb, sock_net(sk), mtu,
1421                         sk->sk_bound_dev_if, sk->sk_mark);
1422 }
1423 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1424
1425 /* Handle redirects */
1426 struct ip6rd_flowi {
1427         struct flowi6 fl6;
1428         struct in6_addr gateway;
1429 };
1430
1431 static struct rt6_info *__ip6_route_redirect(struct net *net,
1432                                              struct fib6_table *table,
1433                                              struct flowi6 *fl6,
1434                                              int flags)
1435 {
1436         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1437         struct rt6_info *rt;
1438         struct fib6_node *fn;
1439
1440         /* Get the "current" route for this destination and
1441          * check if the redirect has come from approriate router.
1442          *
1443          * RFC 4861 specifies that redirects should only be
1444          * accepted if they come from the nexthop to the target.
1445          * Due to the way the routes are chosen, this notion
1446          * is a bit fuzzy and one might need to check all possible
1447          * routes.
1448          */
1449
1450         read_lock_bh(&table->tb6_lock);
1451         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1452 restart:
1453         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1454                 if (rt6_check_expired(rt))
1455                         continue;
1456                 if (rt->dst.error)
1457                         break;
1458                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1459                         continue;
1460                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1461                         continue;
1462                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1463                         continue;
1464                 break;
1465         }
1466
1467         if (!rt)
1468                 rt = net->ipv6.ip6_null_entry;
1469         else if (rt->dst.error) {
1470                 rt = net->ipv6.ip6_null_entry;
1471                 goto out;
1472         }
1473
1474         if (rt == net->ipv6.ip6_null_entry) {
1475                 fn = fib6_backtrack(fn, &fl6->saddr);
1476                 if (fn)
1477                         goto restart;
1478         }
1479
1480 out:
1481         dst_hold(&rt->dst);
1482
1483         read_unlock_bh(&table->tb6_lock);
1484
1485         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1486         return rt;
1487 };
1488
1489 static struct dst_entry *ip6_route_redirect(struct net *net,
1490                                         const struct flowi6 *fl6,
1491                                         const struct in6_addr *gateway)
1492 {
1493         int flags = RT6_LOOKUP_F_HAS_SADDR;
1494         struct ip6rd_flowi rdfl;
1495
1496         rdfl.fl6 = *fl6;
1497         rdfl.gateway = *gateway;
1498
1499         return fib6_rule_lookup(net, &rdfl.fl6,
1500                                 flags, __ip6_route_redirect);
1501 }
1502
1503 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark)
1504 {
1505         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1506         struct dst_entry *dst;
1507         struct flowi6 fl6;
1508
1509         memset(&fl6, 0, sizeof(fl6));
1510         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1511         fl6.flowi6_oif = oif;
1512         fl6.flowi6_mark = mark;
1513         fl6.daddr = iph->daddr;
1514         fl6.saddr = iph->saddr;
1515         fl6.flowlabel = ip6_flowinfo(iph);
1516
1517         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1518         rt6_do_redirect(dst, NULL, skb);
1519         dst_release(dst);
1520 }
1521 EXPORT_SYMBOL_GPL(ip6_redirect);
1522
1523 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1524                             u32 mark)
1525 {
1526         const struct ipv6hdr *iph = ipv6_hdr(skb);
1527         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1528         struct dst_entry *dst;
1529         struct flowi6 fl6;
1530
1531         memset(&fl6, 0, sizeof(fl6));
1532         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1533         fl6.flowi6_oif = oif;
1534         fl6.flowi6_mark = mark;
1535         fl6.daddr = msg->dest;
1536         fl6.saddr = iph->daddr;
1537
1538         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1539         rt6_do_redirect(dst, NULL, skb);
1540         dst_release(dst);
1541 }
1542
1543 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1544 {
1545         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark);
1546 }
1547 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1548
1549 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1550 {
1551         struct net_device *dev = dst->dev;
1552         unsigned int mtu = dst_mtu(dst);
1553         struct net *net = dev_net(dev);
1554
1555         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1556
1557         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1558                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1559
1560         /*
1561          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1562          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1563          * IPV6_MAXPLEN is also valid and means: "any MSS,
1564          * rely only on pmtu discovery"
1565          */
1566         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1567                 mtu = IPV6_MAXPLEN;
1568         return mtu;
1569 }
1570
1571 static unsigned int ip6_mtu(const struct dst_entry *dst)
1572 {
1573         const struct rt6_info *rt = (const struct rt6_info *)dst;
1574         unsigned int mtu = rt->rt6i_pmtu;
1575         struct inet6_dev *idev;
1576
1577         if (mtu)
1578                 goto out;
1579
1580         mtu = dst_metric_raw(dst, RTAX_MTU);
1581         if (mtu)
1582                 goto out;
1583
1584         mtu = IPV6_MIN_MTU;
1585
1586         rcu_read_lock();
1587         idev = __in6_dev_get(dst->dev);
1588         if (idev)
1589                 mtu = idev->cnf.mtu6;
1590         rcu_read_unlock();
1591
1592 out:
1593         return min_t(unsigned int, mtu, IP6_MAX_MTU);
1594 }
1595
1596 static struct dst_entry *icmp6_dst_gc_list;
1597 static DEFINE_SPINLOCK(icmp6_dst_lock);
1598
1599 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1600                                   struct flowi6 *fl6)
1601 {
1602         struct dst_entry *dst;
1603         struct rt6_info *rt;
1604         struct inet6_dev *idev = in6_dev_get(dev);
1605         struct net *net = dev_net(dev);
1606
1607         if (unlikely(!idev))
1608                 return ERR_PTR(-ENODEV);
1609
1610         rt = ip6_dst_alloc(net, dev, 0);
1611         if (unlikely(!rt)) {
1612                 in6_dev_put(idev);
1613                 dst = ERR_PTR(-ENOMEM);
1614                 goto out;
1615         }
1616
1617         rt->dst.flags |= DST_HOST;
1618         rt->dst.output  = ip6_output;
1619         atomic_set(&rt->dst.__refcnt, 1);
1620         rt->rt6i_gateway  = fl6->daddr;
1621         rt->rt6i_dst.addr = fl6->daddr;
1622         rt->rt6i_dst.plen = 128;
1623         rt->rt6i_idev     = idev;
1624         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1625
1626         spin_lock_bh(&icmp6_dst_lock);
1627         rt->dst.next = icmp6_dst_gc_list;
1628         icmp6_dst_gc_list = &rt->dst;
1629         spin_unlock_bh(&icmp6_dst_lock);
1630
1631         fib6_force_start_gc(net);
1632
1633         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1634
1635 out:
1636         return dst;
1637 }
1638
1639 int icmp6_dst_gc(void)
1640 {
1641         struct dst_entry *dst, **pprev;
1642         int more = 0;
1643
1644         spin_lock_bh(&icmp6_dst_lock);
1645         pprev = &icmp6_dst_gc_list;
1646
1647         while ((dst = *pprev) != NULL) {
1648                 if (!atomic_read(&dst->__refcnt)) {
1649                         *pprev = dst->next;
1650                         dst_free(dst);
1651                 } else {
1652                         pprev = &dst->next;
1653                         ++more;
1654                 }
1655         }
1656
1657         spin_unlock_bh(&icmp6_dst_lock);
1658
1659         return more;
1660 }
1661
1662 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1663                             void *arg)
1664 {
1665         struct dst_entry *dst, **pprev;
1666
1667         spin_lock_bh(&icmp6_dst_lock);
1668         pprev = &icmp6_dst_gc_list;
1669         while ((dst = *pprev) != NULL) {
1670                 struct rt6_info *rt = (struct rt6_info *) dst;
1671                 if (func(rt, arg)) {
1672                         *pprev = dst->next;
1673                         dst_free(dst);
1674                 } else {
1675                         pprev = &dst->next;
1676                 }
1677         }
1678         spin_unlock_bh(&icmp6_dst_lock);
1679 }
1680
1681 static int ip6_dst_gc(struct dst_ops *ops)
1682 {
1683         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1684         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1685         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1686         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1687         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1688         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1689         int entries;
1690
1691         entries = dst_entries_get_fast(ops);
1692         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1693             entries <= rt_max_size)
1694                 goto out;
1695
1696         net->ipv6.ip6_rt_gc_expire++;
1697         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1698         entries = dst_entries_get_slow(ops);
1699         if (entries < ops->gc_thresh)
1700                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1701 out:
1702         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1703         return entries > rt_max_size;
1704 }
1705
1706 static int ip6_convert_metrics(struct mx6_config *mxc,
1707                                const struct fib6_config *cfg)
1708 {
1709         bool ecn_ca = false;
1710         struct nlattr *nla;
1711         int remaining;
1712         u32 *mp;
1713
1714         if (!cfg->fc_mx)
1715                 return 0;
1716
1717         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1718         if (unlikely(!mp))
1719                 return -ENOMEM;
1720
1721         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1722                 int type = nla_type(nla);
1723                 u32 val;
1724
1725                 if (!type)
1726                         continue;
1727                 if (unlikely(type > RTAX_MAX))
1728                         goto err;
1729
1730                 if (type == RTAX_CC_ALGO) {
1731                         char tmp[TCP_CA_NAME_MAX];
1732
1733                         nla_strlcpy(tmp, nla, sizeof(tmp));
1734                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1735                         if (val == TCP_CA_UNSPEC)
1736                                 goto err;
1737                 } else {
1738                         val = nla_get_u32(nla);
1739                 }
1740                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1741                         goto err;
1742
1743                 mp[type - 1] = val;
1744                 __set_bit(type - 1, mxc->mx_valid);
1745         }
1746
1747         if (ecn_ca) {
1748                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1749                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1750         }
1751
1752         mxc->mx = mp;
1753         return 0;
1754  err:
1755         kfree(mp);
1756         return -EINVAL;
1757 }
1758
1759 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1760 {
1761         struct net *net = cfg->fc_nlinfo.nl_net;
1762         struct rt6_info *rt = NULL;
1763         struct net_device *dev = NULL;
1764         struct inet6_dev *idev = NULL;
1765         struct fib6_table *table;
1766         int addr_type;
1767         int err = -EINVAL;
1768
1769         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1770                 goto out;
1771 #ifndef CONFIG_IPV6_SUBTREES
1772         if (cfg->fc_src_len)
1773                 goto out;
1774 #endif
1775         if (cfg->fc_ifindex) {
1776                 err = -ENODEV;
1777                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1778                 if (!dev)
1779                         goto out;
1780                 idev = in6_dev_get(dev);
1781                 if (!idev)
1782                         goto out;
1783         }
1784
1785         if (cfg->fc_metric == 0)
1786                 cfg->fc_metric = IP6_RT_PRIO_USER;
1787
1788         err = -ENOBUFS;
1789         if (cfg->fc_nlinfo.nlh &&
1790             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1791                 table = fib6_get_table(net, cfg->fc_table);
1792                 if (!table) {
1793                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1794                         table = fib6_new_table(net, cfg->fc_table);
1795                 }
1796         } else {
1797                 table = fib6_new_table(net, cfg->fc_table);
1798         }
1799
1800         if (!table)
1801                 goto out;
1802
1803         rt = ip6_dst_alloc(net, NULL,
1804                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1805
1806         if (!rt) {
1807                 err = -ENOMEM;
1808                 goto out;
1809         }
1810
1811         if (cfg->fc_flags & RTF_EXPIRES)
1812                 rt6_set_expires(rt, jiffies +
1813                                 clock_t_to_jiffies(cfg->fc_expires));
1814         else
1815                 rt6_clean_expires(rt);
1816
1817         if (cfg->fc_protocol == RTPROT_UNSPEC)
1818                 cfg->fc_protocol = RTPROT_BOOT;
1819         rt->rt6i_protocol = cfg->fc_protocol;
1820
1821         addr_type = ipv6_addr_type(&cfg->fc_dst);
1822
1823         if (addr_type & IPV6_ADDR_MULTICAST)
1824                 rt->dst.input = ip6_mc_input;
1825         else if (cfg->fc_flags & RTF_LOCAL)
1826                 rt->dst.input = ip6_input;
1827         else
1828                 rt->dst.input = ip6_forward;
1829
1830         rt->dst.output = ip6_output;
1831
1832         if (cfg->fc_encap) {
1833                 struct lwtunnel_state *lwtstate;
1834
1835                 err = lwtunnel_build_state(dev, cfg->fc_encap_type,
1836                                            cfg->fc_encap, AF_INET6, cfg,
1837                                            &lwtstate);
1838                 if (err)
1839                         goto out;
1840                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1841                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1842                         rt->dst.lwtstate->orig_output = rt->dst.output;
1843                         rt->dst.output = lwtunnel_output;
1844                 }
1845                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1846                         rt->dst.lwtstate->orig_input = rt->dst.input;
1847                         rt->dst.input = lwtunnel_input;
1848                 }
1849         }
1850
1851         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1852         rt->rt6i_dst.plen = cfg->fc_dst_len;
1853         if (rt->rt6i_dst.plen == 128)
1854                 rt->dst.flags |= DST_HOST;
1855
1856 #ifdef CONFIG_IPV6_SUBTREES
1857         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1858         rt->rt6i_src.plen = cfg->fc_src_len;
1859 #endif
1860
1861         rt->rt6i_metric = cfg->fc_metric;
1862
1863         /* We cannot add true routes via loopback here,
1864            they would result in kernel looping; promote them to reject routes
1865          */
1866         if ((cfg->fc_flags & RTF_REJECT) ||
1867             (dev && (dev->flags & IFF_LOOPBACK) &&
1868              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1869              !(cfg->fc_flags & RTF_LOCAL))) {
1870                 /* hold loopback dev/idev if we haven't done so. */
1871                 if (dev != net->loopback_dev) {
1872                         if (dev) {
1873                                 dev_put(dev);
1874                                 in6_dev_put(idev);
1875                         }
1876                         dev = net->loopback_dev;
1877                         dev_hold(dev);
1878                         idev = in6_dev_get(dev);
1879                         if (!idev) {
1880                                 err = -ENODEV;
1881                                 goto out;
1882                         }
1883                 }
1884                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1885                 switch (cfg->fc_type) {
1886                 case RTN_BLACKHOLE:
1887                         rt->dst.error = -EINVAL;
1888                         rt->dst.output = dst_discard_out;
1889                         rt->dst.input = dst_discard;
1890                         break;
1891                 case RTN_PROHIBIT:
1892                         rt->dst.error = -EACCES;
1893                         rt->dst.output = ip6_pkt_prohibit_out;
1894                         rt->dst.input = ip6_pkt_prohibit;
1895                         break;
1896                 case RTN_THROW:
1897                 case RTN_UNREACHABLE:
1898                 default:
1899                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1900                                         : (cfg->fc_type == RTN_UNREACHABLE)
1901                                         ? -EHOSTUNREACH : -ENETUNREACH;
1902                         rt->dst.output = ip6_pkt_discard_out;
1903                         rt->dst.input = ip6_pkt_discard;
1904                         break;
1905                 }
1906                 goto install_route;
1907         }
1908
1909         if (cfg->fc_flags & RTF_GATEWAY) {
1910                 const struct in6_addr *gw_addr;
1911                 int gwa_type;
1912
1913                 gw_addr = &cfg->fc_gateway;
1914                 gwa_type = ipv6_addr_type(gw_addr);
1915
1916                 /* if gw_addr is local we will fail to detect this in case
1917                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1918                  * will return already-added prefix route via interface that
1919                  * prefix route was assigned to, which might be non-loopback.
1920                  */
1921                 err = -EINVAL;
1922                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1923                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1924                                             dev : NULL, 0, 0))
1925                         goto out;
1926
1927                 rt->rt6i_gateway = *gw_addr;
1928
1929                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1930                         struct rt6_info *grt;
1931
1932                         /* IPv6 strictly inhibits using not link-local
1933                            addresses as nexthop address.
1934                            Otherwise, router will not able to send redirects.
1935                            It is very good, but in some (rare!) circumstances
1936                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1937                            some exceptions. --ANK
1938                          */
1939                         if (!(gwa_type & IPV6_ADDR_UNICAST))
1940                                 goto out;
1941
1942                         grt = rt6_lookup(net, gw_addr, NULL, cfg->fc_ifindex, 1);
1943
1944                         err = -EHOSTUNREACH;
1945                         if (!grt)
1946                                 goto out;
1947                         if (dev) {
1948                                 if (dev != grt->dst.dev) {
1949                                         ip6_rt_put(grt);
1950                                         goto out;
1951                                 }
1952                         } else {
1953                                 dev = grt->dst.dev;
1954                                 idev = grt->rt6i_idev;
1955                                 dev_hold(dev);
1956                                 in6_dev_hold(grt->rt6i_idev);
1957                         }
1958                         if (!(grt->rt6i_flags & RTF_GATEWAY))
1959                                 err = 0;
1960                         ip6_rt_put(grt);
1961
1962                         if (err)
1963                                 goto out;
1964                 }
1965                 err = -EINVAL;
1966                 if (!dev || (dev->flags & IFF_LOOPBACK))
1967                         goto out;
1968         }
1969
1970         err = -ENODEV;
1971         if (!dev)
1972                 goto out;
1973
1974         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
1975                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
1976                         err = -EINVAL;
1977                         goto out;
1978                 }
1979                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
1980                 rt->rt6i_prefsrc.plen = 128;
1981         } else
1982                 rt->rt6i_prefsrc.plen = 0;
1983
1984         rt->rt6i_flags = cfg->fc_flags;
1985
1986 install_route:
1987         rt->dst.dev = dev;
1988         rt->rt6i_idev = idev;
1989         rt->rt6i_table = table;
1990
1991         cfg->fc_nlinfo.nl_net = dev_net(dev);
1992
1993         return rt;
1994 out:
1995         if (dev)
1996                 dev_put(dev);
1997         if (idev)
1998                 in6_dev_put(idev);
1999         if (rt)
2000                 dst_free(&rt->dst);
2001
2002         return ERR_PTR(err);
2003 }
2004
2005 int ip6_route_add(struct fib6_config *cfg)
2006 {
2007         struct mx6_config mxc = { .mx = NULL, };
2008         struct rt6_info *rt;
2009         int err;
2010
2011         rt = ip6_route_info_create(cfg);
2012         if (IS_ERR(rt)) {
2013                 err = PTR_ERR(rt);
2014                 rt = NULL;
2015                 goto out;
2016         }
2017
2018         err = ip6_convert_metrics(&mxc, cfg);
2019         if (err)
2020                 goto out;
2021
2022         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2023
2024         kfree(mxc.mx);
2025
2026         return err;
2027 out:
2028         if (rt)
2029                 dst_free(&rt->dst);
2030
2031         return err;
2032 }
2033
2034 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2035 {
2036         int err;
2037         struct fib6_table *table;
2038         struct net *net = dev_net(rt->dst.dev);
2039
2040         if (rt == net->ipv6.ip6_null_entry ||
2041             rt->dst.flags & DST_NOCACHE) {
2042                 err = -ENOENT;
2043                 goto out;
2044         }
2045
2046         table = rt->rt6i_table;
2047         write_lock_bh(&table->tb6_lock);
2048         err = fib6_del(rt, info);
2049         write_unlock_bh(&table->tb6_lock);
2050
2051 out:
2052         ip6_rt_put(rt);
2053         return err;
2054 }
2055
2056 int ip6_del_rt(struct rt6_info *rt)
2057 {
2058         struct nl_info info = {
2059                 .nl_net = dev_net(rt->dst.dev),
2060         };
2061         return __ip6_del_rt(rt, &info);
2062 }
2063
2064 static int ip6_route_del(struct fib6_config *cfg)
2065 {
2066         struct fib6_table *table;
2067         struct fib6_node *fn;
2068         struct rt6_info *rt;
2069         int err = -ESRCH;
2070
2071         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2072         if (!table)
2073                 return err;
2074
2075         read_lock_bh(&table->tb6_lock);
2076
2077         fn = fib6_locate(&table->tb6_root,
2078                          &cfg->fc_dst, cfg->fc_dst_len,
2079                          &cfg->fc_src, cfg->fc_src_len);
2080
2081         if (fn) {
2082                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2083                         if ((rt->rt6i_flags & RTF_CACHE) &&
2084                             !(cfg->fc_flags & RTF_CACHE))
2085                                 continue;
2086                         if (cfg->fc_ifindex &&
2087                             (!rt->dst.dev ||
2088                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2089                                 continue;
2090                         if (cfg->fc_flags & RTF_GATEWAY &&
2091                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2092                                 continue;
2093                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2094                                 continue;
2095                         dst_hold(&rt->dst);
2096                         read_unlock_bh(&table->tb6_lock);
2097
2098                         return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2099                 }
2100         }
2101         read_unlock_bh(&table->tb6_lock);
2102
2103         return err;
2104 }
2105
2106 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2107 {
2108         struct netevent_redirect netevent;
2109         struct rt6_info *rt, *nrt = NULL;
2110         struct ndisc_options ndopts;
2111         struct inet6_dev *in6_dev;
2112         struct neighbour *neigh;
2113         struct rd_msg *msg;
2114         int optlen, on_link;
2115         u8 *lladdr;
2116
2117         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2118         optlen -= sizeof(*msg);
2119
2120         if (optlen < 0) {
2121                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2122                 return;
2123         }
2124
2125         msg = (struct rd_msg *)icmp6_hdr(skb);
2126
2127         if (ipv6_addr_is_multicast(&msg->dest)) {
2128                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2129                 return;
2130         }
2131
2132         on_link = 0;
2133         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2134                 on_link = 1;
2135         } else if (ipv6_addr_type(&msg->target) !=
2136                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2137                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2138                 return;
2139         }
2140
2141         in6_dev = __in6_dev_get(skb->dev);
2142         if (!in6_dev)
2143                 return;
2144         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2145                 return;
2146
2147         /* RFC2461 8.1:
2148          *      The IP source address of the Redirect MUST be the same as the current
2149          *      first-hop router for the specified ICMP Destination Address.
2150          */
2151
2152         if (!ndisc_parse_options(msg->opt, optlen, &ndopts)) {
2153                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2154                 return;
2155         }
2156
2157         lladdr = NULL;
2158         if (ndopts.nd_opts_tgt_lladdr) {
2159                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2160                                              skb->dev);
2161                 if (!lladdr) {
2162                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2163                         return;
2164                 }
2165         }
2166
2167         rt = (struct rt6_info *) dst;
2168         if (rt->rt6i_flags & RTF_REJECT) {
2169                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2170                 return;
2171         }
2172
2173         /* Redirect received -> path was valid.
2174          * Look, redirects are sent only in response to data packets,
2175          * so that this nexthop apparently is reachable. --ANK
2176          */
2177         dst_confirm(&rt->dst);
2178
2179         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2180         if (!neigh)
2181                 return;
2182
2183         /*
2184          *      We have finally decided to accept it.
2185          */
2186
2187         neigh_update(neigh, lladdr, NUD_STALE,
2188                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2189                      NEIGH_UPDATE_F_OVERRIDE|
2190                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2191                                      NEIGH_UPDATE_F_ISROUTER))
2192                      );
2193
2194         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2195         if (!nrt)
2196                 goto out;
2197
2198         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2199         if (on_link)
2200                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2201
2202         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2203
2204         if (ip6_ins_rt(nrt))
2205                 goto out;
2206
2207         netevent.old = &rt->dst;
2208         netevent.new = &nrt->dst;
2209         netevent.daddr = &msg->dest;
2210         netevent.neigh = neigh;
2211         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2212
2213         if (rt->rt6i_flags & RTF_CACHE) {
2214                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2215                 ip6_del_rt(rt);
2216         }
2217
2218 out:
2219         neigh_release(neigh);
2220 }
2221
2222 /*
2223  *      Misc support functions
2224  */
2225
2226 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2227 {
2228         BUG_ON(from->dst.from);
2229
2230         rt->rt6i_flags &= ~RTF_EXPIRES;
2231         dst_hold(&from->dst);
2232         rt->dst.from = &from->dst;
2233         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2234 }
2235
2236 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2237 {
2238         rt->dst.input = ort->dst.input;
2239         rt->dst.output = ort->dst.output;
2240         rt->rt6i_dst = ort->rt6i_dst;
2241         rt->dst.error = ort->dst.error;
2242         rt->rt6i_idev = ort->rt6i_idev;
2243         if (rt->rt6i_idev)
2244                 in6_dev_hold(rt->rt6i_idev);
2245         rt->dst.lastuse = jiffies;
2246         rt->rt6i_gateway = ort->rt6i_gateway;
2247         rt->rt6i_flags = ort->rt6i_flags;
2248         rt6_set_from(rt, ort);
2249         rt->rt6i_metric = ort->rt6i_metric;
2250 #ifdef CONFIG_IPV6_SUBTREES
2251         rt->rt6i_src = ort->rt6i_src;
2252 #endif
2253         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2254         rt->rt6i_table = ort->rt6i_table;
2255         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2256 }
2257
2258 #ifdef CONFIG_IPV6_ROUTE_INFO
2259 static struct rt6_info *rt6_get_route_info(struct net *net,
2260                                            const struct in6_addr *prefix, int prefixlen,
2261                                            const struct in6_addr *gwaddr, int ifindex)
2262 {
2263         struct fib6_node *fn;
2264         struct rt6_info *rt = NULL;
2265         struct fib6_table *table;
2266
2267         table = fib6_get_table(net, RT6_TABLE_INFO);
2268         if (!table)
2269                 return NULL;
2270
2271         read_lock_bh(&table->tb6_lock);
2272         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2273         if (!fn)
2274                 goto out;
2275
2276         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2277                 if (rt->dst.dev->ifindex != ifindex)
2278                         continue;
2279                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2280                         continue;
2281                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2282                         continue;
2283                 dst_hold(&rt->dst);
2284                 break;
2285         }
2286 out:
2287         read_unlock_bh(&table->tb6_lock);
2288         return rt;
2289 }
2290
2291 static struct rt6_info *rt6_add_route_info(struct net *net,
2292                                            const struct in6_addr *prefix, int prefixlen,
2293                                            const struct in6_addr *gwaddr, int ifindex,
2294                                            unsigned int pref)
2295 {
2296         struct fib6_config cfg = {
2297                 .fc_metric      = IP6_RT_PRIO_USER,
2298                 .fc_ifindex     = ifindex,
2299                 .fc_dst_len     = prefixlen,
2300                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2301                                   RTF_UP | RTF_PREF(pref),
2302                 .fc_nlinfo.portid = 0,
2303                 .fc_nlinfo.nlh = NULL,
2304                 .fc_nlinfo.nl_net = net,
2305         };
2306
2307         cfg.fc_table = l3mdev_fib_table_by_index(net, ifindex) ? : RT6_TABLE_INFO;
2308         cfg.fc_dst = *prefix;
2309         cfg.fc_gateway = *gwaddr;
2310
2311         /* We should treat it as a default route if prefix length is 0. */
2312         if (!prefixlen)
2313                 cfg.fc_flags |= RTF_DEFAULT;
2314
2315         ip6_route_add(&cfg);
2316
2317         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, ifindex);
2318 }
2319 #endif
2320
2321 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2322 {
2323         struct rt6_info *rt;
2324         struct fib6_table *table;
2325
2326         table = fib6_get_table(dev_net(dev), RT6_TABLE_DFLT);
2327         if (!table)
2328                 return NULL;
2329
2330         read_lock_bh(&table->tb6_lock);
2331         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2332                 if (dev == rt->dst.dev &&
2333                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2334                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2335                         break;
2336         }
2337         if (rt)
2338                 dst_hold(&rt->dst);
2339         read_unlock_bh(&table->tb6_lock);
2340         return rt;
2341 }
2342
2343 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2344                                      struct net_device *dev,
2345                                      unsigned int pref)
2346 {
2347         struct fib6_config cfg = {
2348                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2349                 .fc_metric      = IP6_RT_PRIO_USER,
2350                 .fc_ifindex     = dev->ifindex,
2351                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2352                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2353                 .fc_nlinfo.portid = 0,
2354                 .fc_nlinfo.nlh = NULL,
2355                 .fc_nlinfo.nl_net = dev_net(dev),
2356         };
2357
2358         cfg.fc_gateway = *gwaddr;
2359
2360         ip6_route_add(&cfg);
2361
2362         return rt6_get_dflt_router(gwaddr, dev);
2363 }
2364
2365 void rt6_purge_dflt_routers(struct net *net)
2366 {
2367         struct rt6_info *rt;
2368         struct fib6_table *table;
2369
2370         /* NOTE: Keep consistent with rt6_get_dflt_router */
2371         table = fib6_get_table(net, RT6_TABLE_DFLT);
2372         if (!table)
2373                 return;
2374
2375 restart:
2376         read_lock_bh(&table->tb6_lock);
2377         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2378                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2379                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2380                         dst_hold(&rt->dst);
2381                         read_unlock_bh(&table->tb6_lock);
2382                         ip6_del_rt(rt);
2383                         goto restart;
2384                 }
2385         }
2386         read_unlock_bh(&table->tb6_lock);
2387 }
2388
2389 static void rtmsg_to_fib6_config(struct net *net,
2390                                  struct in6_rtmsg *rtmsg,
2391                                  struct fib6_config *cfg)
2392 {
2393         memset(cfg, 0, sizeof(*cfg));
2394
2395         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2396                          : RT6_TABLE_MAIN;
2397         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2398         cfg->fc_metric = rtmsg->rtmsg_metric;
2399         cfg->fc_expires = rtmsg->rtmsg_info;
2400         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2401         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2402         cfg->fc_flags = rtmsg->rtmsg_flags;
2403
2404         cfg->fc_nlinfo.nl_net = net;
2405
2406         cfg->fc_dst = rtmsg->rtmsg_dst;
2407         cfg->fc_src = rtmsg->rtmsg_src;
2408         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2409 }
2410
2411 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2412 {
2413         struct fib6_config cfg;
2414         struct in6_rtmsg rtmsg;
2415         int err;
2416
2417         switch (cmd) {
2418         case SIOCADDRT:         /* Add a route */
2419         case SIOCDELRT:         /* Delete a route */
2420                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2421                         return -EPERM;
2422                 err = copy_from_user(&rtmsg, arg,
2423                                      sizeof(struct in6_rtmsg));
2424                 if (err)
2425                         return -EFAULT;
2426
2427                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2428
2429                 rtnl_lock();
2430                 switch (cmd) {
2431                 case SIOCADDRT:
2432                         err = ip6_route_add(&cfg);
2433                         break;
2434                 case SIOCDELRT:
2435                         err = ip6_route_del(&cfg);
2436                         break;
2437                 default:
2438                         err = -EINVAL;
2439                 }
2440                 rtnl_unlock();
2441
2442                 return err;
2443         }
2444
2445         return -EINVAL;
2446 }
2447
2448 /*
2449  *      Drop the packet on the floor
2450  */
2451
2452 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2453 {
2454         int type;
2455         struct dst_entry *dst = skb_dst(skb);
2456         switch (ipstats_mib_noroutes) {
2457         case IPSTATS_MIB_INNOROUTES:
2458                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2459                 if (type == IPV6_ADDR_ANY) {
2460                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2461                                       IPSTATS_MIB_INADDRERRORS);
2462                         break;
2463                 }
2464                 /* FALLTHROUGH */
2465         case IPSTATS_MIB_OUTNOROUTES:
2466                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2467                               ipstats_mib_noroutes);
2468                 break;
2469         }
2470         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2471         kfree_skb(skb);
2472         return 0;
2473 }
2474
2475 static int ip6_pkt_discard(struct sk_buff *skb)
2476 {
2477         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2478 }
2479
2480 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2481 {
2482         skb->dev = skb_dst(skb)->dev;
2483         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2484 }
2485
2486 static int ip6_pkt_prohibit(struct sk_buff *skb)
2487 {
2488         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2489 }
2490
2491 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2492 {
2493         skb->dev = skb_dst(skb)->dev;
2494         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2495 }
2496
2497 /*
2498  *      Allocate a dst for local (unicast / anycast) address.
2499  */
2500
2501 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2502                                     const struct in6_addr *addr,
2503                                     bool anycast)
2504 {
2505         u32 tb_id;
2506         struct net *net = dev_net(idev->dev);
2507         struct rt6_info *rt = ip6_dst_alloc(net, net->loopback_dev,
2508                                             DST_NOCOUNT);
2509         if (!rt)
2510                 return ERR_PTR(-ENOMEM);
2511
2512         in6_dev_hold(idev);
2513
2514         rt->dst.flags |= DST_HOST;
2515         rt->dst.input = ip6_input;
2516         rt->dst.output = ip6_output;
2517         rt->rt6i_idev = idev;
2518
2519         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2520         if (anycast)
2521                 rt->rt6i_flags |= RTF_ANYCAST;
2522         else
2523                 rt->rt6i_flags |= RTF_LOCAL;
2524
2525         rt->rt6i_gateway  = *addr;
2526         rt->rt6i_dst.addr = *addr;
2527         rt->rt6i_dst.plen = 128;
2528         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2529         rt->rt6i_table = fib6_get_table(net, tb_id);
2530         rt->dst.flags |= DST_NOCACHE;
2531
2532         atomic_set(&rt->dst.__refcnt, 1);
2533
2534         return rt;
2535 }
2536
2537 int ip6_route_get_saddr(struct net *net,
2538                         struct rt6_info *rt,
2539                         const struct in6_addr *daddr,
2540                         unsigned int prefs,
2541                         struct in6_addr *saddr)
2542 {
2543         struct inet6_dev *idev =
2544                 rt ? ip6_dst_idev((struct dst_entry *)rt) : NULL;
2545         int err = 0;
2546         if (rt && rt->rt6i_prefsrc.plen)
2547                 *saddr = rt->rt6i_prefsrc.addr;
2548         else
2549                 err = ipv6_dev_get_saddr(net, idev ? idev->dev : NULL,
2550                                          daddr, prefs, saddr);
2551         return err;
2552 }
2553
2554 /* remove deleted ip from prefsrc entries */
2555 struct arg_dev_net_ip {
2556         struct net_device *dev;
2557         struct net *net;
2558         struct in6_addr *addr;
2559 };
2560
2561 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2562 {
2563         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2564         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2565         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2566
2567         if (((void *)rt->dst.dev == dev || !dev) &&
2568             rt != net->ipv6.ip6_null_entry &&
2569             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2570                 /* remove prefsrc entry */
2571                 rt->rt6i_prefsrc.plen = 0;
2572         }
2573         return 0;
2574 }
2575
2576 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2577 {
2578         struct net *net = dev_net(ifp->idev->dev);
2579         struct arg_dev_net_ip adni = {
2580                 .dev = ifp->idev->dev,
2581                 .net = net,
2582                 .addr = &ifp->addr,
2583         };
2584         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2585 }
2586
2587 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2588 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2589
2590 /* Remove routers and update dst entries when gateway turn into host. */
2591 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2592 {
2593         struct in6_addr *gateway = (struct in6_addr *)arg;
2594
2595         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2596              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2597              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2598                 return -1;
2599         }
2600         return 0;
2601 }
2602
2603 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2604 {
2605         fib6_clean_all(net, fib6_clean_tohost, gateway);
2606 }
2607
2608 struct arg_dev_net {
2609         struct net_device *dev;
2610         struct net *net;
2611 };
2612
2613 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2614 {
2615         const struct arg_dev_net *adn = arg;
2616         const struct net_device *dev = adn->dev;
2617
2618         if ((rt->dst.dev == dev || !dev) &&
2619             rt != adn->net->ipv6.ip6_null_entry)
2620                 return -1;
2621
2622         return 0;
2623 }
2624
2625 void rt6_ifdown(struct net *net, struct net_device *dev)
2626 {
2627         struct arg_dev_net adn = {
2628                 .dev = dev,
2629                 .net = net,
2630         };
2631
2632         fib6_clean_all(net, fib6_ifdown, &adn);
2633         icmp6_clean_all(fib6_ifdown, &adn);
2634         if (dev)
2635                 rt6_uncached_list_flush_dev(net, dev);
2636 }
2637
2638 struct rt6_mtu_change_arg {
2639         struct net_device *dev;
2640         unsigned int mtu;
2641 };
2642
2643 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2644 {
2645         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2646         struct inet6_dev *idev;
2647
2648         /* In IPv6 pmtu discovery is not optional,
2649            so that RTAX_MTU lock cannot disable it.
2650            We still use this lock to block changes
2651            caused by addrconf/ndisc.
2652         */
2653
2654         idev = __in6_dev_get(arg->dev);
2655         if (!idev)
2656                 return 0;
2657
2658         /* For administrative MTU increase, there is no way to discover
2659            IPv6 PMTU increase, so PMTU increase should be updated here.
2660            Since RFC 1981 doesn't include administrative MTU increase
2661            update PMTU increase is a MUST. (i.e. jumbo frame)
2662          */
2663         /*
2664            If new MTU is less than route PMTU, this new MTU will be the
2665            lowest MTU in the path, update the route PMTU to reflect PMTU
2666            decreases; if new MTU is greater than route PMTU, and the
2667            old MTU is the lowest MTU in the path, update the route PMTU
2668            to reflect the increase. In this case if the other nodes' MTU
2669            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2670            PMTU discouvery.
2671          */
2672         if (rt->dst.dev == arg->dev &&
2673             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2674                 if (rt->rt6i_flags & RTF_CACHE) {
2675                         /* For RTF_CACHE with rt6i_pmtu == 0
2676                          * (i.e. a redirected route),
2677                          * the metrics of its rt->dst.from has already
2678                          * been updated.
2679                          */
2680                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2681                                 rt->rt6i_pmtu = arg->mtu;
2682                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2683                            (dst_mtu(&rt->dst) < arg->mtu &&
2684                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2685                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2686                 }
2687         }
2688         return 0;
2689 }
2690
2691 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2692 {
2693         struct rt6_mtu_change_arg arg = {
2694                 .dev = dev,
2695                 .mtu = mtu,
2696         };
2697
2698         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2699 }
2700
2701 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2702         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2703         [RTA_OIF]               = { .type = NLA_U32 },
2704         [RTA_IIF]               = { .type = NLA_U32 },
2705         [RTA_PRIORITY]          = { .type = NLA_U32 },
2706         [RTA_METRICS]           = { .type = NLA_NESTED },
2707         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2708         [RTA_PREF]              = { .type = NLA_U8 },
2709         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2710         [RTA_ENCAP]             = { .type = NLA_NESTED },
2711         [RTA_EXPIRES]           = { .type = NLA_U32 },
2712 };
2713
2714 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2715                               struct fib6_config *cfg)
2716 {
2717         struct rtmsg *rtm;
2718         struct nlattr *tb[RTA_MAX+1];
2719         unsigned int pref;
2720         int err;
2721
2722         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2723         if (err < 0)
2724                 goto errout;
2725
2726         err = -EINVAL;
2727         rtm = nlmsg_data(nlh);
2728         memset(cfg, 0, sizeof(*cfg));
2729
2730         cfg->fc_table = rtm->rtm_table;
2731         cfg->fc_dst_len = rtm->rtm_dst_len;
2732         cfg->fc_src_len = rtm->rtm_src_len;
2733         cfg->fc_flags = RTF_UP;
2734         cfg->fc_protocol = rtm->rtm_protocol;
2735         cfg->fc_type = rtm->rtm_type;
2736
2737         if (rtm->rtm_type == RTN_UNREACHABLE ||
2738             rtm->rtm_type == RTN_BLACKHOLE ||
2739             rtm->rtm_type == RTN_PROHIBIT ||
2740             rtm->rtm_type == RTN_THROW)
2741                 cfg->fc_flags |= RTF_REJECT;
2742
2743         if (rtm->rtm_type == RTN_LOCAL)
2744                 cfg->fc_flags |= RTF_LOCAL;
2745
2746         if (rtm->rtm_flags & RTM_F_CLONED)
2747                 cfg->fc_flags |= RTF_CACHE;
2748
2749         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2750         cfg->fc_nlinfo.nlh = nlh;
2751         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2752
2753         if (tb[RTA_GATEWAY]) {
2754                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2755                 cfg->fc_flags |= RTF_GATEWAY;
2756         }
2757
2758         if (tb[RTA_DST]) {
2759                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2760
2761                 if (nla_len(tb[RTA_DST]) < plen)
2762                         goto errout;
2763
2764                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2765         }
2766
2767         if (tb[RTA_SRC]) {
2768                 int plen = (rtm->rtm_src_len + 7) >> 3;
2769
2770                 if (nla_len(tb[RTA_SRC]) < plen)
2771                         goto errout;
2772
2773                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2774         }
2775
2776         if (tb[RTA_PREFSRC])
2777                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2778
2779         if (tb[RTA_OIF])
2780                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2781
2782         if (tb[RTA_PRIORITY])
2783                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2784
2785         if (tb[RTA_METRICS]) {
2786                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2787                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2788         }
2789
2790         if (tb[RTA_TABLE])
2791                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2792
2793         if (tb[RTA_MULTIPATH]) {
2794                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2795                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2796         }
2797
2798         if (tb[RTA_PREF]) {
2799                 pref = nla_get_u8(tb[RTA_PREF]);
2800                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2801                     pref != ICMPV6_ROUTER_PREF_HIGH)
2802                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2803                 cfg->fc_flags |= RTF_PREF(pref);
2804         }
2805
2806         if (tb[RTA_ENCAP])
2807                 cfg->fc_encap = tb[RTA_ENCAP];
2808
2809         if (tb[RTA_ENCAP_TYPE])
2810                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2811
2812         if (tb[RTA_EXPIRES]) {
2813                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
2814
2815                 if (addrconf_finite_timeout(timeout)) {
2816                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
2817                         cfg->fc_flags |= RTF_EXPIRES;
2818                 }
2819         }
2820
2821         err = 0;
2822 errout:
2823         return err;
2824 }
2825
2826 struct rt6_nh {
2827         struct rt6_info *rt6_info;
2828         struct fib6_config r_cfg;
2829         struct mx6_config mxc;
2830         struct list_head next;
2831 };
2832
2833 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
2834 {
2835         struct rt6_nh *nh;
2836
2837         list_for_each_entry(nh, rt6_nh_list, next) {
2838                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6 nexthop %pI6 ifi %d\n",
2839                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
2840                         nh->r_cfg.fc_ifindex);
2841         }
2842 }
2843
2844 static int ip6_route_info_append(struct list_head *rt6_nh_list,
2845                                  struct rt6_info *rt, struct fib6_config *r_cfg)
2846 {
2847         struct rt6_nh *nh;
2848         struct rt6_info *rtnh;
2849         int err = -EEXIST;
2850
2851         list_for_each_entry(nh, rt6_nh_list, next) {
2852                 /* check if rt6_info already exists */
2853                 rtnh = nh->rt6_info;
2854
2855                 if (rtnh->dst.dev == rt->dst.dev &&
2856                     rtnh->rt6i_idev == rt->rt6i_idev &&
2857                     ipv6_addr_equal(&rtnh->rt6i_gateway,
2858                                     &rt->rt6i_gateway))
2859                         return err;
2860         }
2861
2862         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
2863         if (!nh)
2864                 return -ENOMEM;
2865         nh->rt6_info = rt;
2866         err = ip6_convert_metrics(&nh->mxc, r_cfg);
2867         if (err) {
2868                 kfree(nh);
2869                 return err;
2870         }
2871         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
2872         list_add_tail(&nh->next, rt6_nh_list);
2873
2874         return 0;
2875 }
2876
2877 static int ip6_route_multipath_add(struct fib6_config *cfg)
2878 {
2879         struct fib6_config r_cfg;
2880         struct rtnexthop *rtnh;
2881         struct rt6_info *rt;
2882         struct rt6_nh *err_nh;
2883         struct rt6_nh *nh, *nh_safe;
2884         int remaining;
2885         int attrlen;
2886         int err = 1;
2887         int nhn = 0;
2888         int replace = (cfg->fc_nlinfo.nlh &&
2889                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
2890         LIST_HEAD(rt6_nh_list);
2891
2892         remaining = cfg->fc_mp_len;
2893         rtnh = (struct rtnexthop *)cfg->fc_mp;
2894
2895         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
2896          * rt6_info structs per nexthop
2897          */
2898         while (rtnh_ok(rtnh, remaining)) {
2899                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2900                 if (rtnh->rtnh_ifindex)
2901                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2902
2903                 attrlen = rtnh_attrlen(rtnh);
2904                 if (attrlen > 0) {
2905                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
2906
2907                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
2908                         if (nla) {
2909                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
2910                                 r_cfg.fc_flags |= RTF_GATEWAY;
2911                         }
2912                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
2913                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
2914                         if (nla)
2915                                 r_cfg.fc_encap_type = nla_get_u16(nla);
2916                 }
2917
2918                 rt = ip6_route_info_create(&r_cfg);
2919                 if (IS_ERR(rt)) {
2920                         err = PTR_ERR(rt);
2921                         rt = NULL;
2922                         goto cleanup;
2923                 }
2924
2925                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
2926                 if (err) {
2927                         dst_free(&rt->dst);
2928                         goto cleanup;
2929                 }
2930
2931                 rtnh = rtnh_next(rtnh, &remaining);
2932         }
2933
2934         err_nh = NULL;
2935         list_for_each_entry(nh, &rt6_nh_list, next) {
2936                 err = __ip6_ins_rt(nh->rt6_info, &cfg->fc_nlinfo, &nh->mxc);
2937                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
2938                 nh->rt6_info = NULL;
2939                 if (err) {
2940                         if (replace && nhn)
2941                                 ip6_print_replace_route_err(&rt6_nh_list);
2942                         err_nh = nh;
2943                         goto add_errout;
2944                 }
2945
2946                 /* Because each route is added like a single route we remove
2947                  * these flags after the first nexthop: if there is a collision,
2948                  * we have already failed to add the first nexthop:
2949                  * fib6_add_rt2node() has rejected it; when replacing, old
2950                  * nexthops have been replaced by first new, the rest should
2951                  * be added to it.
2952                  */
2953                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
2954                                                      NLM_F_REPLACE);
2955                 nhn++;
2956         }
2957
2958         goto cleanup;
2959
2960 add_errout:
2961         /* Delete routes that were already added */
2962         list_for_each_entry(nh, &rt6_nh_list, next) {
2963                 if (err_nh == nh)
2964                         break;
2965                 ip6_route_del(&nh->r_cfg);
2966         }
2967
2968 cleanup:
2969         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
2970                 if (nh->rt6_info)
2971                         dst_free(&nh->rt6_info->dst);
2972                 kfree(nh->mxc.mx);
2973                 list_del(&nh->next);
2974                 kfree(nh);
2975         }
2976
2977         return err;
2978 }
2979
2980 static int ip6_route_multipath_del(struct fib6_config *cfg)
2981 {
2982         struct fib6_config r_cfg;
2983         struct rtnexthop *rtnh;
2984         int remaining;
2985         int attrlen;
2986         int err = 1, last_err = 0;
2987
2988         remaining = cfg->fc_mp_len;
2989         rtnh = (struct rtnexthop *)cfg->fc_mp;
2990
2991         /* Parse a Multipath Entry */
2992         while (rtnh_ok(rtnh, remaining)) {
2993                 memcpy(&r_cfg, cfg, sizeof(*cfg));
2994                 if (rtnh->rtnh_ifindex)
2995                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
2996
2997                 attrlen = rtnh_attrlen(rtnh);
2998                 if (attrlen > 0) {
2999                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3000
3001                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3002                         if (nla) {
3003                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3004                                 r_cfg.fc_flags |= RTF_GATEWAY;
3005                         }
3006                 }
3007                 err = ip6_route_del(&r_cfg);
3008                 if (err)
3009                         last_err = err;
3010
3011                 rtnh = rtnh_next(rtnh, &remaining);
3012         }
3013
3014         return last_err;
3015 }
3016
3017 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3018 {
3019         struct fib6_config cfg;
3020         int err;
3021
3022         err = rtm_to_fib6_config(skb, nlh, &cfg);
3023         if (err < 0)
3024                 return err;
3025
3026         if (cfg.fc_mp)
3027                 return ip6_route_multipath_del(&cfg);
3028         else
3029                 return ip6_route_del(&cfg);
3030 }
3031
3032 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3033 {
3034         struct fib6_config cfg;
3035         int err;
3036
3037         err = rtm_to_fib6_config(skb, nlh, &cfg);
3038         if (err < 0)
3039                 return err;
3040
3041         if (cfg.fc_mp)
3042                 return ip6_route_multipath_add(&cfg);
3043         else
3044                 return ip6_route_add(&cfg);
3045 }
3046
3047 static inline size_t rt6_nlmsg_size(struct rt6_info *rt)
3048 {
3049         return NLMSG_ALIGN(sizeof(struct rtmsg))
3050                + nla_total_size(16) /* RTA_SRC */
3051                + nla_total_size(16) /* RTA_DST */
3052                + nla_total_size(16) /* RTA_GATEWAY */
3053                + nla_total_size(16) /* RTA_PREFSRC */
3054                + nla_total_size(4) /* RTA_TABLE */
3055                + nla_total_size(4) /* RTA_IIF */
3056                + nla_total_size(4) /* RTA_OIF */
3057                + nla_total_size(4) /* RTA_PRIORITY */
3058                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3059                + nla_total_size(sizeof(struct rta_cacheinfo))
3060                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3061                + nla_total_size(1) /* RTA_PREF */
3062                + lwtunnel_get_encap_size(rt->dst.lwtstate);
3063 }
3064
3065 static int rt6_fill_node(struct net *net,
3066                          struct sk_buff *skb, struct rt6_info *rt,
3067                          struct in6_addr *dst, struct in6_addr *src,
3068                          int iif, int type, u32 portid, u32 seq,
3069                          int prefix, int nowait, unsigned int flags)
3070 {
3071         u32 metrics[RTAX_MAX];
3072         struct rtmsg *rtm;
3073         struct nlmsghdr *nlh;
3074         long expires;
3075         u32 table;
3076
3077         if (prefix) {   /* user wants prefix routes only */
3078                 if (!(rt->rt6i_flags & RTF_PREFIX_RT)) {
3079                         /* success since this is not a prefix route */
3080                         return 1;
3081                 }
3082         }
3083
3084         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3085         if (!nlh)
3086                 return -EMSGSIZE;
3087
3088         rtm = nlmsg_data(nlh);
3089         rtm->rtm_family = AF_INET6;
3090         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3091         rtm->rtm_src_len = rt->rt6i_src.plen;
3092         rtm->rtm_tos = 0;
3093         if (rt->rt6i_table)
3094                 table = rt->rt6i_table->tb6_id;
3095         else
3096                 table = RT6_TABLE_UNSPEC;
3097         rtm->rtm_table = table;
3098         if (nla_put_u32(skb, RTA_TABLE, table))
3099                 goto nla_put_failure;
3100         if (rt->rt6i_flags & RTF_REJECT) {
3101                 switch (rt->dst.error) {
3102                 case -EINVAL:
3103                         rtm->rtm_type = RTN_BLACKHOLE;
3104                         break;
3105                 case -EACCES:
3106                         rtm->rtm_type = RTN_PROHIBIT;
3107                         break;
3108                 case -EAGAIN:
3109                         rtm->rtm_type = RTN_THROW;
3110                         break;
3111                 default:
3112                         rtm->rtm_type = RTN_UNREACHABLE;
3113                         break;
3114                 }
3115         }
3116         else if (rt->rt6i_flags & RTF_LOCAL)
3117                 rtm->rtm_type = RTN_LOCAL;
3118         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3119                 rtm->rtm_type = RTN_LOCAL;
3120         else
3121                 rtm->rtm_type = RTN_UNICAST;
3122         rtm->rtm_flags = 0;
3123         if (!netif_carrier_ok(rt->dst.dev)) {
3124                 rtm->rtm_flags |= RTNH_F_LINKDOWN;
3125                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3126                         rtm->rtm_flags |= RTNH_F_DEAD;
3127         }
3128         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3129         rtm->rtm_protocol = rt->rt6i_protocol;
3130         if (rt->rt6i_flags & RTF_DYNAMIC)
3131                 rtm->rtm_protocol = RTPROT_REDIRECT;
3132         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3133                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3134                         rtm->rtm_protocol = RTPROT_RA;
3135                 else
3136                         rtm->rtm_protocol = RTPROT_KERNEL;
3137         }
3138
3139         if (rt->rt6i_flags & RTF_CACHE)
3140                 rtm->rtm_flags |= RTM_F_CLONED;
3141
3142         if (dst) {
3143                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3144                         goto nla_put_failure;
3145                 rtm->rtm_dst_len = 128;
3146         } else if (rtm->rtm_dst_len)
3147                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3148                         goto nla_put_failure;
3149 #ifdef CONFIG_IPV6_SUBTREES
3150         if (src) {
3151                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3152                         goto nla_put_failure;
3153                 rtm->rtm_src_len = 128;
3154         } else if (rtm->rtm_src_len &&
3155                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3156                 goto nla_put_failure;
3157 #endif
3158         if (iif) {
3159 #ifdef CONFIG_IPV6_MROUTE
3160                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3161                         int err = ip6mr_get_route(net, skb, rtm, nowait);
3162                         if (err <= 0) {
3163                                 if (!nowait) {
3164                                         if (err == 0)
3165                                                 return 0;
3166                                         goto nla_put_failure;
3167                                 } else {
3168                                         if (err == -EMSGSIZE)
3169                                                 goto nla_put_failure;
3170                                 }
3171                         }
3172                 } else
3173 #endif
3174                         if (nla_put_u32(skb, RTA_IIF, iif))
3175                                 goto nla_put_failure;
3176         } else if (dst) {
3177                 struct in6_addr saddr_buf;
3178                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3179                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3180                         goto nla_put_failure;
3181         }
3182
3183         if (rt->rt6i_prefsrc.plen) {
3184                 struct in6_addr saddr_buf;
3185                 saddr_buf = rt->rt6i_prefsrc.addr;
3186                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3187                         goto nla_put_failure;
3188         }
3189
3190         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3191         if (rt->rt6i_pmtu)
3192                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3193         if (rtnetlink_put_metrics(skb, metrics) < 0)
3194                 goto nla_put_failure;
3195
3196         if (rt->rt6i_flags & RTF_GATEWAY) {
3197                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3198                         goto nla_put_failure;
3199         }
3200
3201         if (rt->dst.dev &&
3202             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3203                 goto nla_put_failure;
3204         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3205                 goto nla_put_failure;
3206
3207         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3208
3209         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3210                 goto nla_put_failure;
3211
3212         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3213                 goto nla_put_failure;
3214
3215         lwtunnel_fill_encap(skb, rt->dst.lwtstate);
3216
3217         nlmsg_end(skb, nlh);
3218         return 0;
3219
3220 nla_put_failure:
3221         nlmsg_cancel(skb, nlh);
3222         return -EMSGSIZE;
3223 }
3224
3225 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3226 {
3227         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3228         int prefix;
3229
3230         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3231                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3232                 prefix = (rtm->rtm_flags & RTM_F_PREFIX) != 0;
3233         } else
3234                 prefix = 0;
3235
3236         return rt6_fill_node(arg->net,
3237                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3238                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3239                      prefix, 0, NLM_F_MULTI);
3240 }
3241
3242 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3243 {
3244         struct net *net = sock_net(in_skb->sk);
3245         struct nlattr *tb[RTA_MAX+1];
3246         struct rt6_info *rt;
3247         struct sk_buff *skb;
3248         struct rtmsg *rtm;
3249         struct flowi6 fl6;
3250         int err, iif = 0, oif = 0;
3251
3252         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3253         if (err < 0)
3254                 goto errout;
3255
3256         err = -EINVAL;
3257         memset(&fl6, 0, sizeof(fl6));
3258
3259         if (tb[RTA_SRC]) {
3260                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3261                         goto errout;
3262
3263                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3264         }
3265
3266         if (tb[RTA_DST]) {
3267                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3268                         goto errout;
3269
3270                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3271         }
3272
3273         if (tb[RTA_IIF])
3274                 iif = nla_get_u32(tb[RTA_IIF]);
3275
3276         if (tb[RTA_OIF])
3277                 oif = nla_get_u32(tb[RTA_OIF]);
3278
3279         if (tb[RTA_MARK])
3280                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3281
3282         if (iif) {
3283                 struct net_device *dev;
3284                 int flags = 0;
3285
3286                 dev = __dev_get_by_index(net, iif);
3287                 if (!dev) {
3288                         err = -ENODEV;
3289                         goto errout;
3290                 }
3291
3292                 fl6.flowi6_iif = iif;
3293
3294                 if (!ipv6_addr_any(&fl6.saddr))
3295                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3296
3297                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3298                                                                flags);
3299         } else {
3300                 fl6.flowi6_oif = oif;
3301
3302                 if (netif_index_is_l3_master(net, oif)) {
3303                         fl6.flowi6_flags = FLOWI_FLAG_L3MDEV_SRC |
3304                                            FLOWI_FLAG_SKIP_NH_OIF;
3305                 }
3306
3307                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3308         }
3309
3310         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3311         if (!skb) {
3312                 ip6_rt_put(rt);
3313                 err = -ENOBUFS;
3314                 goto errout;
3315         }
3316
3317         /* Reserve room for dummy headers, this skb can pass
3318            through good chunk of routing engine.
3319          */
3320         skb_reset_mac_header(skb);
3321         skb_reserve(skb, MAX_HEADER + sizeof(struct ipv6hdr));
3322
3323         skb_dst_set(skb, &rt->dst);
3324
3325         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3326                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3327                             nlh->nlmsg_seq, 0, 0, 0);
3328         if (err < 0) {
3329                 kfree_skb(skb);
3330                 goto errout;
3331         }
3332
3333         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3334 errout:
3335         return err;
3336 }
3337
3338 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3339                      unsigned int nlm_flags)
3340 {
3341         struct sk_buff *skb;
3342         struct net *net = info->nl_net;
3343         u32 seq;
3344         int err;
3345
3346         err = -ENOBUFS;
3347         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3348
3349         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3350         if (!skb)
3351                 goto errout;
3352
3353         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3354                                 event, info->portid, seq, 0, 0, nlm_flags);
3355         if (err < 0) {
3356                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3357                 WARN_ON(err == -EMSGSIZE);
3358                 kfree_skb(skb);
3359                 goto errout;
3360         }
3361         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3362                     info->nlh, gfp_any());
3363         return;
3364 errout:
3365         if (err < 0)
3366                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3367 }
3368
3369 static int ip6_route_dev_notify(struct notifier_block *this,
3370                                 unsigned long event, void *ptr)
3371 {
3372         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3373         struct net *net = dev_net(dev);
3374
3375         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3376                 net->ipv6.ip6_null_entry->dst.dev = dev;
3377                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3378 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3379                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3380                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3381                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3382                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3383 #endif
3384         }
3385
3386         return NOTIFY_OK;
3387 }
3388
3389 /*
3390  *      /proc
3391  */
3392
3393 #ifdef CONFIG_PROC_FS
3394
3395 static const struct file_operations ipv6_route_proc_fops = {
3396         .owner          = THIS_MODULE,
3397         .open           = ipv6_route_open,
3398         .read           = seq_read,
3399         .llseek         = seq_lseek,
3400         .release        = seq_release_net,
3401 };
3402
3403 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3404 {
3405         struct net *net = (struct net *)seq->private;
3406         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3407                    net->ipv6.rt6_stats->fib_nodes,
3408                    net->ipv6.rt6_stats->fib_route_nodes,
3409                    net->ipv6.rt6_stats->fib_rt_alloc,
3410                    net->ipv6.rt6_stats->fib_rt_entries,
3411                    net->ipv6.rt6_stats->fib_rt_cache,
3412                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3413                    net->ipv6.rt6_stats->fib_discarded_routes);
3414
3415         return 0;
3416 }
3417
3418 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3419 {
3420         return single_open_net(inode, file, rt6_stats_seq_show);
3421 }
3422
3423 static const struct file_operations rt6_stats_seq_fops = {
3424         .owner   = THIS_MODULE,
3425         .open    = rt6_stats_seq_open,
3426         .read    = seq_read,
3427         .llseek  = seq_lseek,
3428         .release = single_release_net,
3429 };
3430 #endif  /* CONFIG_PROC_FS */
3431
3432 #ifdef CONFIG_SYSCTL
3433
3434 static
3435 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3436                               void __user *buffer, size_t *lenp, loff_t *ppos)
3437 {
3438         struct net *net;
3439         int delay;
3440         if (!write)
3441                 return -EINVAL;
3442
3443         net = (struct net *)ctl->extra1;
3444         delay = net->ipv6.sysctl.flush_delay;
3445         proc_dointvec(ctl, write, buffer, lenp, ppos);
3446         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3447         return 0;
3448 }
3449
3450 struct ctl_table ipv6_route_table_template[] = {
3451         {
3452                 .procname       =       "flush",
3453                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3454                 .maxlen         =       sizeof(int),
3455                 .mode           =       0200,
3456                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3457         },
3458         {
3459                 .procname       =       "gc_thresh",
3460                 .data           =       &ip6_dst_ops_template.gc_thresh,
3461                 .maxlen         =       sizeof(int),
3462                 .mode           =       0644,
3463                 .proc_handler   =       proc_dointvec,
3464         },
3465         {
3466                 .procname       =       "max_size",
3467                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3468                 .maxlen         =       sizeof(int),
3469                 .mode           =       0644,
3470                 .proc_handler   =       proc_dointvec,
3471         },
3472         {
3473                 .procname       =       "gc_min_interval",
3474                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3475                 .maxlen         =       sizeof(int),
3476                 .mode           =       0644,
3477                 .proc_handler   =       proc_dointvec_jiffies,
3478         },
3479         {
3480                 .procname       =       "gc_timeout",
3481                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3482                 .maxlen         =       sizeof(int),
3483                 .mode           =       0644,
3484                 .proc_handler   =       proc_dointvec_jiffies,
3485         },
3486         {
3487                 .procname       =       "gc_interval",
3488                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3489                 .maxlen         =       sizeof(int),
3490                 .mode           =       0644,
3491                 .proc_handler   =       proc_dointvec_jiffies,
3492         },
3493         {
3494                 .procname       =       "gc_elasticity",
3495                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3496                 .maxlen         =       sizeof(int),
3497                 .mode           =       0644,
3498                 .proc_handler   =       proc_dointvec,
3499         },
3500         {
3501                 .procname       =       "mtu_expires",
3502                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3503                 .maxlen         =       sizeof(int),
3504                 .mode           =       0644,
3505                 .proc_handler   =       proc_dointvec_jiffies,
3506         },
3507         {
3508                 .procname       =       "min_adv_mss",
3509                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3510                 .maxlen         =       sizeof(int),
3511                 .mode           =       0644,
3512                 .proc_handler   =       proc_dointvec,
3513         },
3514         {
3515                 .procname       =       "gc_min_interval_ms",
3516                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3517                 .maxlen         =       sizeof(int),
3518                 .mode           =       0644,
3519                 .proc_handler   =       proc_dointvec_ms_jiffies,
3520         },
3521         { }
3522 };
3523
3524 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3525 {
3526         struct ctl_table *table;
3527
3528         table = kmemdup(ipv6_route_table_template,
3529                         sizeof(ipv6_route_table_template),
3530                         GFP_KERNEL);
3531
3532         if (table) {
3533                 table[0].data = &net->ipv6.sysctl.flush_delay;
3534                 table[0].extra1 = net;
3535                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3536                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3537                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3538                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3539                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3540                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3541                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3542                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3543                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3544
3545                 /* Don't export sysctls to unprivileged users */
3546                 if (net->user_ns != &init_user_ns)
3547                         table[0].procname = NULL;
3548         }
3549
3550         return table;
3551 }
3552 #endif
3553
3554 static int __net_init ip6_route_net_init(struct net *net)
3555 {
3556         int ret = -ENOMEM;
3557
3558         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3559                sizeof(net->ipv6.ip6_dst_ops));
3560
3561         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3562                 goto out_ip6_dst_ops;
3563
3564         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3565                                            sizeof(*net->ipv6.ip6_null_entry),
3566                                            GFP_KERNEL);
3567         if (!net->ipv6.ip6_null_entry)
3568                 goto out_ip6_dst_entries;
3569         net->ipv6.ip6_null_entry->dst.path =
3570                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3571         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3572         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3573                          ip6_template_metrics, true);
3574
3575 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3576         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3577                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3578                                                GFP_KERNEL);
3579         if (!net->ipv6.ip6_prohibit_entry)
3580                 goto out_ip6_null_entry;
3581         net->ipv6.ip6_prohibit_entry->dst.path =
3582                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3583         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3584         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3585                          ip6_template_metrics, true);
3586
3587         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3588                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3589                                                GFP_KERNEL);
3590         if (!net->ipv6.ip6_blk_hole_entry)
3591                 goto out_ip6_prohibit_entry;
3592         net->ipv6.ip6_blk_hole_entry->dst.path =
3593                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3594         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3595         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3596                          ip6_template_metrics, true);
3597 #endif
3598
3599         net->ipv6.sysctl.flush_delay = 0;
3600         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3601         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3602         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3603         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3604         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3605         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3606         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3607
3608         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3609
3610         ret = 0;
3611 out:
3612         return ret;
3613
3614 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3615 out_ip6_prohibit_entry:
3616         kfree(net->ipv6.ip6_prohibit_entry);
3617 out_ip6_null_entry:
3618         kfree(net->ipv6.ip6_null_entry);
3619 #endif
3620 out_ip6_dst_entries:
3621         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3622 out_ip6_dst_ops:
3623         goto out;
3624 }
3625
3626 static void __net_exit ip6_route_net_exit(struct net *net)
3627 {
3628         kfree(net->ipv6.ip6_null_entry);
3629 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3630         kfree(net->ipv6.ip6_prohibit_entry);
3631         kfree(net->ipv6.ip6_blk_hole_entry);
3632 #endif
3633         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3634 }
3635
3636 static int __net_init ip6_route_net_init_late(struct net *net)
3637 {
3638 #ifdef CONFIG_PROC_FS
3639         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3640         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3641 #endif
3642         return 0;
3643 }
3644
3645 static void __net_exit ip6_route_net_exit_late(struct net *net)
3646 {
3647 #ifdef CONFIG_PROC_FS
3648         remove_proc_entry("ipv6_route", net->proc_net);
3649         remove_proc_entry("rt6_stats", net->proc_net);
3650 #endif
3651 }
3652
3653 static struct pernet_operations ip6_route_net_ops = {
3654         .init = ip6_route_net_init,
3655         .exit = ip6_route_net_exit,
3656 };
3657
3658 static int __net_init ipv6_inetpeer_init(struct net *net)
3659 {
3660         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3661
3662         if (!bp)
3663                 return -ENOMEM;
3664         inet_peer_base_init(bp);
3665         net->ipv6.peers = bp;
3666         return 0;
3667 }
3668
3669 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3670 {
3671         struct inet_peer_base *bp = net->ipv6.peers;
3672
3673         net->ipv6.peers = NULL;
3674         inetpeer_invalidate_tree(bp);
3675         kfree(bp);
3676 }
3677
3678 static struct pernet_operations ipv6_inetpeer_ops = {
3679         .init   =       ipv6_inetpeer_init,
3680         .exit   =       ipv6_inetpeer_exit,
3681 };
3682
3683 static struct pernet_operations ip6_route_net_late_ops = {
3684         .init = ip6_route_net_init_late,
3685         .exit = ip6_route_net_exit_late,
3686 };
3687
3688 static struct notifier_block ip6_route_dev_notifier = {
3689         .notifier_call = ip6_route_dev_notify,
3690         .priority = 0,
3691 };
3692
3693 int __init ip6_route_init(void)
3694 {
3695         int ret;
3696         int cpu;
3697
3698         ret = -ENOMEM;
3699         ip6_dst_ops_template.kmem_cachep =
3700                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
3701                                   SLAB_HWCACHE_ALIGN, NULL);
3702         if (!ip6_dst_ops_template.kmem_cachep)
3703                 goto out;
3704
3705         ret = dst_entries_init(&ip6_dst_blackhole_ops);
3706         if (ret)
3707                 goto out_kmem_cache;
3708
3709         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
3710         if (ret)
3711                 goto out_dst_entries;
3712
3713         ret = register_pernet_subsys(&ip6_route_net_ops);
3714         if (ret)
3715                 goto out_register_inetpeer;
3716
3717         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
3718
3719         /* Registering of the loopback is done before this portion of code,
3720          * the loopback reference in rt6_info will not be taken, do it
3721          * manually for init_net */
3722         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
3723         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3724   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3725         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
3726         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3727         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
3728         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
3729   #endif
3730         ret = fib6_init();
3731         if (ret)
3732                 goto out_register_subsys;
3733
3734         ret = xfrm6_init();
3735         if (ret)
3736                 goto out_fib6_init;
3737
3738         ret = fib6_rules_init();
3739         if (ret)
3740                 goto xfrm6_init;
3741
3742         ret = register_pernet_subsys(&ip6_route_net_late_ops);
3743         if (ret)
3744                 goto fib6_rules_init;
3745
3746         ret = -ENOBUFS;
3747         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
3748             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
3749             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
3750                 goto out_register_late_subsys;
3751
3752         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
3753         if (ret)
3754                 goto out_register_late_subsys;
3755
3756         for_each_possible_cpu(cpu) {
3757                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
3758
3759                 INIT_LIST_HEAD(&ul->head);
3760                 spin_lock_init(&ul->lock);
3761         }
3762
3763 out:
3764         return ret;
3765
3766 out_register_late_subsys:
3767         unregister_pernet_subsys(&ip6_route_net_late_ops);
3768 fib6_rules_init:
3769         fib6_rules_cleanup();
3770 xfrm6_init:
3771         xfrm6_fini();
3772 out_fib6_init:
3773         fib6_gc_cleanup();
3774 out_register_subsys:
3775         unregister_pernet_subsys(&ip6_route_net_ops);
3776 out_register_inetpeer:
3777         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3778 out_dst_entries:
3779         dst_entries_destroy(&ip6_dst_blackhole_ops);
3780 out_kmem_cache:
3781         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3782         goto out;
3783 }
3784
3785 void ip6_route_cleanup(void)
3786 {
3787         unregister_netdevice_notifier(&ip6_route_dev_notifier);
3788         unregister_pernet_subsys(&ip6_route_net_late_ops);
3789         fib6_rules_cleanup();
3790         xfrm6_fini();
3791         fib6_gc_cleanup();
3792         unregister_pernet_subsys(&ipv6_inetpeer_ops);
3793         unregister_pernet_subsys(&ip6_route_net_ops);
3794         dst_entries_destroy(&ip6_dst_blackhole_ops);
3795         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
3796 }