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 <linux/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 static size_t rt6_nlmsg_size(struct rt6_info *rt);
 102 static int rt6_fill_node(struct net *net,
 103                          struct sk_buff *skb, struct rt6_info *rt,
 104                          struct in6_addr *dst, struct in6_addr *src,
 105                          int iif, int type, u32 portid, u32 seq,
 106                          unsigned int flags);
 107
 108 #ifdef CONFIG_IPV6_ROUTE_INFO
 109 static struct rt6_info *rt6_add_route_info(struct net *net,
 110                                            const struct in6_addr *prefix, int prefixlen,
 111                                            const struct in6_addr *gwaddr,
 112                                            struct net_device *dev,
 113                                            unsigned int pref);
 114 static struct rt6_info *rt6_get_route_info(struct net *net,
 115                                            const struct in6_addr *prefix, int prefixlen,
 116                                            const struct in6_addr *gwaddr,
 117                                            struct net_device *dev);
 118 #endif
 119
 120 struct uncached_list {
 121         spinlock_t              lock;
 122         struct list_head        head;
 123 };
 124
 125 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 126
 127 static void rt6_uncached_list_add(struct rt6_info *rt)
 128 {
 129         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 130
 131         rt->dst.flags |= DST_NOCACHE;
 132         rt->rt6i_uncached_list = ul;
 133
 134         spin_lock_bh(&ul->lock);
 135         list_add_tail(&rt->rt6i_uncached, &ul->head);
 136         spin_unlock_bh(&ul->lock);
 137 }
 138
 139 static void rt6_uncached_list_del(struct rt6_info *rt)
 140 {
 141         if (!list_empty(&rt->rt6i_uncached)) {
 142                 struct uncached_list *ul = rt->rt6i_uncached_list;
 143
 144                 spin_lock_bh(&ul->lock);
 145                 list_del(&rt->rt6i_uncached);
 146                 spin_unlock_bh(&ul->lock);
 147         }
 148 }
 149
 150 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
 151 {
 152         struct net_device *loopback_dev = net->loopback_dev;
 153         int cpu;
 154
 155         if (dev == loopback_dev)
 156                 return;
 157
 158         for_each_possible_cpu(cpu) {
 159                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 160                 struct rt6_info *rt;
 161
 162                 spin_lock_bh(&ul->lock);
 163                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 164                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 165                         struct net_device *rt_dev = rt->dst.dev;
 166
 167                         if (rt_idev->dev == dev) {
 168                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 169                                 in6_dev_put(rt_idev);
 170                         }
 171
 172                         if (rt_dev == dev) {
 173                                 rt->dst.dev = loopback_dev;
 174                                 dev_hold(rt->dst.dev);
 175                                 dev_put(rt_dev);
 176                         }
 177                 }
 178                 spin_unlock_bh(&ul->lock);
 179         }
 180 }
 181
 182 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 183 {
 184         return dst_metrics_write_ptr(rt->dst.from);
 185 }
 186
 187 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 188 {
 189         struct rt6_info *rt = (struct rt6_info *)dst;
 190
 191         if (rt->rt6i_flags & RTF_PCPU)
 192                 return rt6_pcpu_cow_metrics(rt);
 193         else if (rt->rt6i_flags & RTF_CACHE)
 194                 return NULL;
 195         else
 196                 return dst_cow_metrics_generic(dst, old);
 197 }
 198
 199 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 200                                              struct sk_buff *skb,
 201                                              const void *daddr)
 202 {
 203         struct in6_addr *p = &rt->rt6i_gateway;
 204
 205         if (!ipv6_addr_any(p))
 206                 return (const void *) p;
 207         else if (skb)
 208                 return &ipv6_hdr(skb)->daddr;
 209         return daddr;
 210 }
 211
 212 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 213                                           struct sk_buff *skb,
 214                                           const void *daddr)
 215 {
 216         struct rt6_info *rt = (struct rt6_info *) dst;
 217         struct neighbour *n;
 218
 219         daddr = choose_neigh_daddr(rt, skb, daddr);
 220         n = __ipv6_neigh_lookup(dst->dev, daddr);
 221         if (n)
 222                 return n;
 223         return neigh_create(&nd_tbl, daddr, dst->dev);
 224 }
 225
 226 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
 227 {
 228         struct net_device *dev = dst->dev;
 229         struct rt6_info *rt = (struct rt6_info *)dst;
 230
 231         daddr = choose_neigh_daddr(rt, NULL, daddr);
 232         if (!daddr)
 233                 return;
 234         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 235                 return;
 236         if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
 237                 return;
 238         __ipv6_confirm_neigh(dev, daddr);
 239 }
 240
 241 static struct dst_ops ip6_dst_ops_template = {
 242         .family                 =       AF_INET6,
 243         .gc                     =       ip6_dst_gc,
 244         .gc_thresh              =       1024,
 245         .check                  =       ip6_dst_check,
 246         .default_advmss         =       ip6_default_advmss,
 247         .mtu                    =       ip6_mtu,
 248         .cow_metrics            =       ipv6_cow_metrics,
 249         .destroy                =       ip6_dst_destroy,
 250         .ifdown                 =       ip6_dst_ifdown,
 251         .negative_advice        =       ip6_negative_advice,
 252         .link_failure           =       ip6_link_failure,
 253         .update_pmtu            =       ip6_rt_update_pmtu,
 254         .redirect               =       rt6_do_redirect,
 255         .local_out              =       __ip6_local_out,
 256         .neigh_lookup           =       ip6_neigh_lookup,
 257         .confirm_neigh          =       ip6_confirm_neigh,
 258 };
 259
 260 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 261 {
 262         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 263
 264         return mtu ? : dst->dev->mtu;
 265 }
 266
 267 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 268                                          struct sk_buff *skb, u32 mtu)
 269 {
 270 }
 271
 272 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 273                                       struct sk_buff *skb)
 274 {
 275 }
 276
 277 static struct dst_ops ip6_dst_blackhole_ops = {
 278         .family                 =       AF_INET6,
 279         .destroy                =       ip6_dst_destroy,
 280         .check                  =       ip6_dst_check,
 281         .mtu                    =       ip6_blackhole_mtu,
 282         .default_advmss         =       ip6_default_advmss,
 283         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 284         .redirect               =       ip6_rt_blackhole_redirect,
 285         .cow_metrics            =       dst_cow_metrics_generic,
 286         .neigh_lookup           =       ip6_neigh_lookup,
 287 };
 288
 289 static const u32 ip6_template_metrics[RTAX_MAX] = {
 290         [RTAX_HOPLIMIT - 1] = 0,
 291 };
 292
 293 static const struct rt6_info ip6_null_entry_template = {
 294         .dst = {
 295                 .__refcnt       = ATOMIC_INIT(1),
 296                 .__use          = 1,
 297                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 298                 .error          = -ENETUNREACH,
 299                 .input          = ip6_pkt_discard,
 300                 .output         = ip6_pkt_discard_out,
 301         },
 302         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 303         .rt6i_protocol  = RTPROT_KERNEL,
 304         .rt6i_metric    = ~(u32) 0,
 305         .rt6i_ref       = ATOMIC_INIT(1),
 306 };
 307
 308 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 309
 310 static const struct rt6_info ip6_prohibit_entry_template = {
 311         .dst = {
 312                 .__refcnt       = ATOMIC_INIT(1),
 313                 .__use          = 1,
 314                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 315                 .error          = -EACCES,
 316                 .input          = ip6_pkt_prohibit,
 317                 .output         = ip6_pkt_prohibit_out,
 318         },
 319         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 320         .rt6i_protocol  = RTPROT_KERNEL,
 321         .rt6i_metric    = ~(u32) 0,
 322         .rt6i_ref       = ATOMIC_INIT(1),
 323 };
 324
 325 static const struct rt6_info ip6_blk_hole_entry_template = {
 326         .dst = {
 327                 .__refcnt       = ATOMIC_INIT(1),
 328                 .__use          = 1,
 329                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 330                 .error          = -EINVAL,
 331                 .input          = dst_discard,
 332                 .output         = dst_discard_out,
 333         },
 334         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 335         .rt6i_protocol  = RTPROT_KERNEL,
 336         .rt6i_metric    = ~(u32) 0,
 337         .rt6i_ref       = ATOMIC_INIT(1),
 338 };
 339
 340 #endif
 341
 342 static void rt6_info_init(struct rt6_info *rt)
 343 {
 344         struct dst_entry *dst = &rt->dst;
 345
 346         memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 347         INIT_LIST_HEAD(&rt->rt6i_siblings);
 348         INIT_LIST_HEAD(&rt->rt6i_uncached);
 349 }
 350
 351 /* allocate dst with ip6_dst_ops */
 352 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 353                                         struct net_device *dev,
 354                                         int flags)
 355 {
 356         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 357                                         0, DST_OBSOLETE_FORCE_CHK, flags);
 358
 359         if (rt)
 360                 rt6_info_init(rt);
 361
 362         return rt;
 363 }
 364
 365 struct rt6_info *ip6_dst_alloc(struct net *net,
 366                                struct net_device *dev,
 367                                int flags)
 368 {
 369         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 370
 371         if (rt) {
 372                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 373                 if (rt->rt6i_pcpu) {
 374                         int cpu;
 375
 376                         for_each_possible_cpu(cpu) {
 377                                 struct rt6_info **p;
 378
 379                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 380                                 /* no one shares rt */
 381                                 *p =  NULL;
 382                         }
 383                 } else {
 384                         dst_destroy((struct dst_entry *)rt);
 385                         return NULL;
 386                 }
 387         }
 388
 389         return rt;
 390 }
 391 EXPORT_SYMBOL(ip6_dst_alloc);
 392
 393 static void ip6_dst_destroy(struct dst_entry *dst)
 394 {
 395         struct rt6_info *rt = (struct rt6_info *)dst;
 396         struct dst_entry *from = dst->from;
 397         struct inet6_dev *idev;
 398
 399         dst_destroy_metrics_generic(dst);
 400         free_percpu(rt->rt6i_pcpu);
 401         rt6_uncached_list_del(rt);
 402
 403         idev = rt->rt6i_idev;
 404         if (idev) {
 405                 rt->rt6i_idev = NULL;
 406                 in6_dev_put(idev);
 407         }
 408
 409         dst->from = NULL;
 410         dst_release(from);
 411 }
 412
 413 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 414                            int how)
 415 {
 416         struct rt6_info *rt = (struct rt6_info *)dst;
 417         struct inet6_dev *idev = rt->rt6i_idev;
 418         struct net_device *loopback_dev =
 419                 dev_net(dev)->loopback_dev;
 420
 421         if (dev != loopback_dev) {
 422                 if (idev && idev->dev == dev) {
 423                         struct inet6_dev *loopback_idev =
 424                                 in6_dev_get(loopback_dev);
 425                         if (loopback_idev) {
 426                                 rt->rt6i_idev = loopback_idev;
 427                                 in6_dev_put(idev);
 428                         }
 429                 }
 430         }
 431 }
 432
 433 static bool __rt6_check_expired(const struct rt6_info *rt)
 434 {
 435         if (rt->rt6i_flags & RTF_EXPIRES)
 436                 return time_after(jiffies, rt->dst.expires);
 437         else
 438                 return false;
 439 }
 440
 441 static bool rt6_check_expired(const struct rt6_info *rt)
 442 {
 443         if (rt->rt6i_flags & RTF_EXPIRES) {
 444                 if (time_after(jiffies, rt->dst.expires))
 445                         return true;
 446         } else if (rt->dst.from) {
 447                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 448         }
 449         return false;
 450 }
 451
 452 /* Multipath route selection:
 453  *   Hash based function using packet header and flowlabel.
 454  * Adapted from fib_info_hashfn()
 455  */
 456 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 457                                const struct flowi6 *fl6)
 458 {
 459         return get_hash_from_flowi6(fl6) % candidate_count;
 460 }
 461
 462 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 463                                              struct flowi6 *fl6, int oif,
 464                                              int strict)
 465 {
 466         struct rt6_info *sibling, *next_sibling;
 467         int route_choosen;
 468
 469         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 470         /* Don't change the route, if route_choosen == 0
 471          * (siblings does not include ourself)
 472          */
 473         if (route_choosen)
 474                 list_for_each_entry_safe(sibling, next_sibling,
 475                                 &match->rt6i_siblings, rt6i_siblings) {
 476                         route_choosen--;
 477                         if (route_choosen == 0) {
 478                                 if (rt6_score_route(sibling, oif, strict) < 0)
 479                                         break;
 480                                 match = sibling;
 481                                 break;
 482                         }
 483                 }
 484         return match;
 485 }
 486
 487 /*
 488  *      Route lookup. Any table->tb6_lock is implied.
 489  */
 490
 491 static inline struct rt6_info *rt6_device_match(struct net *net,
 492                                                     struct rt6_info *rt,
 493                                                     const struct in6_addr *saddr,
 494                                                     int oif,
 495                                                     int flags)
 496 {
 497         struct rt6_info *local = NULL;
 498         struct rt6_info *sprt;
 499
 500         if (!oif && ipv6_addr_any(saddr))
 501                 goto out;
 502
 503         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 504                 struct net_device *dev = sprt->dst.dev;
 505
 506                 if (oif) {
 507                         if (dev->ifindex == oif)
 508                                 return sprt;
 509                         if (dev->flags & IFF_LOOPBACK) {
 510                                 if (!sprt->rt6i_idev ||
 511                                     sprt->rt6i_idev->dev->ifindex != oif) {
 512                                         if (flags & RT6_LOOKUP_F_IFACE)
 513                                                 continue;
 514                                         if (local &&
 515                                             local->rt6i_idev->dev->ifindex == oif)
 516                                                 continue;
 517                                 }
 518                                 local = sprt;
 519                         }
 520                 } else {
 521                         if (ipv6_chk_addr(net, saddr, dev,
 522                                           flags & RT6_LOOKUP_F_IFACE))
 523                                 return sprt;
 524                 }
 525         }
 526
 527         if (oif) {
 528                 if (local)
 529                         return local;
 530
 531                 if (flags & RT6_LOOKUP_F_IFACE)
 532                         return net->ipv6.ip6_null_entry;
 533         }
 534 out:
 535         return rt;
 536 }
 537
 538 #ifdef CONFIG_IPV6_ROUTER_PREF
 539 struct __rt6_probe_work {
 540         struct work_struct work;
 541         struct in6_addr target;
 542         struct net_device *dev;
 543 };
 544
 545 static void rt6_probe_deferred(struct work_struct *w)
 546 {
 547         struct in6_addr mcaddr;
 548         struct __rt6_probe_work *work =
 549                 container_of(w, struct __rt6_probe_work, work);
 550
 551         addrconf_addr_solict_mult(&work->target, &mcaddr);
 552         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
 553         dev_put(work->dev);
 554         kfree(work);
 555 }
 556
 557 static void rt6_probe(struct rt6_info *rt)
 558 {
 559         struct __rt6_probe_work *work;
 560         struct neighbour *neigh;
 561         /*
 562          * Okay, this does not seem to be appropriate
 563          * for now, however, we need to check if it
 564          * is really so; aka Router Reachability Probing.
 565          *
 566          * Router Reachability Probe MUST be rate-limited
 567          * to no more than one per minute.
 568          */
 569         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 570                 return;
 571         rcu_read_lock_bh();
 572         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 573         if (neigh) {
 574                 if (neigh->nud_state & NUD_VALID)
 575                         goto out;
 576
 577                 work = NULL;
 578                 write_lock(&neigh->lock);
 579                 if (!(neigh->nud_state & NUD_VALID) &&
 580                     time_after(jiffies,
 581                                neigh->updated +
 582                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 583                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 584                         if (work)
 585                                 __neigh_set_probe_once(neigh);
 586                 }
 587                 write_unlock(&neigh->lock);
 588         } else {
 589                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 590         }
 591
 592         if (work) {
 593                 INIT_WORK(&work->work, rt6_probe_deferred);
 594                 work->target = rt->rt6i_gateway;
 595                 dev_hold(rt->dst.dev);
 596                 work->dev = rt->dst.dev;
 597                 schedule_work(&work->work);
 598         }
 599
 600 out:
 601         rcu_read_unlock_bh();
 602 }
 603 #else
 604 static inline void rt6_probe(struct rt6_info *rt)
 605 {
 606 }
 607 #endif
 608
 609 /*
 610  * Default Router Selection (RFC 2461 6.3.6)
 611  */
 612 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 613 {
 614         struct net_device *dev = rt->dst.dev;
 615         if (!oif || dev->ifindex == oif)
 616                 return 2;
 617         if ((dev->flags & IFF_LOOPBACK) &&
 618             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 619                 return 1;
 620         return 0;
 621 }
 622
 623 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 624 {
 625         struct neighbour *neigh;
 626         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 627
 628         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 629             !(rt->rt6i_flags & RTF_GATEWAY))
 630                 return RT6_NUD_SUCCEED;
 631
 632         rcu_read_lock_bh();
 633         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 634         if (neigh) {
 635                 read_lock(&neigh->lock);
 636                 if (neigh->nud_state & NUD_VALID)
 637                         ret = RT6_NUD_SUCCEED;
 638 #ifdef CONFIG_IPV6_ROUTER_PREF
 639                 else if (!(neigh->nud_state & NUD_FAILED))
 640                         ret = RT6_NUD_SUCCEED;
 641                 else
 642                         ret = RT6_NUD_FAIL_PROBE;
 643 #endif
 644                 read_unlock(&neigh->lock);
 645         } else {
 646                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 647                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 648         }
 649         rcu_read_unlock_bh();
 650
 651         return ret;
 652 }
 653
 654 static int rt6_score_route(struct rt6_info *rt, int oif,
 655                            int strict)
 656 {
 657         int m;
 658
 659         m = rt6_check_dev(rt, oif);
 660         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 661                 return RT6_NUD_FAIL_HARD;
 662 #ifdef CONFIG_IPV6_ROUTER_PREF
 663         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 664 #endif
 665         if (strict & RT6_LOOKUP_F_REACHABLE) {
 666                 int n = rt6_check_neigh(rt);
 667                 if (n < 0)
 668                         return n;
 669         }
 670         return m;
 671 }
 672
 673 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 674                                    int *mpri, struct rt6_info *match,
 675                                    bool *do_rr)
 676 {
 677         int m;
 678         bool match_do_rr = false;
 679         struct inet6_dev *idev = rt->rt6i_idev;
 680         struct net_device *dev = rt->dst.dev;
 681
 682         if (dev && !netif_carrier_ok(dev) &&
 683             idev->cnf.ignore_routes_with_linkdown &&
 684             !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
 685                 goto out;
 686
 687         if (rt6_check_expired(rt))
 688                 goto out;
 689
 690         m = rt6_score_route(rt, oif, strict);
 691         if (m == RT6_NUD_FAIL_DO_RR) {
 692                 match_do_rr = true;
 693                 m = 0; /* lowest valid score */
 694         } else if (m == RT6_NUD_FAIL_HARD) {
 695                 goto out;
 696         }
 697
 698         if (strict & RT6_LOOKUP_F_REACHABLE)
 699                 rt6_probe(rt);
 700
 701         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 702         if (m > *mpri) {
 703                 *do_rr = match_do_rr;
 704                 *mpri = m;
 705                 match = rt;
 706         }
 707 out:
 708         return match;
 709 }
 710
 711 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 712                                      struct rt6_info *rr_head,
 713                                      u32 metric, int oif, int strict,
 714                                      bool *do_rr)
 715 {
 716         struct rt6_info *rt, *match, *cont;
 717         int mpri = -1;
 718
 719         match = NULL;
 720         cont = NULL;
 721         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 722                 if (rt->rt6i_metric != metric) {
 723                         cont = rt;
 724                         break;
 725                 }
 726
 727                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 728         }
 729
 730         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 731                 if (rt->rt6i_metric != metric) {
 732                         cont = rt;
 733                         break;
 734                 }
 735
 736                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 737         }
 738
 739         if (match || !cont)
 740                 return match;
 741
 742         for (rt = cont; rt; rt = rt->dst.rt6_next)
 743                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 744
 745         return match;
 746 }
 747
 748 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 749 {
 750         struct rt6_info *match, *rt0;
 751         struct net *net;
 752         bool do_rr = false;
 753
 754         rt0 = fn->rr_ptr;
 755         if (!rt0)
 756                 fn->rr_ptr = rt0 = fn->leaf;
 757
 758         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 759                              &do_rr);
 760
 761         if (do_rr) {
 762                 struct rt6_info *next = rt0->dst.rt6_next;
 763
 764                 /* no entries matched; do round-robin */
 765                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 766                         next = fn->leaf;
 767
 768                 if (next != rt0)
 769                         fn->rr_ptr = next;
 770         }
 771
 772         net = dev_net(rt0->dst.dev);
 773         return match ? match : net->ipv6.ip6_null_entry;
 774 }
 775
 776 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 777 {
 778         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 779 }
 780
 781 #ifdef CONFIG_IPV6_ROUTE_INFO
 782 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 783                   const struct in6_addr *gwaddr)
 784 {
 785         struct net *net = dev_net(dev);
 786         struct route_info *rinfo = (struct route_info *) opt;
 787         struct in6_addr prefix_buf, *prefix;
 788         unsigned int pref;
 789         unsigned long lifetime;
 790         struct rt6_info *rt;
 791
 792         if (len < sizeof(struct route_info)) {
 793                 return -EINVAL;
 794         }
 795
 796         /* Sanity check for prefix_len and length */
 797         if (rinfo->length > 3) {
 798                 return -EINVAL;
 799         } else if (rinfo->prefix_len > 128) {
 800                 return -EINVAL;
 801         } else if (rinfo->prefix_len > 64) {
 802                 if (rinfo->length < 2) {
 803                         return -EINVAL;
 804                 }
 805         } else if (rinfo->prefix_len > 0) {
 806                 if (rinfo->length < 1) {
 807                         return -EINVAL;
 808                 }
 809         }
 810
 811         pref = rinfo->route_pref;
 812         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 813                 return -EINVAL;
 814
 815         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 816
 817         if (rinfo->length == 3)
 818                 prefix = (struct in6_addr *)rinfo->prefix;
 819         else {
 820                 /* this function is safe */
 821                 ipv6_addr_prefix(&prefix_buf,
 822                                  (struct in6_addr *)rinfo->prefix,
 823                                  rinfo->prefix_len);
 824                 prefix = &prefix_buf;
 825         }
 826
 827         if (rinfo->prefix_len == 0)
 828                 rt = rt6_get_dflt_router(gwaddr, dev);
 829         else
 830                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 831                                         gwaddr, dev);
 832
 833         if (rt && !lifetime) {
 834                 ip6_del_rt(rt);
 835                 rt = NULL;
 836         }
 837
 838         if (!rt && lifetime)
 839                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 840                                         dev, pref);
 841         else if (rt)
 842                 rt->rt6i_flags = RTF_ROUTEINFO |
 843                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 844
 845         if (rt) {
 846                 if (!addrconf_finite_timeout(lifetime))
 847                         rt6_clean_expires(rt);
 848                 else
 849                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 850
 851                 ip6_rt_put(rt);
 852         }
 853         return 0;
 854 }
 855 #endif
 856
 857 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 858                                         struct in6_addr *saddr)
 859 {
 860         struct fib6_node *pn;
 861         while (1) {
 862                 if (fn->fn_flags & RTN_TL_ROOT)
 863                         return NULL;
 864                 pn = fn->parent;
 865                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 866                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 867                 else
 868                         fn = pn;
 869                 if (fn->fn_flags & RTN_RTINFO)
 870                         return fn;
 871         }
 872 }
 873
 874 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 875                                              struct fib6_table *table,
 876                                              struct flowi6 *fl6, int flags)
 877 {
 878         struct fib6_node *fn;
 879         struct rt6_info *rt;
 880
 881         read_lock_bh(&table->tb6_lock);
 882         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 883 restart:
 884         rt = fn->leaf;
 885         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 886         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 887                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 888         if (rt == net->ipv6.ip6_null_entry) {
 889                 fn = fib6_backtrack(fn, &fl6->saddr);
 890                 if (fn)
 891                         goto restart;
 892         }
 893         dst_use(&rt->dst, jiffies);
 894         read_unlock_bh(&table->tb6_lock);
 895
 896         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
 897
 898         return rt;
 899
 900 }
 901
 902 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 903                                     int flags)
 904 {
 905         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 906 }
 907 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 908
 909 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 910                             const struct in6_addr *saddr, int oif, int strict)
 911 {
 912         struct flowi6 fl6 = {
 913                 .flowi6_oif = oif,
 914                 .daddr = *daddr,
 915         };
 916         struct dst_entry *dst;
 917         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 918
 919         if (saddr) {
 920                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 921                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 922         }
 923
 924         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 925         if (dst->error == 0)
 926                 return (struct rt6_info *) dst;
 927
 928         dst_release(dst);
 929
 930         return NULL;
 931 }
 932 EXPORT_SYMBOL(rt6_lookup);
 933
 934 /* ip6_ins_rt is called with FREE table->tb6_lock.
 935    It takes new route entry, the addition fails by any reason the
 936    route is freed. In any case, if caller does not hold it, it may
 937    be destroyed.
 938  */
 939
 940 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 941                         struct mx6_config *mxc)
 942 {
 943         int err;
 944         struct fib6_table *table;
 945
 946         table = rt->rt6i_table;
 947         write_lock_bh(&table->tb6_lock);
 948         err = fib6_add(&table->tb6_root, rt, info, mxc);
 949         write_unlock_bh(&table->tb6_lock);
 950
 951         return err;
 952 }
 953
 954 int ip6_ins_rt(struct rt6_info *rt)
 955 {
 956         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 957         struct mx6_config mxc = { .mx = NULL, };
 958
 959         return __ip6_ins_rt(rt, &info, &mxc);
 960 }
 961
 962 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 963                                            const struct in6_addr *daddr,
 964                                            const struct in6_addr *saddr)
 965 {
 966         struct rt6_info *rt;
 967
 968         /*
 969          *      Clone the route.
 970          */
 971
 972         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 973                 ort = (struct rt6_info *)ort->dst.from;
 974
 975         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 976
 977         if (!rt)
 978                 return NULL;
 979
 980         ip6_rt_copy_init(rt, ort);
 981         rt->rt6i_flags |= RTF_CACHE;
 982         rt->rt6i_metric = 0;
 983         rt->dst.flags |= DST_HOST;
 984         rt->rt6i_dst.addr = *daddr;
 985         rt->rt6i_dst.plen = 128;
 986
 987         if (!rt6_is_gw_or_nonexthop(ort)) {
 988                 if (ort->rt6i_dst.plen != 128 &&
 989                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 990                         rt->rt6i_flags |= RTF_ANYCAST;
 991 #ifdef CONFIG_IPV6_SUBTREES
 992                 if (rt->rt6i_src.plen && saddr) {
 993                         rt->rt6i_src.addr = *saddr;
 994                         rt->rt6i_src.plen = 128;
 995                 }
 996 #endif
 997         }
 998
 999         return rt;
1000 }
1001
1002 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
1003 {
1004         struct rt6_info *pcpu_rt;
1005
1006         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
1007                                   rt->dst.dev, rt->dst.flags);
1008
1009         if (!pcpu_rt)
1010                 return NULL;
1011         ip6_rt_copy_init(pcpu_rt, rt);
1012         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
1013         pcpu_rt->rt6i_flags |= RTF_PCPU;
1014         return pcpu_rt;
1015 }
1016
1017 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1018 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1019 {
1020         struct rt6_info *pcpu_rt, **p;
1021
1022         p = this_cpu_ptr(rt->rt6i_pcpu);
1023         pcpu_rt = *p;
1024
1025         if (pcpu_rt) {
1026                 dst_hold(&pcpu_rt->dst);
1027                 rt6_dst_from_metrics_check(pcpu_rt);
1028         }
1029         return pcpu_rt;
1030 }
1031
1032 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1033 {
1034         struct fib6_table *table = rt->rt6i_table;
1035         struct rt6_info *pcpu_rt, *prev, **p;
1036
1037         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1038         if (!pcpu_rt) {
1039                 struct net *net = dev_net(rt->dst.dev);
1040
1041                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1042                 return net->ipv6.ip6_null_entry;
1043         }
1044
1045         read_lock_bh(&table->tb6_lock);
1046         if (rt->rt6i_pcpu) {
1047                 p = this_cpu_ptr(rt->rt6i_pcpu);
1048                 prev = cmpxchg(p, NULL, pcpu_rt);
1049                 if (prev) {
1050                         /* If someone did it before us, return prev instead */
1051                         dst_destroy(&pcpu_rt->dst);
1052                         pcpu_rt = prev;
1053                 }
1054         } else {
1055                 /* rt has been removed from the fib6 tree
1056                  * before we have a chance to acquire the read_lock.
1057                  * In this case, don't brother to create a pcpu rt
1058                  * since rt is going away anyway.  The next
1059                  * dst_check() will trigger a re-lookup.
1060                  */
1061                 dst_destroy(&pcpu_rt->dst);
1062                 pcpu_rt = rt;
1063         }
1064         dst_hold(&pcpu_rt->dst);
1065         rt6_dst_from_metrics_check(pcpu_rt);
1066         read_unlock_bh(&table->tb6_lock);
1067         return pcpu_rt;
1068 }
1069
1070 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1071                                int oif, struct flowi6 *fl6, int flags)
1072 {
1073         struct fib6_node *fn, *saved_fn;
1074         struct rt6_info *rt;
1075         int strict = 0;
1076
1077         strict |= flags & RT6_LOOKUP_F_IFACE;
1078         strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1079         if (net->ipv6.devconf_all->forwarding == 0)
1080                 strict |= RT6_LOOKUP_F_REACHABLE;
1081
1082         read_lock_bh(&table->tb6_lock);
1083
1084         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1085         saved_fn = fn;
1086
1087         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1088                 oif = 0;
1089
1090 redo_rt6_select:
1091         rt = rt6_select(fn, oif, strict);
1092         if (rt->rt6i_nsiblings)
1093                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1094         if (rt == net->ipv6.ip6_null_entry) {
1095                 fn = fib6_backtrack(fn, &fl6->saddr);
1096                 if (fn)
1097                         goto redo_rt6_select;
1098                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1099                         /* also consider unreachable route */
1100                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1101                         fn = saved_fn;
1102                         goto redo_rt6_select;
1103                 }
1104         }
1105
1106
1107         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1108                 dst_use(&rt->dst, jiffies);
1109                 read_unlock_bh(&table->tb6_lock);
1110
1111                 rt6_dst_from_metrics_check(rt);
1112
1113                 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1114                 return rt;
1115         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1116                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1117                 /* Create a RTF_CACHE clone which will not be
1118                  * owned by the fib6 tree.  It is for the special case where
1119                  * the daddr in the skb during the neighbor look-up is different
1120                  * from the fl6->daddr used to look-up route here.
1121                  */
1122
1123                 struct rt6_info *uncached_rt;
1124
1125                 dst_use(&rt->dst, jiffies);
1126                 read_unlock_bh(&table->tb6_lock);
1127
1128                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1129                 dst_release(&rt->dst);
1130
1131                 if (uncached_rt)
1132                         rt6_uncached_list_add(uncached_rt);
1133                 else
1134                         uncached_rt = net->ipv6.ip6_null_entry;
1135
1136                 dst_hold(&uncached_rt->dst);
1137
1138                 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1139                 return uncached_rt;
1140
1141         } else {
1142                 /* Get a percpu copy */
1143
1144                 struct rt6_info *pcpu_rt;
1145
1146                 rt->dst.lastuse = jiffies;
1147                 rt->dst.__use++;
1148                 pcpu_rt = rt6_get_pcpu_route(rt);
1149
1150                 if (pcpu_rt) {
1151                         read_unlock_bh(&table->tb6_lock);
1152                 } else {
1153                         /* We have to do the read_unlock first
1154                          * because rt6_make_pcpu_route() may trigger
1155                          * ip6_dst_gc() which will take the write_lock.
1156                          */
1157                         dst_hold(&rt->dst);
1158                         read_unlock_bh(&table->tb6_lock);
1159                         pcpu_rt = rt6_make_pcpu_route(rt);
1160                         dst_release(&rt->dst);
1161                 }
1162
1163                 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1164                 return pcpu_rt;
1165
1166         }
1167 }
1168 EXPORT_SYMBOL_GPL(ip6_pol_route);
1169
1170 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1171                                             struct flowi6 *fl6, int flags)
1172 {
1173         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1174 }
1175
1176 struct dst_entry *ip6_route_input_lookup(struct net *net,
1177                                          struct net_device *dev,
1178                                          struct flowi6 *fl6, int flags)
1179 {
1180         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1181                 flags |= RT6_LOOKUP_F_IFACE;
1182
1183         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1184 }
1185 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1186
1187 void ip6_route_input(struct sk_buff *skb)
1188 {
1189         const struct ipv6hdr *iph = ipv6_hdr(skb);
1190         struct net *net = dev_net(skb->dev);
1191         int flags = RT6_LOOKUP_F_HAS_SADDR;
1192         struct ip_tunnel_info *tun_info;
1193         struct flowi6 fl6 = {
1194                 .flowi6_iif = skb->dev->ifindex,
1195                 .daddr = iph->daddr,
1196                 .saddr = iph->saddr,
1197                 .flowlabel = ip6_flowinfo(iph),
1198                 .flowi6_mark = skb->mark,
1199                 .flowi6_proto = iph->nexthdr,
1200         };
1201
1202         tun_info = skb_tunnel_info(skb);
1203         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1204                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1205         skb_dst_drop(skb);
1206         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1207 }
1208
1209 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1210                                              struct flowi6 *fl6, int flags)
1211 {
1212         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1213 }
1214
1215 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1216                                          struct flowi6 *fl6, int flags)
1217 {
1218         bool any_src;
1219
1220         if (rt6_need_strict(&fl6->daddr)) {
1221                 struct dst_entry *dst;
1222
1223                 dst = l3mdev_link_scope_lookup(net, fl6);
1224                 if (dst)
1225                         return dst;
1226         }
1227
1228         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1229
1230         any_src = ipv6_addr_any(&fl6->saddr);
1231         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1232             (fl6->flowi6_oif && any_src))
1233                 flags |= RT6_LOOKUP_F_IFACE;
1234
1235         if (!any_src)
1236                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1237         else if (sk)
1238                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1239
1240         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1241 }
1242 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1243
1244 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1245 {
1246         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1247         struct dst_entry *new = NULL;
1248
1249         rt = dst_alloc(&ip6_dst_blackhole_ops, ort->dst.dev, 1, DST_OBSOLETE_NONE, 0);
1250         if (rt) {
1251                 rt6_info_init(rt);
1252
1253                 new = &rt->dst;
1254                 new->__use = 1;
1255                 new->input = dst_discard;
1256                 new->output = dst_discard_out;
1257
1258                 dst_copy_metrics(new, &ort->dst);
1259                 rt->rt6i_idev = ort->rt6i_idev;
1260                 if (rt->rt6i_idev)
1261                         in6_dev_hold(rt->rt6i_idev);
1262
1263                 rt->rt6i_gateway = ort->rt6i_gateway;
1264                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1265                 rt->rt6i_metric = 0;
1266
1267                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1268 #ifdef CONFIG_IPV6_SUBTREES
1269                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1270 #endif
1271
1272                 dst_free(new);
1273         }
1274
1275         dst_release(dst_orig);
1276         return new ? new : ERR_PTR(-ENOMEM);
1277 }
1278
1279 /*
1280  *      Destination cache support functions
1281  */
1282
1283 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1284 {
1285         if (rt->dst.from &&
1286             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1287                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1288 }
1289
1290 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1291 {
1292         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1293                 return NULL;
1294
1295         if (rt6_check_expired(rt))
1296                 return NULL;
1297
1298         return &rt->dst;
1299 }
1300
1301 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1302 {
1303         if (!__rt6_check_expired(rt) &&
1304             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1305             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1306                 return &rt->dst;
1307         else
1308                 return NULL;
1309 }
1310
1311 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1312 {
1313         struct rt6_info *rt;
1314
1315         rt = (struct rt6_info *) dst;
1316
1317         /* All IPV6 dsts are created with ->obsolete set to the value
1318          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1319          * into this function always.
1320          */
1321
1322         rt6_dst_from_metrics_check(rt);
1323
1324         if (rt->rt6i_flags & RTF_PCPU ||
1325             (unlikely(dst->flags & DST_NOCACHE) && rt->dst.from))
1326                 return rt6_dst_from_check(rt, cookie);
1327         else
1328                 return rt6_check(rt, cookie);
1329 }
1330
1331 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1332 {
1333         struct rt6_info *rt = (struct rt6_info *) dst;
1334
1335         if (rt) {
1336                 if (rt->rt6i_flags & RTF_CACHE) {
1337                         if (rt6_check_expired(rt)) {
1338                                 ip6_del_rt(rt);
1339                                 dst = NULL;
1340                         }
1341                 } else {
1342                         dst_release(dst);
1343                         dst = NULL;
1344                 }
1345         }
1346         return dst;
1347 }
1348
1349 static void ip6_link_failure(struct sk_buff *skb)
1350 {
1351         struct rt6_info *rt;
1352
1353         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1354
1355         rt = (struct rt6_info *) skb_dst(skb);
1356         if (rt) {
1357                 if (rt->rt6i_flags & RTF_CACHE) {
1358                         dst_hold(&rt->dst);
1359                         ip6_del_rt(rt);
1360                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1361                         rt->rt6i_node->fn_sernum = -1;
1362                 }
1363         }
1364 }
1365
1366 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1367 {
1368         struct net *net = dev_net(rt->dst.dev);
1369
1370         rt->rt6i_flags |= RTF_MODIFIED;
1371         rt->rt6i_pmtu = mtu;
1372         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1373 }
1374
1375 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1376 {
1377         return !(rt->rt6i_flags & RTF_CACHE) &&
1378                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1379 }
1380
1381 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1382                                  const struct ipv6hdr *iph, u32 mtu)
1383 {
1384         const struct in6_addr *daddr, *saddr;
1385         struct rt6_info *rt6 = (struct rt6_info *)dst;
1386
1387         if (rt6->rt6i_flags & RTF_LOCAL)
1388                 return;
1389
1390         if (dst_metric_locked(dst, RTAX_MTU))
1391                 return;
1392
1393         if (iph) {
1394                 daddr = &iph->daddr;
1395                 saddr = &iph->saddr;
1396         } else if (sk) {
1397                 daddr = &sk->sk_v6_daddr;
1398                 saddr = &inet6_sk(sk)->saddr;
1399         } else {
1400                 daddr = NULL;
1401                 saddr = NULL;
1402         }
1403         dst_confirm_neigh(dst, daddr);
1404         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1405         if (mtu >= dst_mtu(dst))
1406                 return;
1407
1408         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1409                 rt6_do_update_pmtu(rt6, mtu);
1410         } else if (daddr) {
1411                 struct rt6_info *nrt6;
1412
1413                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1414                 if (nrt6) {
1415                         rt6_do_update_pmtu(nrt6, mtu);
1416
1417                         /* ip6_ins_rt(nrt6) will bump the
1418                          * rt6->rt6i_node->fn_sernum
1419                          * which will fail the next rt6_check() and
1420                          * invalidate the sk->sk_dst_cache.
1421                          */
1422                         ip6_ins_rt(nrt6);
1423                 }
1424         }
1425 }
1426
1427 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1428                                struct sk_buff *skb, u32 mtu)
1429 {
1430         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1431 }
1432
1433 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1434                      int oif, u32 mark, kuid_t uid)
1435 {
1436         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1437         struct dst_entry *dst;
1438         struct flowi6 fl6;
1439
1440         memset(&fl6, 0, sizeof(fl6));
1441         fl6.flowi6_oif = oif;
1442         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1443         fl6.daddr = iph->daddr;
1444         fl6.saddr = iph->saddr;
1445         fl6.flowlabel = ip6_flowinfo(iph);
1446         fl6.flowi6_uid = uid;
1447
1448         dst = ip6_route_output(net, NULL, &fl6);
1449         if (!dst->error)
1450                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1451         dst_release(dst);
1452 }
1453 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1454
1455 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1456 {
1457         struct dst_entry *dst;
1458
1459         ip6_update_pmtu(skb, sock_net(sk), mtu,
1460                         sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
1461
1462         dst = __sk_dst_get(sk);
1463         if (!dst || !dst->obsolete ||
1464             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1465                 return;
1466
1467         bh_lock_sock(sk);
1468         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1469                 ip6_datagram_dst_update(sk, false);
1470         bh_unlock_sock(sk);
1471 }
1472 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1473
1474 /* Handle redirects */
1475 struct ip6rd_flowi {
1476         struct flowi6 fl6;
1477         struct in6_addr gateway;
1478 };
1479
1480 static struct rt6_info *__ip6_route_redirect(struct net *net,
1481                                              struct fib6_table *table,
1482                                              struct flowi6 *fl6,
1483                                              int flags)
1484 {
1485         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1486         struct rt6_info *rt;
1487         struct fib6_node *fn;
1488
1489         /* Get the "current" route for this destination and
1490          * check if the redirect has come from appropriate router.
1491          *
1492          * RFC 4861 specifies that redirects should only be
1493          * accepted if they come from the nexthop to the target.
1494          * Due to the way the routes are chosen, this notion
1495          * is a bit fuzzy and one might need to check all possible
1496          * routes.
1497          */
1498
1499         read_lock_bh(&table->tb6_lock);
1500         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1501 restart:
1502         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1503                 if (rt6_check_expired(rt))
1504                         continue;
1505                 if (rt->dst.error)
1506                         break;
1507                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1508                         continue;
1509                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1510                         continue;
1511                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1512                         continue;
1513                 break;
1514         }
1515
1516         if (!rt)
1517                 rt = net->ipv6.ip6_null_entry;
1518         else if (rt->dst.error) {
1519                 rt = net->ipv6.ip6_null_entry;
1520                 goto out;
1521         }
1522
1523         if (rt == net->ipv6.ip6_null_entry) {
1524                 fn = fib6_backtrack(fn, &fl6->saddr);
1525                 if (fn)
1526                         goto restart;
1527         }
1528
1529 out:
1530         dst_hold(&rt->dst);
1531
1532         read_unlock_bh(&table->tb6_lock);
1533
1534         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1535         return rt;
1536 };
1537
1538 static struct dst_entry *ip6_route_redirect(struct net *net,
1539                                         const struct flowi6 *fl6,
1540                                         const struct in6_addr *gateway)
1541 {
1542         int flags = RT6_LOOKUP_F_HAS_SADDR;
1543         struct ip6rd_flowi rdfl;
1544
1545         rdfl.fl6 = *fl6;
1546         rdfl.gateway = *gateway;
1547
1548         return fib6_rule_lookup(net, &rdfl.fl6,
1549                                 flags, __ip6_route_redirect);
1550 }
1551
1552 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1553                   kuid_t uid)
1554 {
1555         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1556         struct dst_entry *dst;
1557         struct flowi6 fl6;
1558
1559         memset(&fl6, 0, sizeof(fl6));
1560         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1561         fl6.flowi6_oif = oif;
1562         fl6.flowi6_mark = mark;
1563         fl6.daddr = iph->daddr;
1564         fl6.saddr = iph->saddr;
1565         fl6.flowlabel = ip6_flowinfo(iph);
1566         fl6.flowi6_uid = uid;
1567
1568         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1569         rt6_do_redirect(dst, NULL, skb);
1570         dst_release(dst);
1571 }
1572 EXPORT_SYMBOL_GPL(ip6_redirect);
1573
1574 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1575                             u32 mark)
1576 {
1577         const struct ipv6hdr *iph = ipv6_hdr(skb);
1578         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1579         struct dst_entry *dst;
1580         struct flowi6 fl6;
1581
1582         memset(&fl6, 0, sizeof(fl6));
1583         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1584         fl6.flowi6_oif = oif;
1585         fl6.flowi6_mark = mark;
1586         fl6.daddr = msg->dest;
1587         fl6.saddr = iph->daddr;
1588         fl6.flowi6_uid = sock_net_uid(net, NULL);
1589
1590         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1591         rt6_do_redirect(dst, NULL, skb);
1592         dst_release(dst);
1593 }
1594
1595 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1596 {
1597         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1598                      sk->sk_uid);
1599 }
1600 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1601
1602 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1603 {
1604         struct net_device *dev = dst->dev;
1605         unsigned int mtu = dst_mtu(dst);
1606         struct net *net = dev_net(dev);
1607
1608         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1609
1610         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1611                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1612
1613         /*
1614          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1615          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1616          * IPV6_MAXPLEN is also valid and means: "any MSS,
1617          * rely only on pmtu discovery"
1618          */
1619         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1620                 mtu = IPV6_MAXPLEN;
1621         return mtu;
1622 }
1623
1624 static unsigned int ip6_mtu(const struct dst_entry *dst)
1625 {
1626         const struct rt6_info *rt = (const struct rt6_info *)dst;
1627         unsigned int mtu = rt->rt6i_pmtu;
1628         struct inet6_dev *idev;
1629
1630         if (mtu)
1631                 goto out;
1632
1633         mtu = dst_metric_raw(dst, RTAX_MTU);
1634         if (mtu)
1635                 goto out;
1636
1637         mtu = IPV6_MIN_MTU;
1638
1639         rcu_read_lock();
1640         idev = __in6_dev_get(dst->dev);
1641         if (idev)
1642                 mtu = idev->cnf.mtu6;
1643         rcu_read_unlock();
1644
1645 out:
1646         mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1647
1648         return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1649 }
1650
1651 static struct dst_entry *icmp6_dst_gc_list;
1652 static DEFINE_SPINLOCK(icmp6_dst_lock);
1653
1654 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1655                                   struct flowi6 *fl6)
1656 {
1657         struct dst_entry *dst;
1658         struct rt6_info *rt;
1659         struct inet6_dev *idev = in6_dev_get(dev);
1660         struct net *net = dev_net(dev);
1661
1662         if (unlikely(!idev))
1663                 return ERR_PTR(-ENODEV);
1664
1665         rt = ip6_dst_alloc(net, dev, 0);
1666         if (unlikely(!rt)) {
1667                 in6_dev_put(idev);
1668                 dst = ERR_PTR(-ENOMEM);
1669                 goto out;
1670         }
1671
1672         rt->dst.flags |= DST_HOST;
1673         rt->dst.output  = ip6_output;
1674         atomic_set(&rt->dst.__refcnt, 1);
1675         rt->rt6i_gateway  = fl6->daddr;
1676         rt->rt6i_dst.addr = fl6->daddr;
1677         rt->rt6i_dst.plen = 128;
1678         rt->rt6i_idev     = idev;
1679         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1680
1681         spin_lock_bh(&icmp6_dst_lock);
1682         rt->dst.next = icmp6_dst_gc_list;
1683         icmp6_dst_gc_list = &rt->dst;
1684         spin_unlock_bh(&icmp6_dst_lock);
1685
1686         fib6_force_start_gc(net);
1687
1688         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1689
1690 out:
1691         return dst;
1692 }
1693
1694 int icmp6_dst_gc(void)
1695 {
1696         struct dst_entry *dst, **pprev;
1697         int more = 0;
1698
1699         spin_lock_bh(&icmp6_dst_lock);
1700         pprev = &icmp6_dst_gc_list;
1701
1702         while ((dst = *pprev) != NULL) {
1703                 if (!atomic_read(&dst->__refcnt)) {
1704                         *pprev = dst->next;
1705                         dst_free(dst);
1706                 } else {
1707                         pprev = &dst->next;
1708                         ++more;
1709                 }
1710         }
1711
1712         spin_unlock_bh(&icmp6_dst_lock);
1713
1714         return more;
1715 }
1716
1717 static void icmp6_clean_all(int (*func)(struct rt6_info *rt, void *arg),
1718                             void *arg)
1719 {
1720         struct dst_entry *dst, **pprev;
1721
1722         spin_lock_bh(&icmp6_dst_lock);
1723         pprev = &icmp6_dst_gc_list;
1724         while ((dst = *pprev) != NULL) {
1725                 struct rt6_info *rt = (struct rt6_info *) dst;
1726                 if (func(rt, arg)) {
1727                         *pprev = dst->next;
1728                         dst_free(dst);
1729                 } else {
1730                         pprev = &dst->next;
1731                 }
1732         }
1733         spin_unlock_bh(&icmp6_dst_lock);
1734 }
1735
1736 static int ip6_dst_gc(struct dst_ops *ops)
1737 {
1738         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1739         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1740         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1741         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1742         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1743         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1744         int entries;
1745
1746         entries = dst_entries_get_fast(ops);
1747         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1748             entries <= rt_max_size)
1749                 goto out;
1750
1751         net->ipv6.ip6_rt_gc_expire++;
1752         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1753         entries = dst_entries_get_slow(ops);
1754         if (entries < ops->gc_thresh)
1755                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1756 out:
1757         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1758         return entries > rt_max_size;
1759 }
1760
1761 static int ip6_convert_metrics(struct mx6_config *mxc,
1762                                const struct fib6_config *cfg)
1763 {
1764         bool ecn_ca = false;
1765         struct nlattr *nla;
1766         int remaining;
1767         u32 *mp;
1768
1769         if (!cfg->fc_mx)
1770                 return 0;
1771
1772         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1773         if (unlikely(!mp))
1774                 return -ENOMEM;
1775
1776         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1777                 int type = nla_type(nla);
1778                 u32 val;
1779
1780                 if (!type)
1781                         continue;
1782                 if (unlikely(type > RTAX_MAX))
1783                         goto err;
1784
1785                 if (type == RTAX_CC_ALGO) {
1786                         char tmp[TCP_CA_NAME_MAX];
1787
1788                         nla_strlcpy(tmp, nla, sizeof(tmp));
1789                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1790                         if (val == TCP_CA_UNSPEC)
1791                                 goto err;
1792                 } else {
1793                         val = nla_get_u32(nla);
1794                 }
1795                 if (type == RTAX_HOPLIMIT && val > 255)
1796                         val = 255;
1797                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1798                         goto err;
1799
1800                 mp[type - 1] = val;
1801                 __set_bit(type - 1, mxc->mx_valid);
1802         }
1803
1804         if (ecn_ca) {
1805                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1806                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1807         }
1808
1809         mxc->mx = mp;
1810         return 0;
1811  err:
1812         kfree(mp);
1813         return -EINVAL;
1814 }
1815
1816 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1817                                             struct fib6_config *cfg,
1818                                             const struct in6_addr *gw_addr)
1819 {
1820         struct flowi6 fl6 = {
1821                 .flowi6_oif = cfg->fc_ifindex,
1822                 .daddr = *gw_addr,
1823                 .saddr = cfg->fc_prefsrc,
1824         };
1825         struct fib6_table *table;
1826         struct rt6_info *rt;
1827         int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
1828
1829         table = fib6_get_table(net, cfg->fc_table);
1830         if (!table)
1831                 return NULL;
1832
1833         if (!ipv6_addr_any(&cfg->fc_prefsrc))
1834                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1835
1836         rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1837
1838         /* if table lookup failed, fall back to full lookup */
1839         if (rt == net->ipv6.ip6_null_entry) {
1840                 ip6_rt_put(rt);
1841                 rt = NULL;
1842         }
1843
1844         return rt;
1845 }
1846
1847 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg)
1848 {
1849         struct net *net = cfg->fc_nlinfo.nl_net;
1850         struct rt6_info *rt = NULL;
1851         struct net_device *dev = NULL;
1852         struct inet6_dev *idev = NULL;
1853         struct fib6_table *table;
1854         int addr_type;
1855         int err = -EINVAL;
1856
1857         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1858                 goto out;
1859 #ifndef CONFIG_IPV6_SUBTREES
1860         if (cfg->fc_src_len)
1861                 goto out;
1862 #endif
1863         if (cfg->fc_ifindex) {
1864                 err = -ENODEV;
1865                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1866                 if (!dev)
1867                         goto out;
1868                 idev = in6_dev_get(dev);
1869                 if (!idev)
1870                         goto out;
1871         }
1872
1873         if (cfg->fc_metric == 0)
1874                 cfg->fc_metric = IP6_RT_PRIO_USER;
1875
1876         err = -ENOBUFS;
1877         if (cfg->fc_nlinfo.nlh &&
1878             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1879                 table = fib6_get_table(net, cfg->fc_table);
1880                 if (!table) {
1881                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1882                         table = fib6_new_table(net, cfg->fc_table);
1883                 }
1884         } else {
1885                 table = fib6_new_table(net, cfg->fc_table);
1886         }
1887
1888         if (!table)
1889                 goto out;
1890
1891         rt = ip6_dst_alloc(net, NULL,
1892                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1893
1894         if (!rt) {
1895                 err = -ENOMEM;
1896                 goto out;
1897         }
1898
1899         if (cfg->fc_flags & RTF_EXPIRES)
1900                 rt6_set_expires(rt, jiffies +
1901                                 clock_t_to_jiffies(cfg->fc_expires));
1902         else
1903                 rt6_clean_expires(rt);
1904
1905         if (cfg->fc_protocol == RTPROT_UNSPEC)
1906                 cfg->fc_protocol = RTPROT_BOOT;
1907         rt->rt6i_protocol = cfg->fc_protocol;
1908
1909         addr_type = ipv6_addr_type(&cfg->fc_dst);
1910
1911         if (addr_type & IPV6_ADDR_MULTICAST)
1912                 rt->dst.input = ip6_mc_input;
1913         else if (cfg->fc_flags & RTF_LOCAL)
1914                 rt->dst.input = ip6_input;
1915         else
1916                 rt->dst.input = ip6_forward;
1917
1918         rt->dst.output = ip6_output;
1919
1920         if (cfg->fc_encap) {
1921                 struct lwtunnel_state *lwtstate;
1922
1923                 err = lwtunnel_build_state(cfg->fc_encap_type,
1924                                            cfg->fc_encap, AF_INET6, cfg,
1925                                            &lwtstate);
1926                 if (err)
1927                         goto out;
1928                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1929                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1930                         rt->dst.lwtstate->orig_output = rt->dst.output;
1931                         rt->dst.output = lwtunnel_output;
1932                 }
1933                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1934                         rt->dst.lwtstate->orig_input = rt->dst.input;
1935                         rt->dst.input = lwtunnel_input;
1936                 }
1937         }
1938
1939         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1940         rt->rt6i_dst.plen = cfg->fc_dst_len;
1941         if (rt->rt6i_dst.plen == 128)
1942                 rt->dst.flags |= DST_HOST;
1943
1944 #ifdef CONFIG_IPV6_SUBTREES
1945         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1946         rt->rt6i_src.plen = cfg->fc_src_len;
1947 #endif
1948
1949         rt->rt6i_metric = cfg->fc_metric;
1950
1951         /* We cannot add true routes via loopback here,
1952            they would result in kernel looping; promote them to reject routes
1953          */
1954         if ((cfg->fc_flags & RTF_REJECT) ||
1955             (dev && (dev->flags & IFF_LOOPBACK) &&
1956              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1957              !(cfg->fc_flags & RTF_LOCAL))) {
1958                 /* hold loopback dev/idev if we haven't done so. */
1959                 if (dev != net->loopback_dev) {
1960                         if (dev) {
1961                                 dev_put(dev);
1962                                 in6_dev_put(idev);
1963                         }
1964                         dev = net->loopback_dev;
1965                         dev_hold(dev);
1966                         idev = in6_dev_get(dev);
1967                         if (!idev) {
1968                                 err = -ENODEV;
1969                                 goto out;
1970                         }
1971                 }
1972                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1973                 switch (cfg->fc_type) {
1974                 case RTN_BLACKHOLE:
1975                         rt->dst.error = -EINVAL;
1976                         rt->dst.output = dst_discard_out;
1977                         rt->dst.input = dst_discard;
1978                         break;
1979                 case RTN_PROHIBIT:
1980                         rt->dst.error = -EACCES;
1981                         rt->dst.output = ip6_pkt_prohibit_out;
1982                         rt->dst.input = ip6_pkt_prohibit;
1983                         break;
1984                 case RTN_THROW:
1985                 case RTN_UNREACHABLE:
1986                 default:
1987                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1988                                         : (cfg->fc_type == RTN_UNREACHABLE)
1989                                         ? -EHOSTUNREACH : -ENETUNREACH;
1990                         rt->dst.output = ip6_pkt_discard_out;
1991                         rt->dst.input = ip6_pkt_discard;
1992                         break;
1993                 }
1994                 goto install_route;
1995         }
1996
1997         if (cfg->fc_flags & RTF_GATEWAY) {
1998                 const struct in6_addr *gw_addr;
1999                 int gwa_type;
2000
2001                 gw_addr = &cfg->fc_gateway;
2002                 gwa_type = ipv6_addr_type(gw_addr);
2003
2004                 /* if gw_addr is local we will fail to detect this in case
2005                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
2006                  * will return already-added prefix route via interface that
2007                  * prefix route was assigned to, which might be non-loopback.
2008                  */
2009                 err = -EINVAL;
2010                 if (ipv6_chk_addr_and_flags(net, gw_addr,
2011                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
2012                                             dev : NULL, 0, 0))
2013                         goto out;
2014
2015                 rt->rt6i_gateway = *gw_addr;
2016
2017                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
2018                         struct rt6_info *grt = NULL;
2019
2020                         /* IPv6 strictly inhibits using not link-local
2021                            addresses as nexthop address.
2022                            Otherwise, router will not able to send redirects.
2023                            It is very good, but in some (rare!) circumstances
2024                            (SIT, PtP, NBMA NOARP links) it is handy to allow
2025                            some exceptions. --ANK
2026                            We allow IPv4-mapped nexthops to support RFC4798-type
2027                            addressing
2028                          */
2029                         if (!(gwa_type & (IPV6_ADDR_UNICAST |
2030                                           IPV6_ADDR_MAPPED)))
2031                                 goto out;
2032
2033                         if (cfg->fc_table) {
2034                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2035
2036                                 if (grt) {
2037                                         if (grt->rt6i_flags & RTF_GATEWAY ||
2038                                             (dev && dev != grt->dst.dev)) {
2039                                                 ip6_rt_put(grt);
2040                                                 grt = NULL;
2041                                         }
2042                                 }
2043                         }
2044
2045                         if (!grt)
2046                                 grt = rt6_lookup(net, gw_addr, NULL,
2047                                                  cfg->fc_ifindex, 1);
2048
2049                         err = -EHOSTUNREACH;
2050                         if (!grt)
2051                                 goto out;
2052                         if (dev) {
2053                                 if (dev != grt->dst.dev) {
2054                                         ip6_rt_put(grt);
2055                                         goto out;
2056                                 }
2057                         } else {
2058                                 dev = grt->dst.dev;
2059                                 idev = grt->rt6i_idev;
2060                                 dev_hold(dev);
2061                                 in6_dev_hold(grt->rt6i_idev);
2062                         }
2063                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2064                                 err = 0;
2065                         ip6_rt_put(grt);
2066
2067                         if (err)
2068                                 goto out;
2069                 }
2070                 err = -EINVAL;
2071                 if (!dev || (dev->flags & IFF_LOOPBACK))
2072                         goto out;
2073         }
2074
2075         err = -ENODEV;
2076         if (!dev)
2077                 goto out;
2078
2079         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2080                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2081                         err = -EINVAL;
2082                         goto out;
2083                 }
2084                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2085                 rt->rt6i_prefsrc.plen = 128;
2086         } else
2087                 rt->rt6i_prefsrc.plen = 0;
2088
2089         rt->rt6i_flags = cfg->fc_flags;
2090
2091 install_route:
2092         rt->dst.dev = dev;
2093         rt->rt6i_idev = idev;
2094         rt->rt6i_table = table;
2095
2096         cfg->fc_nlinfo.nl_net = dev_net(dev);
2097
2098         return rt;
2099 out:
2100         if (dev)
2101                 dev_put(dev);
2102         if (idev)
2103                 in6_dev_put(idev);
2104         if (rt)
2105                 dst_free(&rt->dst);
2106
2107         return ERR_PTR(err);
2108 }
2109
2110 int ip6_route_add(struct fib6_config *cfg)
2111 {
2112         struct mx6_config mxc = { .mx = NULL, };
2113         struct rt6_info *rt;
2114         int err;
2115
2116         rt = ip6_route_info_create(cfg);
2117         if (IS_ERR(rt)) {
2118                 err = PTR_ERR(rt);
2119                 rt = NULL;
2120                 goto out;
2121         }
2122
2123         err = ip6_convert_metrics(&mxc, cfg);
2124         if (err)
2125                 goto out;
2126
2127         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2128
2129         kfree(mxc.mx);
2130
2131         return err;
2132 out:
2133         if (rt)
2134                 dst_free(&rt->dst);
2135
2136         return err;
2137 }
2138
2139 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2140 {
2141         int err;
2142         struct fib6_table *table;
2143         struct net *net = dev_net(rt->dst.dev);
2144
2145         if (rt == net->ipv6.ip6_null_entry ||
2146             rt->dst.flags & DST_NOCACHE) {
2147                 err = -ENOENT;
2148                 goto out;
2149         }
2150
2151         table = rt->rt6i_table;
2152         write_lock_bh(&table->tb6_lock);
2153         err = fib6_del(rt, info);
2154         write_unlock_bh(&table->tb6_lock);
2155
2156 out:
2157         ip6_rt_put(rt);
2158         return err;
2159 }
2160
2161 int ip6_del_rt(struct rt6_info *rt)
2162 {
2163         struct nl_info info = {
2164                 .nl_net = dev_net(rt->dst.dev),
2165         };
2166         return __ip6_del_rt(rt, &info);
2167 }
2168
2169 static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
2170 {
2171         struct nl_info *info = &cfg->fc_nlinfo;
2172         struct sk_buff *skb = NULL;
2173         struct fib6_table *table;
2174         int err;
2175
2176         table = rt->rt6i_table;
2177         write_lock_bh(&table->tb6_lock);
2178
2179         if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
2180                 struct rt6_info *sibling, *next_sibling;
2181
2182                 /* prefer to send a single notification with all hops */
2183                 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
2184                 if (skb) {
2185                         u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2186
2187                         if (rt6_fill_node(info->nl_net, skb, rt,
2188                                           NULL, NULL, 0, RTM_DELROUTE,
2189                                           info->portid, seq, 0) < 0) {
2190                                 kfree_skb(skb);
2191                                 skb = NULL;
2192                         } else
2193                                 info->skip_notify = 1;
2194                 }
2195
2196                 list_for_each_entry_safe(sibling, next_sibling,
2197                                          &rt->rt6i_siblings,
2198                                          rt6i_siblings) {
2199                         err = fib6_del(sibling, info);
2200                         if (err)
2201                                 goto out;
2202                 }
2203         }
2204
2205         err = fib6_del(rt, info);
2206 out:
2207         write_unlock_bh(&table->tb6_lock);
2208         ip6_rt_put(rt);
2209
2210         if (skb) {
2211                 rtnl_notify(skb, info->nl_net, info->portid, RTNLGRP_IPV6_ROUTE,
2212                             info->nlh, gfp_any());
2213         }
2214         return err;
2215 }
2216
2217 static int ip6_route_del(struct fib6_config *cfg)
2218 {
2219         struct fib6_table *table;
2220         struct fib6_node *fn;
2221         struct rt6_info *rt;
2222         int err = -ESRCH;
2223
2224         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2225         if (!table)
2226                 return err;
2227
2228         read_lock_bh(&table->tb6_lock);
2229
2230         fn = fib6_locate(&table->tb6_root,
2231                          &cfg->fc_dst, cfg->fc_dst_len,
2232                          &cfg->fc_src, cfg->fc_src_len);
2233
2234         if (fn) {
2235                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2236                         if ((rt->rt6i_flags & RTF_CACHE) &&
2237                             !(cfg->fc_flags & RTF_CACHE))
2238                                 continue;
2239                         if (cfg->fc_ifindex &&
2240                             (!rt->dst.dev ||
2241                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2242                                 continue;
2243                         if (cfg->fc_flags & RTF_GATEWAY &&
2244                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2245                                 continue;
2246                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2247                                 continue;
2248                         if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2249                                 continue;
2250                         dst_hold(&rt->dst);
2251                         read_unlock_bh(&table->tb6_lock);
2252
2253                         /* if gateway was specified only delete the one hop */
2254                         if (cfg->fc_flags & RTF_GATEWAY)
2255                                 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2256
2257                         return __ip6_del_rt_siblings(rt, cfg);
2258                 }
2259         }
2260         read_unlock_bh(&table->tb6_lock);
2261
2262         return err;
2263 }
2264
2265 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2266 {
2267         struct netevent_redirect netevent;
2268         struct rt6_info *rt, *nrt = NULL;
2269         struct ndisc_options ndopts;
2270         struct inet6_dev *in6_dev;
2271         struct neighbour *neigh;
2272         struct rd_msg *msg;
2273         int optlen, on_link;
2274         u8 *lladdr;
2275
2276         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2277         optlen -= sizeof(*msg);
2278
2279         if (optlen < 0) {
2280                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2281                 return;
2282         }
2283
2284         msg = (struct rd_msg *)icmp6_hdr(skb);
2285
2286         if (ipv6_addr_is_multicast(&msg->dest)) {
2287                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2288                 return;
2289         }
2290
2291         on_link = 0;
2292         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2293                 on_link = 1;
2294         } else if (ipv6_addr_type(&msg->target) !=
2295                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2296                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2297                 return;
2298         }
2299
2300         in6_dev = __in6_dev_get(skb->dev);
2301         if (!in6_dev)
2302                 return;
2303         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2304                 return;
2305
2306         /* RFC2461 8.1:
2307          *      The IP source address of the Redirect MUST be the same as the current
2308          *      first-hop router for the specified ICMP Destination Address.
2309          */
2310
2311         if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2312                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2313                 return;
2314         }
2315
2316         lladdr = NULL;
2317         if (ndopts.nd_opts_tgt_lladdr) {
2318                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2319                                              skb->dev);
2320                 if (!lladdr) {
2321                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2322                         return;
2323                 }
2324         }
2325
2326         rt = (struct rt6_info *) dst;
2327         if (rt->rt6i_flags & RTF_REJECT) {
2328                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2329                 return;
2330         }
2331
2332         /* Redirect received -> path was valid.
2333          * Look, redirects are sent only in response to data packets,
2334          * so that this nexthop apparently is reachable. --ANK
2335          */
2336         dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
2337
2338         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2339         if (!neigh)
2340                 return;
2341
2342         /*
2343          *      We have finally decided to accept it.
2344          */
2345
2346         ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2347                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2348                      NEIGH_UPDATE_F_OVERRIDE|
2349                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2350                                      NEIGH_UPDATE_F_ISROUTER)),
2351                      NDISC_REDIRECT, &ndopts);
2352
2353         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2354         if (!nrt)
2355                 goto out;
2356
2357         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2358         if (on_link)
2359                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2360
2361         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2362
2363         if (ip6_ins_rt(nrt))
2364                 goto out;
2365
2366         netevent.old = &rt->dst;
2367         netevent.new = &nrt->dst;
2368         netevent.daddr = &msg->dest;
2369         netevent.neigh = neigh;
2370         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2371
2372         if (rt->rt6i_flags & RTF_CACHE) {
2373                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2374                 ip6_del_rt(rt);
2375         }
2376
2377 out:
2378         neigh_release(neigh);
2379 }
2380
2381 /*
2382  *      Misc support functions
2383  */
2384
2385 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2386 {
2387         BUG_ON(from->dst.from);
2388
2389         rt->rt6i_flags &= ~RTF_EXPIRES;
2390         dst_hold(&from->dst);
2391         rt->dst.from = &from->dst;
2392         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2393 }
2394
2395 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2396 {
2397         rt->dst.input = ort->dst.input;
2398         rt->dst.output = ort->dst.output;
2399         rt->rt6i_dst = ort->rt6i_dst;
2400         rt->dst.error = ort->dst.error;
2401         rt->rt6i_idev = ort->rt6i_idev;
2402         if (rt->rt6i_idev)
2403                 in6_dev_hold(rt->rt6i_idev);
2404         rt->dst.lastuse = jiffies;
2405         rt->rt6i_gateway = ort->rt6i_gateway;
2406         rt->rt6i_flags = ort->rt6i_flags;
2407         rt6_set_from(rt, ort);
2408         rt->rt6i_metric = ort->rt6i_metric;
2409 #ifdef CONFIG_IPV6_SUBTREES
2410         rt->rt6i_src = ort->rt6i_src;
2411 #endif
2412         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2413         rt->rt6i_table = ort->rt6i_table;
2414         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2415 }
2416
2417 #ifdef CONFIG_IPV6_ROUTE_INFO
2418 static struct rt6_info *rt6_get_route_info(struct net *net,
2419                                            const struct in6_addr *prefix, int prefixlen,
2420                                            const struct in6_addr *gwaddr,
2421                                            struct net_device *dev)
2422 {
2423         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2424         int ifindex = dev->ifindex;
2425         struct fib6_node *fn;
2426         struct rt6_info *rt = NULL;
2427         struct fib6_table *table;
2428
2429         table = fib6_get_table(net, tb_id);
2430         if (!table)
2431                 return NULL;
2432
2433         read_lock_bh(&table->tb6_lock);
2434         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2435         if (!fn)
2436                 goto out;
2437
2438         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2439                 if (rt->dst.dev->ifindex != ifindex)
2440                         continue;
2441                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2442                         continue;
2443                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2444                         continue;
2445                 dst_hold(&rt->dst);
2446                 break;
2447         }
2448 out:
2449         read_unlock_bh(&table->tb6_lock);
2450         return rt;
2451 }
2452
2453 static struct rt6_info *rt6_add_route_info(struct net *net,
2454                                            const struct in6_addr *prefix, int prefixlen,
2455                                            const struct in6_addr *gwaddr,
2456                                            struct net_device *dev,
2457                                            unsigned int pref)
2458 {
2459         struct fib6_config cfg = {
2460                 .fc_metric      = IP6_RT_PRIO_USER,
2461                 .fc_ifindex     = dev->ifindex,
2462                 .fc_dst_len     = prefixlen,
2463                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2464                                   RTF_UP | RTF_PREF(pref),
2465                 .fc_nlinfo.portid = 0,
2466                 .fc_nlinfo.nlh = NULL,
2467                 .fc_nlinfo.nl_net = net,
2468         };
2469
2470         cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2471         cfg.fc_dst = *prefix;
2472         cfg.fc_gateway = *gwaddr;
2473
2474         /* We should treat it as a default route if prefix length is 0. */
2475         if (!prefixlen)
2476                 cfg.fc_flags |= RTF_DEFAULT;
2477
2478         ip6_route_add(&cfg);
2479
2480         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2481 }
2482 #endif
2483
2484 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2485 {
2486         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2487         struct rt6_info *rt;
2488         struct fib6_table *table;
2489
2490         table = fib6_get_table(dev_net(dev), tb_id);
2491         if (!table)
2492                 return NULL;
2493
2494         read_lock_bh(&table->tb6_lock);
2495         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2496                 if (dev == rt->dst.dev &&
2497                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2498                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2499                         break;
2500         }
2501         if (rt)
2502                 dst_hold(&rt->dst);
2503         read_unlock_bh(&table->tb6_lock);
2504         return rt;
2505 }
2506
2507 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2508                                      struct net_device *dev,
2509                                      unsigned int pref)
2510 {
2511         struct fib6_config cfg = {
2512                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2513                 .fc_metric      = IP6_RT_PRIO_USER,
2514                 .fc_ifindex     = dev->ifindex,
2515                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2516                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2517                 .fc_nlinfo.portid = 0,
2518                 .fc_nlinfo.nlh = NULL,
2519                 .fc_nlinfo.nl_net = dev_net(dev),
2520         };
2521
2522         cfg.fc_gateway = *gwaddr;
2523
2524         if (!ip6_route_add(&cfg)) {
2525                 struct fib6_table *table;
2526
2527                 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2528                 if (table)
2529                         table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2530         }
2531
2532         return rt6_get_dflt_router(gwaddr, dev);
2533 }
2534
2535 static void __rt6_purge_dflt_routers(struct fib6_table *table)
2536 {
2537         struct rt6_info *rt;
2538
2539 restart:
2540         read_lock_bh(&table->tb6_lock);
2541         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2542                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2543                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2544                         dst_hold(&rt->dst);
2545                         read_unlock_bh(&table->tb6_lock);
2546                         ip6_del_rt(rt);
2547                         goto restart;
2548                 }
2549         }
2550         read_unlock_bh(&table->tb6_lock);
2551
2552         table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2553 }
2554
2555 void rt6_purge_dflt_routers(struct net *net)
2556 {
2557         struct fib6_table *table;
2558         struct hlist_head *head;
2559         unsigned int h;
2560
2561         rcu_read_lock();
2562
2563         for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2564                 head = &net->ipv6.fib_table_hash[h];
2565                 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2566                         if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2567                                 __rt6_purge_dflt_routers(table);
2568                 }
2569         }
2570
2571         rcu_read_unlock();
2572 }
2573
2574 static void rtmsg_to_fib6_config(struct net *net,
2575                                  struct in6_rtmsg *rtmsg,
2576                                  struct fib6_config *cfg)
2577 {
2578         memset(cfg, 0, sizeof(*cfg));
2579
2580         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2581                          : RT6_TABLE_MAIN;
2582         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2583         cfg->fc_metric = rtmsg->rtmsg_metric;
2584         cfg->fc_expires = rtmsg->rtmsg_info;
2585         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2586         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2587         cfg->fc_flags = rtmsg->rtmsg_flags;
2588
2589         cfg->fc_nlinfo.nl_net = net;
2590
2591         cfg->fc_dst = rtmsg->rtmsg_dst;
2592         cfg->fc_src = rtmsg->rtmsg_src;
2593         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2594 }
2595
2596 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2597 {
2598         struct fib6_config cfg;
2599         struct in6_rtmsg rtmsg;
2600         int err;
2601
2602         switch (cmd) {
2603         case SIOCADDRT:         /* Add a route */
2604         case SIOCDELRT:         /* Delete a route */
2605                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2606                         return -EPERM;
2607                 err = copy_from_user(&rtmsg, arg,
2608                                      sizeof(struct in6_rtmsg));
2609                 if (err)
2610                         return -EFAULT;
2611
2612                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2613
2614                 rtnl_lock();
2615                 switch (cmd) {
2616                 case SIOCADDRT:
2617                         err = ip6_route_add(&cfg);
2618                         break;
2619                 case SIOCDELRT:
2620                         err = ip6_route_del(&cfg);
2621                         break;
2622                 default:
2623                         err = -EINVAL;
2624                 }
2625                 rtnl_unlock();
2626
2627                 return err;
2628         }
2629
2630         return -EINVAL;
2631 }
2632
2633 /*
2634  *      Drop the packet on the floor
2635  */
2636
2637 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2638 {
2639         int type;
2640         struct dst_entry *dst = skb_dst(skb);
2641         switch (ipstats_mib_noroutes) {
2642         case IPSTATS_MIB_INNOROUTES:
2643                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2644                 if (type == IPV6_ADDR_ANY) {
2645                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2646                                       IPSTATS_MIB_INADDRERRORS);
2647                         break;
2648                 }
2649                 /* FALLTHROUGH */
2650         case IPSTATS_MIB_OUTNOROUTES:
2651                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2652                               ipstats_mib_noroutes);
2653                 break;
2654         }
2655         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2656         kfree_skb(skb);
2657         return 0;
2658 }
2659
2660 static int ip6_pkt_discard(struct sk_buff *skb)
2661 {
2662         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2663 }
2664
2665 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2666 {
2667         skb->dev = skb_dst(skb)->dev;
2668         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2669 }
2670
2671 static int ip6_pkt_prohibit(struct sk_buff *skb)
2672 {
2673         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2674 }
2675
2676 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2677 {
2678         skb->dev = skb_dst(skb)->dev;
2679         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2680 }
2681
2682 /*
2683  *      Allocate a dst for local (unicast / anycast) address.
2684  */
2685
2686 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2687                                     const struct in6_addr *addr,
2688                                     bool anycast)
2689 {
2690         u32 tb_id;
2691         struct net *net = dev_net(idev->dev);
2692         struct net_device *dev = net->loopback_dev;
2693         struct rt6_info *rt;
2694
2695         /* use L3 Master device as loopback for host routes if device
2696          * is enslaved and address is not link local or multicast
2697          */
2698         if (!rt6_need_strict(addr))
2699                 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2700
2701         rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2702         if (!rt)
2703                 return ERR_PTR(-ENOMEM);
2704
2705         in6_dev_hold(idev);
2706
2707         rt->dst.flags |= DST_HOST;
2708         rt->dst.input = ip6_input;
2709         rt->dst.output = ip6_output;
2710         rt->rt6i_idev = idev;
2711
2712         rt->rt6i_protocol = RTPROT_KERNEL;
2713         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2714         if (anycast)
2715                 rt->rt6i_flags |= RTF_ANYCAST;
2716         else
2717                 rt->rt6i_flags |= RTF_LOCAL;
2718
2719         rt->rt6i_gateway  = *addr;
2720         rt->rt6i_dst.addr = *addr;
2721         rt->rt6i_dst.plen = 128;
2722         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2723         rt->rt6i_table = fib6_get_table(net, tb_id);
2724         rt->dst.flags |= DST_NOCACHE;
2725
2726         atomic_set(&rt->dst.__refcnt, 1);
2727
2728         return rt;
2729 }
2730
2731 /* remove deleted ip from prefsrc entries */
2732 struct arg_dev_net_ip {
2733         struct net_device *dev;
2734         struct net *net;
2735         struct in6_addr *addr;
2736 };
2737
2738 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2739 {
2740         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2741         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2742         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2743
2744         if (((void *)rt->dst.dev == dev || !dev) &&
2745             rt != net->ipv6.ip6_null_entry &&
2746             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2747                 /* remove prefsrc entry */
2748                 rt->rt6i_prefsrc.plen = 0;
2749         }
2750         return 0;
2751 }
2752
2753 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2754 {
2755         struct net *net = dev_net(ifp->idev->dev);
2756         struct arg_dev_net_ip adni = {
2757                 .dev = ifp->idev->dev,
2758                 .net = net,
2759                 .addr = &ifp->addr,
2760         };
2761         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2762 }
2763
2764 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2765 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2766
2767 /* Remove routers and update dst entries when gateway turn into host. */
2768 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2769 {
2770         struct in6_addr *gateway = (struct in6_addr *)arg;
2771
2772         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2773              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2774              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2775                 return -1;
2776         }
2777         return 0;
2778 }
2779
2780 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2781 {
2782         fib6_clean_all(net, fib6_clean_tohost, gateway);
2783 }
2784
2785 struct arg_dev_net {
2786         struct net_device *dev;
2787         struct net *net;
2788 };
2789
2790 /* called with write lock held for table with rt */
2791 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2792 {
2793         const struct arg_dev_net *adn = arg;
2794         const struct net_device *dev = adn->dev;
2795
2796         if ((rt->dst.dev == dev || !dev) &&
2797             rt != adn->net->ipv6.ip6_null_entry &&
2798             (rt->rt6i_nsiblings == 0 ||
2799              !rt->rt6i_idev->cnf.ignore_routes_with_linkdown))
2800                 return -1;
2801
2802         return 0;
2803 }
2804
2805 void rt6_ifdown(struct net *net, struct net_device *dev)
2806 {
2807         struct arg_dev_net adn = {
2808                 .dev = dev,
2809                 .net = net,
2810         };
2811
2812         fib6_clean_all(net, fib6_ifdown, &adn);
2813         icmp6_clean_all(fib6_ifdown, &adn);
2814         if (dev)
2815                 rt6_uncached_list_flush_dev(net, dev);
2816 }
2817
2818 struct rt6_mtu_change_arg {
2819         struct net_device *dev;
2820         unsigned int mtu;
2821 };
2822
2823 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2824 {
2825         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2826         struct inet6_dev *idev;
2827
2828         /* In IPv6 pmtu discovery is not optional,
2829            so that RTAX_MTU lock cannot disable it.
2830            We still use this lock to block changes
2831            caused by addrconf/ndisc.
2832         */
2833
2834         idev = __in6_dev_get(arg->dev);
2835         if (!idev)
2836                 return 0;
2837
2838         /* For administrative MTU increase, there is no way to discover
2839            IPv6 PMTU increase, so PMTU increase should be updated here.
2840            Since RFC 1981 doesn't include administrative MTU increase
2841            update PMTU increase is a MUST. (i.e. jumbo frame)
2842          */
2843         /*
2844            If new MTU is less than route PMTU, this new MTU will be the
2845            lowest MTU in the path, update the route PMTU to reflect PMTU
2846            decreases; if new MTU is greater than route PMTU, and the
2847            old MTU is the lowest MTU in the path, update the route PMTU
2848            to reflect the increase. In this case if the other nodes' MTU
2849            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2850            PMTU discovery.
2851          */
2852         if (rt->dst.dev == arg->dev &&
2853             dst_metric_raw(&rt->dst, RTAX_MTU) &&
2854             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2855                 if (rt->rt6i_flags & RTF_CACHE) {
2856                         /* For RTF_CACHE with rt6i_pmtu == 0
2857                          * (i.e. a redirected route),
2858                          * the metrics of its rt->dst.from has already
2859                          * been updated.
2860                          */
2861                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2862                                 rt->rt6i_pmtu = arg->mtu;
2863                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2864                            (dst_mtu(&rt->dst) < arg->mtu &&
2865                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2866                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2867                 }
2868         }
2869         return 0;
2870 }
2871
2872 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2873 {
2874         struct rt6_mtu_change_arg arg = {
2875                 .dev = dev,
2876                 .mtu = mtu,
2877         };
2878
2879         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2880 }
2881
2882 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2883         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2884         [RTA_OIF]               = { .type = NLA_U32 },
2885         [RTA_IIF]               = { .type = NLA_U32 },
2886         [RTA_PRIORITY]          = { .type = NLA_U32 },
2887         [RTA_METRICS]           = { .type = NLA_NESTED },
2888         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2889         [RTA_PREF]              = { .type = NLA_U8 },
2890         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2891         [RTA_ENCAP]             = { .type = NLA_NESTED },
2892         [RTA_EXPIRES]           = { .type = NLA_U32 },
2893         [RTA_UID]               = { .type = NLA_U32 },
2894 };
2895
2896 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2897                               struct fib6_config *cfg)
2898 {
2899         struct rtmsg *rtm;
2900         struct nlattr *tb[RTA_MAX+1];
2901         unsigned int pref;
2902         int err;
2903
2904         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
2905         if (err < 0)
2906                 goto errout;
2907
2908         err = -EINVAL;
2909         rtm = nlmsg_data(nlh);
2910         memset(cfg, 0, sizeof(*cfg));
2911
2912         cfg->fc_table = rtm->rtm_table;
2913         cfg->fc_dst_len = rtm->rtm_dst_len;
2914         cfg->fc_src_len = rtm->rtm_src_len;
2915         cfg->fc_flags = RTF_UP;
2916         cfg->fc_protocol = rtm->rtm_protocol;
2917         cfg->fc_type = rtm->rtm_type;
2918
2919         if (rtm->rtm_type == RTN_UNREACHABLE ||
2920             rtm->rtm_type == RTN_BLACKHOLE ||
2921             rtm->rtm_type == RTN_PROHIBIT ||
2922             rtm->rtm_type == RTN_THROW)
2923                 cfg->fc_flags |= RTF_REJECT;
2924
2925         if (rtm->rtm_type == RTN_LOCAL)
2926                 cfg->fc_flags |= RTF_LOCAL;
2927
2928         if (rtm->rtm_flags & RTM_F_CLONED)
2929                 cfg->fc_flags |= RTF_CACHE;
2930
2931         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2932         cfg->fc_nlinfo.nlh = nlh;
2933         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2934
2935         if (tb[RTA_GATEWAY]) {
2936                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2937                 cfg->fc_flags |= RTF_GATEWAY;
2938         }
2939
2940         if (tb[RTA_DST]) {
2941                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2942
2943                 if (nla_len(tb[RTA_DST]) < plen)
2944                         goto errout;
2945
2946                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2947         }
2948
2949         if (tb[RTA_SRC]) {
2950                 int plen = (rtm->rtm_src_len + 7) >> 3;
2951
2952                 if (nla_len(tb[RTA_SRC]) < plen)
2953                         goto errout;
2954
2955                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2956         }
2957
2958         if (tb[RTA_PREFSRC])
2959                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2960
2961         if (tb[RTA_OIF])
2962                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2963
2964         if (tb[RTA_PRIORITY])
2965                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2966
2967         if (tb[RTA_METRICS]) {
2968                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2969                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2970         }
2971
2972         if (tb[RTA_TABLE])
2973                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2974
2975         if (tb[RTA_MULTIPATH]) {
2976                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2977                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2978
2979                 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
2980                                                      cfg->fc_mp_len);
2981                 if (err < 0)
2982                         goto errout;
2983         }
2984
2985         if (tb[RTA_PREF]) {
2986                 pref = nla_get_u8(tb[RTA_PREF]);
2987                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2988                     pref != ICMPV6_ROUTER_PREF_HIGH)
2989                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2990                 cfg->fc_flags |= RTF_PREF(pref);
2991         }
2992
2993         if (tb[RTA_ENCAP])
2994                 cfg->fc_encap = tb[RTA_ENCAP];
2995
2996         if (tb[RTA_ENCAP_TYPE]) {
2997                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2998
2999                 err = lwtunnel_valid_encap_type(cfg->fc_encap_type);
3000                 if (err < 0)
3001                         goto errout;
3002         }
3003
3004         if (tb[RTA_EXPIRES]) {
3005                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
3006
3007                 if (addrconf_finite_timeout(timeout)) {
3008                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
3009                         cfg->fc_flags |= RTF_EXPIRES;
3010                 }
3011         }
3012
3013         err = 0;
3014 errout:
3015         return err;
3016 }
3017
3018 struct rt6_nh {
3019         struct rt6_info *rt6_info;
3020         struct fib6_config r_cfg;
3021         struct mx6_config mxc;
3022         struct list_head next;
3023 };
3024
3025 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
3026 {
3027         struct rt6_nh *nh;
3028
3029         list_for_each_entry(nh, rt6_nh_list, next) {
3030                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
3031                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
3032                         nh->r_cfg.fc_ifindex);
3033         }
3034 }
3035
3036 static int ip6_route_info_append(struct list_head *rt6_nh_list,
3037                                  struct rt6_info *rt, struct fib6_config *r_cfg)
3038 {
3039         struct rt6_nh *nh;
3040         struct rt6_info *rtnh;
3041         int err = -EEXIST;
3042
3043         list_for_each_entry(nh, rt6_nh_list, next) {
3044                 /* check if rt6_info already exists */
3045                 rtnh = nh->rt6_info;
3046
3047                 if (rtnh->dst.dev == rt->dst.dev &&
3048                     rtnh->rt6i_idev == rt->rt6i_idev &&
3049                     ipv6_addr_equal(&rtnh->rt6i_gateway,
3050                                     &rt->rt6i_gateway))
3051                         return err;
3052         }
3053
3054         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
3055         if (!nh)
3056                 return -ENOMEM;
3057         nh->rt6_info = rt;
3058         err = ip6_convert_metrics(&nh->mxc, r_cfg);
3059         if (err) {
3060                 kfree(nh);
3061                 return err;
3062         }
3063         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
3064         list_add_tail(&nh->next, rt6_nh_list);
3065
3066         return 0;
3067 }
3068
3069 static void ip6_route_mpath_notify(struct rt6_info *rt,
3070                                    struct rt6_info *rt_last,
3071                                    struct nl_info *info,
3072                                    __u16 nlflags)
3073 {
3074         /* if this is an APPEND route, then rt points to the first route
3075          * inserted and rt_last points to last route inserted. Userspace
3076          * wants a consistent dump of the route which starts at the first
3077          * nexthop. Since sibling routes are always added at the end of
3078          * the list, find the first sibling of the last route appended
3079          */
3080         if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->rt6i_nsiblings) {
3081                 rt = list_first_entry(&rt_last->rt6i_siblings,
3082                                       struct rt6_info,
3083                                       rt6i_siblings);
3084         }
3085
3086         if (rt)
3087                 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
3088 }
3089
3090 static int ip6_route_multipath_add(struct fib6_config *cfg)
3091 {
3092         struct rt6_info *rt_notif = NULL, *rt_last = NULL;
3093         struct nl_info *info = &cfg->fc_nlinfo;
3094         struct fib6_config r_cfg;
3095         struct rtnexthop *rtnh;
3096         struct rt6_info *rt;
3097         struct rt6_nh *err_nh;
3098         struct rt6_nh *nh, *nh_safe;
3099         __u16 nlflags;
3100         int remaining;
3101         int attrlen;
3102         int err = 1;
3103         int nhn = 0;
3104         int replace = (cfg->fc_nlinfo.nlh &&
3105                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
3106         LIST_HEAD(rt6_nh_list);
3107
3108         nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
3109         if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
3110                 nlflags |= NLM_F_APPEND;
3111
3112         remaining = cfg->fc_mp_len;
3113         rtnh = (struct rtnexthop *)cfg->fc_mp;
3114
3115         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3116          * rt6_info structs per nexthop
3117          */
3118         while (rtnh_ok(rtnh, remaining)) {
3119                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3120                 if (rtnh->rtnh_ifindex)
3121                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3122
3123                 attrlen = rtnh_attrlen(rtnh);
3124                 if (attrlen > 0) {
3125                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3126
3127                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3128                         if (nla) {
3129                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
3130                                 r_cfg.fc_flags |= RTF_GATEWAY;
3131                         }
3132                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3133                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3134                         if (nla)
3135                                 r_cfg.fc_encap_type = nla_get_u16(nla);
3136                 }
3137
3138                 rt = ip6_route_info_create(&r_cfg);
3139                 if (IS_ERR(rt)) {
3140                         err = PTR_ERR(rt);
3141                         rt = NULL;
3142                         goto cleanup;
3143                 }
3144
3145                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
3146                 if (err) {
3147                         dst_free(&rt->dst);
3148                         goto cleanup;
3149                 }
3150
3151                 rtnh = rtnh_next(rtnh, &remaining);
3152         }
3153
3154         /* for add and replace send one notification with all nexthops.
3155          * Skip the notification in fib6_add_rt2node and send one with
3156          * the full route when done
3157          */
3158         info->skip_notify = 1;
3159
3160         err_nh = NULL;
3161         list_for_each_entry(nh, &rt6_nh_list, next) {
3162                 rt_last = nh->rt6_info;
3163                 err = __ip6_ins_rt(nh->rt6_info, info, &nh->mxc);
3164                 /* save reference to first route for notification */
3165                 if (!rt_notif && !err)
3166                         rt_notif = nh->rt6_info;
3167
3168                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3169                 nh->rt6_info = NULL;
3170                 if (err) {
3171                         if (replace && nhn)
3172                                 ip6_print_replace_route_err(&rt6_nh_list);
3173                         err_nh = nh;
3174                         goto add_errout;
3175                 }
3176
3177                 /* Because each route is added like a single route we remove
3178                  * these flags after the first nexthop: if there is a collision,
3179                  * we have already failed to add the first nexthop:
3180                  * fib6_add_rt2node() has rejected it; when replacing, old
3181                  * nexthops have been replaced by first new, the rest should
3182                  * be added to it.
3183                  */
3184                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3185                                                      NLM_F_REPLACE);
3186                 nhn++;
3187         }
3188
3189         /* success ... tell user about new route */
3190         ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3191         goto cleanup;
3192
3193 add_errout:
3194         /* send notification for routes that were added so that
3195          * the delete notifications sent by ip6_route_del are
3196          * coherent
3197          */
3198         if (rt_notif)
3199                 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3200
3201         /* Delete routes that were already added */
3202         list_for_each_entry(nh, &rt6_nh_list, next) {
3203                 if (err_nh == nh)
3204                         break;
3205                 ip6_route_del(&nh->r_cfg);
3206         }
3207
3208 cleanup:
3209         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3210                 if (nh->rt6_info)
3211                         dst_free(&nh->rt6_info->dst);
3212                 kfree(nh->mxc.mx);
3213                 list_del(&nh->next);
3214                 kfree(nh);
3215         }
3216
3217         return err;
3218 }
3219
3220 static int ip6_route_multipath_del(struct fib6_config *cfg)
3221 {
3222         struct fib6_config r_cfg;
3223         struct rtnexthop *rtnh;
3224         int remaining;
3225         int attrlen;
3226         int err = 1, last_err = 0;
3227
3228         remaining = cfg->fc_mp_len;
3229         rtnh = (struct rtnexthop *)cfg->fc_mp;
3230
3231         /* Parse a Multipath Entry */
3232         while (rtnh_ok(rtnh, remaining)) {
3233                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3234                 if (rtnh->rtnh_ifindex)
3235                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3236
3237                 attrlen = rtnh_attrlen(rtnh);
3238                 if (attrlen > 0) {
3239                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3240
3241                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3242                         if (nla) {
3243                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3244                                 r_cfg.fc_flags |= RTF_GATEWAY;
3245                         }
3246                 }
3247                 err = ip6_route_del(&r_cfg);
3248                 if (err)
3249                         last_err = err;
3250
3251                 rtnh = rtnh_next(rtnh, &remaining);
3252         }
3253
3254         return last_err;
3255 }
3256
3257 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3258 {
3259         struct fib6_config cfg;
3260         int err;
3261
3262         err = rtm_to_fib6_config(skb, nlh, &cfg);
3263         if (err < 0)
3264                 return err;
3265
3266         if (cfg.fc_mp)
3267                 return ip6_route_multipath_del(&cfg);
3268         else {
3269                 cfg.fc_delete_all_nh = 1;
3270                 return ip6_route_del(&cfg);
3271         }
3272 }
3273
3274 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh)
3275 {
3276         struct fib6_config cfg;
3277         int err;
3278
3279         err = rtm_to_fib6_config(skb, nlh, &cfg);
3280         if (err < 0)
3281                 return err;
3282
3283         if (cfg.fc_mp)
3284                 return ip6_route_multipath_add(&cfg);
3285         else
3286                 return ip6_route_add(&cfg);
3287 }
3288
3289 static size_t rt6_nlmsg_size(struct rt6_info *rt)
3290 {
3291         int nexthop_len = 0;
3292
3293         if (rt->rt6i_nsiblings) {
3294                 nexthop_len = nla_total_size(0)  /* RTA_MULTIPATH */
3295                             + NLA_ALIGN(sizeof(struct rtnexthop))
3296                             + nla_total_size(16) /* RTA_GATEWAY */
3297                             + nla_total_size(4)  /* RTA_OIF */
3298                             + lwtunnel_get_encap_size(rt->dst.lwtstate);
3299
3300                 nexthop_len *= rt->rt6i_nsiblings;
3301         }
3302
3303         return NLMSG_ALIGN(sizeof(struct rtmsg))
3304                + nla_total_size(16) /* RTA_SRC */
3305                + nla_total_size(16) /* RTA_DST */
3306                + nla_total_size(16) /* RTA_GATEWAY */
3307                + nla_total_size(16) /* RTA_PREFSRC */
3308                + nla_total_size(4) /* RTA_TABLE */
3309                + nla_total_size(4) /* RTA_IIF */
3310                + nla_total_size(4) /* RTA_OIF */
3311                + nla_total_size(4) /* RTA_PRIORITY */
3312                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3313                + nla_total_size(sizeof(struct rta_cacheinfo))
3314                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3315                + nla_total_size(1) /* RTA_PREF */
3316                + lwtunnel_get_encap_size(rt->dst.lwtstate)
3317                + nexthop_len;
3318 }
3319
3320 static int rt6_nexthop_info(struct sk_buff *skb, struct rt6_info *rt,
3321                             unsigned int *flags)
3322 {
3323         if (!netif_running(rt->dst.dev) || !netif_carrier_ok(rt->dst.dev)) {
3324                 *flags |= RTNH_F_LINKDOWN;
3325                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3326                         *flags |= RTNH_F_DEAD;
3327         }
3328
3329         if (rt->rt6i_flags & RTF_GATEWAY) {
3330                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3331                         goto nla_put_failure;
3332         }
3333
3334         if (rt->dst.dev &&
3335             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3336                 goto nla_put_failure;
3337
3338         if (rt->dst.lwtstate &&
3339             lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3340                 goto nla_put_failure;
3341
3342         return 0;
3343
3344 nla_put_failure:
3345         return -EMSGSIZE;
3346 }
3347
3348 static int rt6_add_nexthop(struct sk_buff *skb, struct rt6_info *rt)
3349 {
3350         struct rtnexthop *rtnh;
3351         unsigned int flags = 0;
3352
3353         rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
3354         if (!rtnh)
3355                 goto nla_put_failure;
3356
3357         rtnh->rtnh_hops = 0;
3358         rtnh->rtnh_ifindex = rt->dst.dev ? rt->dst.dev->ifindex : 0;
3359
3360         if (rt6_nexthop_info(skb, rt, &flags) < 0)
3361                 goto nla_put_failure;
3362
3363         rtnh->rtnh_flags = flags;
3364
3365         /* length of rtnetlink header + attributes */
3366         rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
3367
3368         return 0;
3369
3370 nla_put_failure:
3371         return -EMSGSIZE;
3372 }
3373
3374 static int rt6_fill_node(struct net *net,
3375                          struct sk_buff *skb, struct rt6_info *rt,
3376                          struct in6_addr *dst, struct in6_addr *src,
3377                          int iif, int type, u32 portid, u32 seq,
3378                          unsigned int flags)
3379 {
3380         u32 metrics[RTAX_MAX];
3381         struct rtmsg *rtm;
3382         struct nlmsghdr *nlh;
3383         long expires;
3384         u32 table;
3385
3386         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3387         if (!nlh)
3388                 return -EMSGSIZE;
3389
3390         rtm = nlmsg_data(nlh);
3391         rtm->rtm_family = AF_INET6;
3392         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3393         rtm->rtm_src_len = rt->rt6i_src.plen;
3394         rtm->rtm_tos = 0;
3395         if (rt->rt6i_table)
3396                 table = rt->rt6i_table->tb6_id;
3397         else
3398                 table = RT6_TABLE_UNSPEC;
3399         rtm->rtm_table = table;
3400         if (nla_put_u32(skb, RTA_TABLE, table))
3401                 goto nla_put_failure;
3402         if (rt->rt6i_flags & RTF_REJECT) {
3403                 switch (rt->dst.error) {
3404                 case -EINVAL:
3405                         rtm->rtm_type = RTN_BLACKHOLE;
3406                         break;
3407                 case -EACCES:
3408                         rtm->rtm_type = RTN_PROHIBIT;
3409                         break;
3410                 case -EAGAIN:
3411                         rtm->rtm_type = RTN_THROW;
3412                         break;
3413                 default:
3414                         rtm->rtm_type = RTN_UNREACHABLE;
3415                         break;
3416                 }
3417         }
3418         else if (rt->rt6i_flags & RTF_LOCAL)
3419                 rtm->rtm_type = RTN_LOCAL;
3420         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3421                 rtm->rtm_type = RTN_LOCAL;
3422         else
3423                 rtm->rtm_type = RTN_UNICAST;
3424         rtm->rtm_flags = 0;
3425         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3426         rtm->rtm_protocol = rt->rt6i_protocol;
3427         if (rt->rt6i_flags & RTF_DYNAMIC)
3428                 rtm->rtm_protocol = RTPROT_REDIRECT;
3429         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3430                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3431                         rtm->rtm_protocol = RTPROT_RA;
3432                 else
3433                         rtm->rtm_protocol = RTPROT_KERNEL;
3434         }
3435
3436         if (rt->rt6i_flags & RTF_CACHE)
3437                 rtm->rtm_flags |= RTM_F_CLONED;
3438
3439         if (dst) {
3440                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3441                         goto nla_put_failure;
3442                 rtm->rtm_dst_len = 128;
3443         } else if (rtm->rtm_dst_len)
3444                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3445                         goto nla_put_failure;
3446 #ifdef CONFIG_IPV6_SUBTREES
3447         if (src) {
3448                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3449                         goto nla_put_failure;
3450                 rtm->rtm_src_len = 128;
3451         } else if (rtm->rtm_src_len &&
3452                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3453                 goto nla_put_failure;
3454 #endif
3455         if (iif) {
3456 #ifdef CONFIG_IPV6_MROUTE
3457                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3458                         int err = ip6mr_get_route(net, skb, rtm, portid);
3459
3460                         if (err == 0)
3461                                 return 0;
3462                         if (err < 0)
3463                                 goto nla_put_failure;
3464                 } else
3465 #endif
3466                         if (nla_put_u32(skb, RTA_IIF, iif))
3467                                 goto nla_put_failure;
3468         } else if (dst) {
3469                 struct in6_addr saddr_buf;
3470                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3471                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3472                         goto nla_put_failure;
3473         }
3474
3475         if (rt->rt6i_prefsrc.plen) {
3476                 struct in6_addr saddr_buf;
3477                 saddr_buf = rt->rt6i_prefsrc.addr;
3478                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3479                         goto nla_put_failure;
3480         }
3481
3482         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3483         if (rt->rt6i_pmtu)
3484                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3485         if (rtnetlink_put_metrics(skb, metrics) < 0)
3486                 goto nla_put_failure;
3487
3488         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3489                 goto nla_put_failure;
3490
3491         /* For multipath routes, walk the siblings list and add
3492          * each as a nexthop within RTA_MULTIPATH.
3493          */
3494         if (rt->rt6i_nsiblings) {
3495                 struct rt6_info *sibling, *next_sibling;
3496                 struct nlattr *mp;
3497
3498                 mp = nla_nest_start(skb, RTA_MULTIPATH);
3499                 if (!mp)
3500                         goto nla_put_failure;
3501
3502                 if (rt6_add_nexthop(skb, rt) < 0)
3503                         goto nla_put_failure;
3504
3505                 list_for_each_entry_safe(sibling, next_sibling,
3506                                          &rt->rt6i_siblings, rt6i_siblings) {
3507                         if (rt6_add_nexthop(skb, sibling) < 0)
3508                                 goto nla_put_failure;
3509                 }
3510
3511                 nla_nest_end(skb, mp);
3512         } else {
3513                 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags) < 0)
3514                         goto nla_put_failure;
3515         }
3516
3517         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3518
3519         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3520                 goto nla_put_failure;
3521
3522         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3523                 goto nla_put_failure;
3524
3525
3526         nlmsg_end(skb, nlh);
3527         return 0;
3528
3529 nla_put_failure:
3530         nlmsg_cancel(skb, nlh);
3531         return -EMSGSIZE;
3532 }
3533
3534 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3535 {
3536         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3537         struct net *net = arg->net;
3538
3539         if (rt == net->ipv6.ip6_null_entry)
3540                 return 0;
3541
3542         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3543                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3544
3545                 /* user wants prefix routes only */
3546                 if (rtm->rtm_flags & RTM_F_PREFIX &&
3547                     !(rt->rt6i_flags & RTF_PREFIX_RT)) {
3548                         /* success since this is not a prefix route */
3549                         return 1;
3550                 }
3551         }
3552
3553         return rt6_fill_node(net,
3554                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3555                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3556                      NLM_F_MULTI);
3557 }
3558
3559 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh)
3560 {
3561         struct net *net = sock_net(in_skb->sk);
3562         struct nlattr *tb[RTA_MAX+1];
3563         struct rt6_info *rt;
3564         struct sk_buff *skb;
3565         struct rtmsg *rtm;
3566         struct flowi6 fl6;
3567         int err, iif = 0, oif = 0;
3568
3569         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy);
3570         if (err < 0)
3571                 goto errout;
3572
3573         err = -EINVAL;
3574         memset(&fl6, 0, sizeof(fl6));
3575         rtm = nlmsg_data(nlh);
3576         fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3577
3578         if (tb[RTA_SRC]) {
3579                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3580                         goto errout;
3581
3582                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3583         }
3584
3585         if (tb[RTA_DST]) {
3586                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3587                         goto errout;
3588
3589                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3590         }
3591
3592         if (tb[RTA_IIF])
3593                 iif = nla_get_u32(tb[RTA_IIF]);
3594
3595         if (tb[RTA_OIF])
3596                 oif = nla_get_u32(tb[RTA_OIF]);
3597
3598         if (tb[RTA_MARK])
3599                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3600
3601         if (tb[RTA_UID])
3602                 fl6.flowi6_uid = make_kuid(current_user_ns(),
3603                                            nla_get_u32(tb[RTA_UID]));
3604         else
3605                 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3606
3607         if (iif) {
3608                 struct net_device *dev;
3609                 int flags = 0;
3610
3611                 dev = __dev_get_by_index(net, iif);
3612                 if (!dev) {
3613                         err = -ENODEV;
3614                         goto errout;
3615                 }
3616
3617                 fl6.flowi6_iif = iif;
3618
3619                 if (!ipv6_addr_any(&fl6.saddr))
3620                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3621
3622                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3623                                                                flags);
3624         } else {
3625                 fl6.flowi6_oif = oif;
3626
3627                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3628         }
3629
3630         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3631         if (!skb) {
3632                 ip6_rt_put(rt);
3633                 err = -ENOBUFS;
3634                 goto errout;
3635         }
3636
3637         skb_dst_set(skb, &rt->dst);
3638
3639         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3640                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3641                             nlh->nlmsg_seq, 0);
3642         if (err < 0) {
3643                 kfree_skb(skb);
3644                 goto errout;
3645         }
3646
3647         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3648 errout:
3649         return err;
3650 }
3651
3652 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3653                      unsigned int nlm_flags)
3654 {
3655         struct sk_buff *skb;
3656         struct net *net = info->nl_net;
3657         u32 seq;
3658         int err;
3659
3660         err = -ENOBUFS;
3661         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3662
3663         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3664         if (!skb)
3665                 goto errout;
3666
3667         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3668                                 event, info->portid, seq, nlm_flags);
3669         if (err < 0) {
3670                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3671                 WARN_ON(err == -EMSGSIZE);
3672                 kfree_skb(skb);
3673                 goto errout;
3674         }
3675         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3676                     info->nlh, gfp_any());
3677         return;
3678 errout:
3679         if (err < 0)
3680                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3681 }
3682
3683 static int ip6_route_dev_notify(struct notifier_block *this,
3684                                 unsigned long event, void *ptr)
3685 {
3686         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3687         struct net *net = dev_net(dev);
3688
3689         if (event == NETDEV_REGISTER && (dev->flags & IFF_LOOPBACK)) {
3690                 net->ipv6.ip6_null_entry->dst.dev = dev;
3691                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3692 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3693                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3694                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3695                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3696                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3697 #endif
3698         }
3699
3700         return NOTIFY_OK;
3701 }
3702
3703 /*
3704  *      /proc
3705  */
3706
3707 #ifdef CONFIG_PROC_FS
3708
3709 static const struct file_operations ipv6_route_proc_fops = {
3710         .owner          = THIS_MODULE,
3711         .open           = ipv6_route_open,
3712         .read           = seq_read,
3713         .llseek         = seq_lseek,
3714         .release        = seq_release_net,
3715 };
3716
3717 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3718 {
3719         struct net *net = (struct net *)seq->private;
3720         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3721                    net->ipv6.rt6_stats->fib_nodes,
3722                    net->ipv6.rt6_stats->fib_route_nodes,
3723                    net->ipv6.rt6_stats->fib_rt_alloc,
3724                    net->ipv6.rt6_stats->fib_rt_entries,
3725                    net->ipv6.rt6_stats->fib_rt_cache,
3726                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3727                    net->ipv6.rt6_stats->fib_discarded_routes);
3728
3729         return 0;
3730 }
3731
3732 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3733 {
3734         return single_open_net(inode, file, rt6_stats_seq_show);
3735 }
3736
3737 static const struct file_operations rt6_stats_seq_fops = {
3738         .owner   = THIS_MODULE,
3739         .open    = rt6_stats_seq_open,
3740         .read    = seq_read,
3741         .llseek  = seq_lseek,
3742         .release = single_release_net,
3743 };
3744 #endif  /* CONFIG_PROC_FS */
3745
3746 #ifdef CONFIG_SYSCTL
3747
3748 static
3749 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3750                               void __user *buffer, size_t *lenp, loff_t *ppos)
3751 {
3752         struct net *net;
3753         int delay;
3754         if (!write)
3755                 return -EINVAL;
3756
3757         net = (struct net *)ctl->extra1;
3758         delay = net->ipv6.sysctl.flush_delay;
3759         proc_dointvec(ctl, write, buffer, lenp, ppos);
3760         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3761         return 0;
3762 }
3763
3764 struct ctl_table ipv6_route_table_template[] = {
3765         {
3766                 .procname       =       "flush",
3767                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3768                 .maxlen         =       sizeof(int),
3769                 .mode           =       0200,
3770                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3771         },
3772         {
3773                 .procname       =       "gc_thresh",
3774                 .data           =       &ip6_dst_ops_template.gc_thresh,
3775                 .maxlen         =       sizeof(int),
3776                 .mode           =       0644,
3777                 .proc_handler   =       proc_dointvec,
3778         },
3779         {
3780                 .procname       =       "max_size",
3781                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3782                 .maxlen         =       sizeof(int),
3783                 .mode           =       0644,
3784                 .proc_handler   =       proc_dointvec,
3785         },
3786         {
3787                 .procname       =       "gc_min_interval",
3788                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3789                 .maxlen         =       sizeof(int),
3790                 .mode           =       0644,
3791                 .proc_handler   =       proc_dointvec_jiffies,
3792         },
3793         {
3794                 .procname       =       "gc_timeout",
3795                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3796                 .maxlen         =       sizeof(int),
3797                 .mode           =       0644,
3798                 .proc_handler   =       proc_dointvec_jiffies,
3799         },
3800         {
3801                 .procname       =       "gc_interval",
3802                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3803                 .maxlen         =       sizeof(int),
3804                 .mode           =       0644,
3805                 .proc_handler   =       proc_dointvec_jiffies,
3806         },
3807         {
3808                 .procname       =       "gc_elasticity",
3809                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3810                 .maxlen         =       sizeof(int),
3811                 .mode           =       0644,
3812                 .proc_handler   =       proc_dointvec,
3813         },
3814         {
3815                 .procname       =       "mtu_expires",
3816                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3817                 .maxlen         =       sizeof(int),
3818                 .mode           =       0644,
3819                 .proc_handler   =       proc_dointvec_jiffies,
3820         },
3821         {
3822                 .procname       =       "min_adv_mss",
3823                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3824                 .maxlen         =       sizeof(int),
3825                 .mode           =       0644,
3826                 .proc_handler   =       proc_dointvec,
3827         },
3828         {
3829                 .procname       =       "gc_min_interval_ms",
3830                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3831                 .maxlen         =       sizeof(int),
3832                 .mode           =       0644,
3833                 .proc_handler   =       proc_dointvec_ms_jiffies,
3834         },
3835         { }
3836 };
3837
3838 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3839 {
3840         struct ctl_table *table;
3841
3842         table = kmemdup(ipv6_route_table_template,
3843                         sizeof(ipv6_route_table_template),
3844                         GFP_KERNEL);
3845
3846         if (table) {
3847                 table[0].data = &net->ipv6.sysctl.flush_delay;
3848                 table[0].extra1 = net;
3849                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3850                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3851                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3852                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3853                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3854                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3855                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3856                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3857                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3858
3859                 /* Don't export sysctls to unprivileged users */
3860                 if (net->user_ns != &init_user_ns)
3861                         table[0].procname = NULL;
3862         }
3863
3864         return table;
3865 }
3866 #endif
3867
3868 static int __net_init ip6_route_net_init(struct net *net)
3869 {
3870         int ret = -ENOMEM;
3871
3872         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3873                sizeof(net->ipv6.ip6_dst_ops));
3874
3875         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3876                 goto out_ip6_dst_ops;
3877
3878         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3879                                            sizeof(*net->ipv6.ip6_null_entry),
3880                                            GFP_KERNEL);
3881         if (!net->ipv6.ip6_null_entry)
3882                 goto out_ip6_dst_entries;
3883         net->ipv6.ip6_null_entry->dst.path =
3884                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3885         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3886         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3887                          ip6_template_metrics, true);
3888
3889 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3890         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3891                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3892                                                GFP_KERNEL);
3893         if (!net->ipv6.ip6_prohibit_entry)
3894                 goto out_ip6_null_entry;
3895         net->ipv6.ip6_prohibit_entry->dst.path =
3896                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3897         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3898         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3899                          ip6_template_metrics, true);
3900
3901         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3902                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3903                                                GFP_KERNEL);
3904         if (!net->ipv6.ip6_blk_hole_entry)
3905                 goto out_ip6_prohibit_entry;
3906         net->ipv6.ip6_blk_hole_entry->dst.path =
3907                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3908         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3909         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3910                          ip6_template_metrics, true);
3911 #endif
3912
3913         net->ipv6.sysctl.flush_delay = 0;
3914         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3915         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3916         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3917         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3918         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3919         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3920         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3921
3922         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3923
3924         ret = 0;
3925 out:
3926         return ret;
3927
3928 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3929 out_ip6_prohibit_entry:
3930         kfree(net->ipv6.ip6_prohibit_entry);
3931 out_ip6_null_entry:
3932         kfree(net->ipv6.ip6_null_entry);
3933 #endif
3934 out_ip6_dst_entries:
3935         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3936 out_ip6_dst_ops:
3937         goto out;
3938 }
3939
3940 static void __net_exit ip6_route_net_exit(struct net *net)
3941 {
3942         kfree(net->ipv6.ip6_null_entry);
3943 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3944         kfree(net->ipv6.ip6_prohibit_entry);
3945         kfree(net->ipv6.ip6_blk_hole_entry);
3946 #endif
3947         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3948 }
3949
3950 static int __net_init ip6_route_net_init_late(struct net *net)
3951 {
3952 #ifdef CONFIG_PROC_FS
3953         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3954         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3955 #endif
3956         return 0;
3957 }
3958
3959 static void __net_exit ip6_route_net_exit_late(struct net *net)
3960 {
3961 #ifdef CONFIG_PROC_FS
3962         remove_proc_entry("ipv6_route", net->proc_net);
3963         remove_proc_entry("rt6_stats", net->proc_net);
3964 #endif
3965 }
3966
3967 static struct pernet_operations ip6_route_net_ops = {
3968         .init = ip6_route_net_init,
3969         .exit = ip6_route_net_exit,
3970 };
3971
3972 static int __net_init ipv6_inetpeer_init(struct net *net)
3973 {
3974         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
3975
3976         if (!bp)
3977                 return -ENOMEM;
3978         inet_peer_base_init(bp);
3979         net->ipv6.peers = bp;
3980         return 0;
3981 }
3982
3983 static void __net_exit ipv6_inetpeer_exit(struct net *net)
3984 {
3985         struct inet_peer_base *bp = net->ipv6.peers;
3986
3987         net->ipv6.peers = NULL;
3988         inetpeer_invalidate_tree(bp);
3989         kfree(bp);
3990 }
3991
3992 static struct pernet_operations ipv6_inetpeer_ops = {
3993         .init   =       ipv6_inetpeer_init,
3994         .exit   =       ipv6_inetpeer_exit,
3995 };
3996
3997 static struct pernet_operations ip6_route_net_late_ops = {
3998         .init = ip6_route_net_init_late,
3999         .exit = ip6_route_net_exit_late,
4000 };
4001
4002 static struct notifier_block ip6_route_dev_notifier = {
4003         .notifier_call = ip6_route_dev_notify,
4004         .priority = 0,
4005 };
4006
4007 int __init ip6_route_init(void)
4008 {
4009         int ret;
4010         int cpu;
4011
4012         ret = -ENOMEM;
4013         ip6_dst_ops_template.kmem_cachep =
4014                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
4015                                   SLAB_HWCACHE_ALIGN, NULL);
4016         if (!ip6_dst_ops_template.kmem_cachep)
4017                 goto out;
4018
4019         ret = dst_entries_init(&ip6_dst_blackhole_ops);
4020         if (ret)
4021                 goto out_kmem_cache;
4022
4023         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
4024         if (ret)
4025                 goto out_dst_entries;
4026
4027         ret = register_pernet_subsys(&ip6_route_net_ops);
4028         if (ret)
4029                 goto out_register_inetpeer;
4030
4031         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
4032
4033         /* Registering of the loopback is done before this portion of code,
4034          * the loopback reference in rt6_info will not be taken, do it
4035          * manually for init_net */
4036         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
4037         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4038   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4039         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
4040         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4041         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
4042         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4043   #endif
4044         ret = fib6_init();
4045         if (ret)
4046                 goto out_register_subsys;
4047
4048         ret = xfrm6_init();
4049         if (ret)
4050                 goto out_fib6_init;
4051
4052         ret = fib6_rules_init();
4053         if (ret)
4054                 goto xfrm6_init;
4055
4056         ret = register_pernet_subsys(&ip6_route_net_late_ops);
4057         if (ret)
4058                 goto fib6_rules_init;
4059
4060         ret = -ENOBUFS;
4061         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
4062             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
4063             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
4064                 goto out_register_late_subsys;
4065
4066         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
4067         if (ret)
4068                 goto out_register_late_subsys;
4069
4070         for_each_possible_cpu(cpu) {
4071                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
4072
4073                 INIT_LIST_HEAD(&ul->head);
4074                 spin_lock_init(&ul->lock);
4075         }
4076
4077 out:
4078         return ret;
4079
4080 out_register_late_subsys:
4081         unregister_pernet_subsys(&ip6_route_net_late_ops);
4082 fib6_rules_init:
4083         fib6_rules_cleanup();
4084 xfrm6_init:
4085         xfrm6_fini();
4086 out_fib6_init:
4087         fib6_gc_cleanup();
4088 out_register_subsys:
4089         unregister_pernet_subsys(&ip6_route_net_ops);
4090 out_register_inetpeer:
4091         unregister_pernet_subsys(&ipv6_inetpeer_ops);
4092 out_dst_entries:
4093         dst_entries_destroy(&ip6_dst_blackhole_ops);
4094 out_kmem_cache:
4095         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4096         goto out;
4097 }
4098
4099 void ip6_route_cleanup(void)
4100 {
4101         unregister_netdevice_notifier(&ip6_route_dev_notifier);
4102         unregister_pernet_subsys(&ip6_route_net_late_ops);
4103         fib6_rules_cleanup();
4104         xfrm6_fini();
4105         fib6_gc_cleanup();
4106         unregister_pernet_subsys(&ipv6_inetpeer_ops);
4107         unregister_pernet_subsys(&ip6_route_net_ops);
4108         dst_entries_destroy(&ip6_dst_blackhole_ops);
4109         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4110 }