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         /* RTF_PCPU is an internal flag; can not be set by userspace */
1858         if (cfg->fc_flags & RTF_PCPU)
1859                 goto out;
1860
1861         if (cfg->fc_dst_len > 128 || cfg->fc_src_len > 128)
1862                 goto out;
1863 #ifndef CONFIG_IPV6_SUBTREES
1864         if (cfg->fc_src_len)
1865                 goto out;
1866 #endif
1867         if (cfg->fc_ifindex) {
1868                 err = -ENODEV;
1869                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1870                 if (!dev)
1871                         goto out;
1872                 idev = in6_dev_get(dev);
1873                 if (!idev)
1874                         goto out;
1875         }
1876
1877         if (cfg->fc_metric == 0)
1878                 cfg->fc_metric = IP6_RT_PRIO_USER;
1879
1880         err = -ENOBUFS;
1881         if (cfg->fc_nlinfo.nlh &&
1882             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1883                 table = fib6_get_table(net, cfg->fc_table);
1884                 if (!table) {
1885                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1886                         table = fib6_new_table(net, cfg->fc_table);
1887                 }
1888         } else {
1889                 table = fib6_new_table(net, cfg->fc_table);
1890         }
1891
1892         if (!table)
1893                 goto out;
1894
1895         rt = ip6_dst_alloc(net, NULL,
1896                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1897
1898         if (!rt) {
1899                 err = -ENOMEM;
1900                 goto out;
1901         }
1902
1903         if (cfg->fc_flags & RTF_EXPIRES)
1904                 rt6_set_expires(rt, jiffies +
1905                                 clock_t_to_jiffies(cfg->fc_expires));
1906         else
1907                 rt6_clean_expires(rt);
1908
1909         if (cfg->fc_protocol == RTPROT_UNSPEC)
1910                 cfg->fc_protocol = RTPROT_BOOT;
1911         rt->rt6i_protocol = cfg->fc_protocol;
1912
1913         addr_type = ipv6_addr_type(&cfg->fc_dst);
1914
1915         if (addr_type & IPV6_ADDR_MULTICAST)
1916                 rt->dst.input = ip6_mc_input;
1917         else if (cfg->fc_flags & RTF_LOCAL)
1918                 rt->dst.input = ip6_input;
1919         else
1920                 rt->dst.input = ip6_forward;
1921
1922         rt->dst.output = ip6_output;
1923
1924         if (cfg->fc_encap) {
1925                 struct lwtunnel_state *lwtstate;
1926
1927                 err = lwtunnel_build_state(cfg->fc_encap_type,
1928                                            cfg->fc_encap, AF_INET6, cfg,
1929                                            &lwtstate);
1930                 if (err)
1931                         goto out;
1932                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1933                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1934                         rt->dst.lwtstate->orig_output = rt->dst.output;
1935                         rt->dst.output = lwtunnel_output;
1936                 }
1937                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1938                         rt->dst.lwtstate->orig_input = rt->dst.input;
1939                         rt->dst.input = lwtunnel_input;
1940                 }
1941         }
1942
1943         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1944         rt->rt6i_dst.plen = cfg->fc_dst_len;
1945         if (rt->rt6i_dst.plen == 128)
1946                 rt->dst.flags |= DST_HOST;
1947
1948 #ifdef CONFIG_IPV6_SUBTREES
1949         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1950         rt->rt6i_src.plen = cfg->fc_src_len;
1951 #endif
1952
1953         rt->rt6i_metric = cfg->fc_metric;
1954
1955         /* We cannot add true routes via loopback here,
1956            they would result in kernel looping; promote them to reject routes
1957          */
1958         if ((cfg->fc_flags & RTF_REJECT) ||
1959             (dev && (dev->flags & IFF_LOOPBACK) &&
1960              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1961              !(cfg->fc_flags & RTF_LOCAL))) {
1962                 /* hold loopback dev/idev if we haven't done so. */
1963                 if (dev != net->loopback_dev) {
1964                         if (dev) {
1965                                 dev_put(dev);
1966                                 in6_dev_put(idev);
1967                         }
1968                         dev = net->loopback_dev;
1969                         dev_hold(dev);
1970                         idev = in6_dev_get(dev);
1971                         if (!idev) {
1972                                 err = -ENODEV;
1973                                 goto out;
1974                         }
1975                 }
1976                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1977                 switch (cfg->fc_type) {
1978                 case RTN_BLACKHOLE:
1979                         rt->dst.error = -EINVAL;
1980                         rt->dst.output = dst_discard_out;
1981                         rt->dst.input = dst_discard;
1982                         break;
1983                 case RTN_PROHIBIT:
1984                         rt->dst.error = -EACCES;
1985                         rt->dst.output = ip6_pkt_prohibit_out;
1986                         rt->dst.input = ip6_pkt_prohibit;
1987                         break;
1988                 case RTN_THROW:
1989                 case RTN_UNREACHABLE:
1990                 default:
1991                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1992                                         : (cfg->fc_type == RTN_UNREACHABLE)
1993                                         ? -EHOSTUNREACH : -ENETUNREACH;
1994                         rt->dst.output = ip6_pkt_discard_out;
1995                         rt->dst.input = ip6_pkt_discard;
1996                         break;
1997                 }
1998                 goto install_route;
1999         }
2000
2001         if (cfg->fc_flags & RTF_GATEWAY) {
2002                 const struct in6_addr *gw_addr;
2003                 int gwa_type;
2004
2005                 gw_addr = &cfg->fc_gateway;
2006                 gwa_type = ipv6_addr_type(gw_addr);
2007
2008                 /* if gw_addr is local we will fail to detect this in case
2009                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
2010                  * will return already-added prefix route via interface that
2011                  * prefix route was assigned to, which might be non-loopback.
2012                  */
2013                 err = -EINVAL;
2014                 if (ipv6_chk_addr_and_flags(net, gw_addr,
2015                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
2016                                             dev : NULL, 0, 0))
2017                         goto out;
2018
2019                 rt->rt6i_gateway = *gw_addr;
2020
2021                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
2022                         struct rt6_info *grt = NULL;
2023
2024                         /* IPv6 strictly inhibits using not link-local
2025                            addresses as nexthop address.
2026                            Otherwise, router will not able to send redirects.
2027                            It is very good, but in some (rare!) circumstances
2028                            (SIT, PtP, NBMA NOARP links) it is handy to allow
2029                            some exceptions. --ANK
2030                            We allow IPv4-mapped nexthops to support RFC4798-type
2031                            addressing
2032                          */
2033                         if (!(gwa_type & (IPV6_ADDR_UNICAST |
2034                                           IPV6_ADDR_MAPPED)))
2035                                 goto out;
2036
2037                         if (cfg->fc_table) {
2038                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2039
2040                                 if (grt) {
2041                                         if (grt->rt6i_flags & RTF_GATEWAY ||
2042                                             (dev && dev != grt->dst.dev)) {
2043                                                 ip6_rt_put(grt);
2044                                                 grt = NULL;
2045                                         }
2046                                 }
2047                         }
2048
2049                         if (!grt)
2050                                 grt = rt6_lookup(net, gw_addr, NULL,
2051                                                  cfg->fc_ifindex, 1);
2052
2053                         err = -EHOSTUNREACH;
2054                         if (!grt)
2055                                 goto out;
2056                         if (dev) {
2057                                 if (dev != grt->dst.dev) {
2058                                         ip6_rt_put(grt);
2059                                         goto out;
2060                                 }
2061                         } else {
2062                                 dev = grt->dst.dev;
2063                                 idev = grt->rt6i_idev;
2064                                 dev_hold(dev);
2065                                 in6_dev_hold(grt->rt6i_idev);
2066                         }
2067                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2068                                 err = 0;
2069                         ip6_rt_put(grt);
2070
2071                         if (err)
2072                                 goto out;
2073                 }
2074                 err = -EINVAL;
2075                 if (!dev || (dev->flags & IFF_LOOPBACK))
2076                         goto out;
2077         }
2078
2079         err = -ENODEV;
2080         if (!dev)
2081                 goto out;
2082
2083         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2084                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2085                         err = -EINVAL;
2086                         goto out;
2087                 }
2088                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2089                 rt->rt6i_prefsrc.plen = 128;
2090         } else
2091                 rt->rt6i_prefsrc.plen = 0;
2092
2093         rt->rt6i_flags = cfg->fc_flags;
2094
2095 install_route:
2096         rt->dst.dev = dev;
2097         rt->rt6i_idev = idev;
2098         rt->rt6i_table = table;
2099
2100         cfg->fc_nlinfo.nl_net = dev_net(dev);
2101
2102         return rt;
2103 out:
2104         if (dev)
2105                 dev_put(dev);
2106         if (idev)
2107                 in6_dev_put(idev);
2108         if (rt)
2109                 dst_free(&rt->dst);
2110
2111         return ERR_PTR(err);
2112 }
2113
2114 int ip6_route_add(struct fib6_config *cfg)
2115 {
2116         struct mx6_config mxc = { .mx = NULL, };
2117         struct rt6_info *rt;
2118         int err;
2119
2120         rt = ip6_route_info_create(cfg);
2121         if (IS_ERR(rt)) {
2122                 err = PTR_ERR(rt);
2123                 rt = NULL;
2124                 goto out;
2125         }
2126
2127         err = ip6_convert_metrics(&mxc, cfg);
2128         if (err)
2129                 goto out;
2130
2131         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc);
2132
2133         kfree(mxc.mx);
2134
2135         return err;
2136 out:
2137         if (rt)
2138                 dst_free(&rt->dst);
2139
2140         return err;
2141 }
2142
2143 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2144 {
2145         int err;
2146         struct fib6_table *table;
2147         struct net *net = dev_net(rt->dst.dev);
2148
2149         if (rt == net->ipv6.ip6_null_entry ||
2150             rt->dst.flags & DST_NOCACHE) {
2151                 err = -ENOENT;
2152                 goto out;
2153         }
2154
2155         table = rt->rt6i_table;
2156         write_lock_bh(&table->tb6_lock);
2157         err = fib6_del(rt, info);
2158         write_unlock_bh(&table->tb6_lock);
2159
2160 out:
2161         ip6_rt_put(rt);
2162         return err;
2163 }
2164
2165 int ip6_del_rt(struct rt6_info *rt)
2166 {
2167         struct nl_info info = {
2168                 .nl_net = dev_net(rt->dst.dev),
2169         };
2170         return __ip6_del_rt(rt, &info);
2171 }
2172
2173 static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
2174 {
2175         struct nl_info *info = &cfg->fc_nlinfo;
2176         struct net *net = info->nl_net;
2177         struct sk_buff *skb = NULL;
2178         struct fib6_table *table;
2179         int err = -ENOENT;
2180
2181         if (rt == net->ipv6.ip6_null_entry)
2182                 goto out_put;
2183         table = rt->rt6i_table;
2184         write_lock_bh(&table->tb6_lock);
2185
2186         if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
2187                 struct rt6_info *sibling, *next_sibling;
2188
2189                 /* prefer to send a single notification with all hops */
2190                 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
2191                 if (skb) {
2192                         u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2193
2194                         if (rt6_fill_node(net, skb, rt,
2195                                           NULL, NULL, 0, RTM_DELROUTE,
2196                                           info->portid, seq, 0) < 0) {
2197                                 kfree_skb(skb);
2198                                 skb = NULL;
2199                         } else
2200                                 info->skip_notify = 1;
2201                 }
2202
2203                 list_for_each_entry_safe(sibling, next_sibling,
2204                                          &rt->rt6i_siblings,
2205                                          rt6i_siblings) {
2206                         err = fib6_del(sibling, info);
2207                         if (err)
2208                                 goto out_unlock;
2209                 }
2210         }
2211
2212         err = fib6_del(rt, info);
2213 out_unlock:
2214         write_unlock_bh(&table->tb6_lock);
2215 out_put:
2216         ip6_rt_put(rt);
2217
2218         if (skb) {
2219                 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2220                             info->nlh, gfp_any());
2221         }
2222         return err;
2223 }
2224
2225 static int ip6_route_del(struct fib6_config *cfg)
2226 {
2227         struct fib6_table *table;
2228         struct fib6_node *fn;
2229         struct rt6_info *rt;
2230         int err = -ESRCH;
2231
2232         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2233         if (!table)
2234                 return err;
2235
2236         read_lock_bh(&table->tb6_lock);
2237
2238         fn = fib6_locate(&table->tb6_root,
2239                          &cfg->fc_dst, cfg->fc_dst_len,
2240                          &cfg->fc_src, cfg->fc_src_len);
2241
2242         if (fn) {
2243                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2244                         if ((rt->rt6i_flags & RTF_CACHE) &&
2245                             !(cfg->fc_flags & RTF_CACHE))
2246                                 continue;
2247                         if (cfg->fc_ifindex &&
2248                             (!rt->dst.dev ||
2249                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2250                                 continue;
2251                         if (cfg->fc_flags & RTF_GATEWAY &&
2252                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2253                                 continue;
2254                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2255                                 continue;
2256                         if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2257                                 continue;
2258                         dst_hold(&rt->dst);
2259                         read_unlock_bh(&table->tb6_lock);
2260
2261                         /* if gateway was specified only delete the one hop */
2262                         if (cfg->fc_flags & RTF_GATEWAY)
2263                                 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2264
2265                         return __ip6_del_rt_siblings(rt, cfg);
2266                 }
2267         }
2268         read_unlock_bh(&table->tb6_lock);
2269
2270         return err;
2271 }
2272
2273 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2274 {
2275         struct netevent_redirect netevent;
2276         struct rt6_info *rt, *nrt = NULL;
2277         struct ndisc_options ndopts;
2278         struct inet6_dev *in6_dev;
2279         struct neighbour *neigh;
2280         struct rd_msg *msg;
2281         int optlen, on_link;
2282         u8 *lladdr;
2283
2284         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2285         optlen -= sizeof(*msg);
2286
2287         if (optlen < 0) {
2288                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2289                 return;
2290         }
2291
2292         msg = (struct rd_msg *)icmp6_hdr(skb);
2293
2294         if (ipv6_addr_is_multicast(&msg->dest)) {
2295                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2296                 return;
2297         }
2298
2299         on_link = 0;
2300         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2301                 on_link = 1;
2302         } else if (ipv6_addr_type(&msg->target) !=
2303                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2304                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2305                 return;
2306         }
2307
2308         in6_dev = __in6_dev_get(skb->dev);
2309         if (!in6_dev)
2310                 return;
2311         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2312                 return;
2313
2314         /* RFC2461 8.1:
2315          *      The IP source address of the Redirect MUST be the same as the current
2316          *      first-hop router for the specified ICMP Destination Address.
2317          */
2318
2319         if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2320                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2321                 return;
2322         }
2323
2324         lladdr = NULL;
2325         if (ndopts.nd_opts_tgt_lladdr) {
2326                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2327                                              skb->dev);
2328                 if (!lladdr) {
2329                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2330                         return;
2331                 }
2332         }
2333
2334         rt = (struct rt6_info *) dst;
2335         if (rt->rt6i_flags & RTF_REJECT) {
2336                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2337                 return;
2338         }
2339
2340         /* Redirect received -> path was valid.
2341          * Look, redirects are sent only in response to data packets,
2342          * so that this nexthop apparently is reachable. --ANK
2343          */
2344         dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
2345
2346         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2347         if (!neigh)
2348                 return;
2349
2350         /*
2351          *      We have finally decided to accept it.
2352          */
2353
2354         ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2355                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2356                      NEIGH_UPDATE_F_OVERRIDE|
2357                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2358                                      NEIGH_UPDATE_F_ISROUTER)),
2359                      NDISC_REDIRECT, &ndopts);
2360
2361         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2362         if (!nrt)
2363                 goto out;
2364
2365         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2366         if (on_link)
2367                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2368
2369         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2370
2371         if (ip6_ins_rt(nrt))
2372                 goto out;
2373
2374         netevent.old = &rt->dst;
2375         netevent.new = &nrt->dst;
2376         netevent.daddr = &msg->dest;
2377         netevent.neigh = neigh;
2378         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2379
2380         if (rt->rt6i_flags & RTF_CACHE) {
2381                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2382                 ip6_del_rt(rt);
2383         }
2384
2385 out:
2386         neigh_release(neigh);
2387 }
2388
2389 /*
2390  *      Misc support functions
2391  */
2392
2393 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2394 {
2395         BUG_ON(from->dst.from);
2396
2397         rt->rt6i_flags &= ~RTF_EXPIRES;
2398         dst_hold(&from->dst);
2399         rt->dst.from = &from->dst;
2400         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2401 }
2402
2403 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2404 {
2405         rt->dst.input = ort->dst.input;
2406         rt->dst.output = ort->dst.output;
2407         rt->rt6i_dst = ort->rt6i_dst;
2408         rt->dst.error = ort->dst.error;
2409         rt->rt6i_idev = ort->rt6i_idev;
2410         if (rt->rt6i_idev)
2411                 in6_dev_hold(rt->rt6i_idev);
2412         rt->dst.lastuse = jiffies;
2413         rt->rt6i_gateway = ort->rt6i_gateway;
2414         rt->rt6i_flags = ort->rt6i_flags;
2415         rt6_set_from(rt, ort);
2416         rt->rt6i_metric = ort->rt6i_metric;
2417 #ifdef CONFIG_IPV6_SUBTREES
2418         rt->rt6i_src = ort->rt6i_src;
2419 #endif
2420         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2421         rt->rt6i_table = ort->rt6i_table;
2422         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2423 }
2424
2425 #ifdef CONFIG_IPV6_ROUTE_INFO
2426 static struct rt6_info *rt6_get_route_info(struct net *net,
2427                                            const struct in6_addr *prefix, int prefixlen,
2428                                            const struct in6_addr *gwaddr,
2429                                            struct net_device *dev)
2430 {
2431         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2432         int ifindex = dev->ifindex;
2433         struct fib6_node *fn;
2434         struct rt6_info *rt = NULL;
2435         struct fib6_table *table;
2436
2437         table = fib6_get_table(net, tb_id);
2438         if (!table)
2439                 return NULL;
2440
2441         read_lock_bh(&table->tb6_lock);
2442         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2443         if (!fn)
2444                 goto out;
2445
2446         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2447                 if (rt->dst.dev->ifindex != ifindex)
2448                         continue;
2449                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2450                         continue;
2451                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2452                         continue;
2453                 dst_hold(&rt->dst);
2454                 break;
2455         }
2456 out:
2457         read_unlock_bh(&table->tb6_lock);
2458         return rt;
2459 }
2460
2461 static struct rt6_info *rt6_add_route_info(struct net *net,
2462                                            const struct in6_addr *prefix, int prefixlen,
2463                                            const struct in6_addr *gwaddr,
2464                                            struct net_device *dev,
2465                                            unsigned int pref)
2466 {
2467         struct fib6_config cfg = {
2468                 .fc_metric      = IP6_RT_PRIO_USER,
2469                 .fc_ifindex     = dev->ifindex,
2470                 .fc_dst_len     = prefixlen,
2471                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2472                                   RTF_UP | RTF_PREF(pref),
2473                 .fc_nlinfo.portid = 0,
2474                 .fc_nlinfo.nlh = NULL,
2475                 .fc_nlinfo.nl_net = net,
2476         };
2477
2478         cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2479         cfg.fc_dst = *prefix;
2480         cfg.fc_gateway = *gwaddr;
2481
2482         /* We should treat it as a default route if prefix length is 0. */
2483         if (!prefixlen)
2484                 cfg.fc_flags |= RTF_DEFAULT;
2485
2486         ip6_route_add(&cfg);
2487
2488         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2489 }
2490 #endif
2491
2492 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2493 {
2494         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2495         struct rt6_info *rt;
2496         struct fib6_table *table;
2497
2498         table = fib6_get_table(dev_net(dev), tb_id);
2499         if (!table)
2500                 return NULL;
2501
2502         read_lock_bh(&table->tb6_lock);
2503         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2504                 if (dev == rt->dst.dev &&
2505                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2506                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2507                         break;
2508         }
2509         if (rt)
2510                 dst_hold(&rt->dst);
2511         read_unlock_bh(&table->tb6_lock);
2512         return rt;
2513 }
2514
2515 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2516                                      struct net_device *dev,
2517                                      unsigned int pref)
2518 {
2519         struct fib6_config cfg = {
2520                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2521                 .fc_metric      = IP6_RT_PRIO_USER,
2522                 .fc_ifindex     = dev->ifindex,
2523                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2524                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2525                 .fc_nlinfo.portid = 0,
2526                 .fc_nlinfo.nlh = NULL,
2527                 .fc_nlinfo.nl_net = dev_net(dev),
2528         };
2529
2530         cfg.fc_gateway = *gwaddr;
2531
2532         if (!ip6_route_add(&cfg)) {
2533                 struct fib6_table *table;
2534
2535                 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2536                 if (table)
2537                         table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2538         }
2539
2540         return rt6_get_dflt_router(gwaddr, dev);
2541 }
2542
2543 static void __rt6_purge_dflt_routers(struct fib6_table *table)
2544 {
2545         struct rt6_info *rt;
2546
2547 restart:
2548         read_lock_bh(&table->tb6_lock);
2549         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2550                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2551                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2552                         dst_hold(&rt->dst);
2553                         read_unlock_bh(&table->tb6_lock);
2554                         ip6_del_rt(rt);
2555                         goto restart;
2556                 }
2557         }
2558         read_unlock_bh(&table->tb6_lock);
2559
2560         table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2561 }
2562
2563 void rt6_purge_dflt_routers(struct net *net)
2564 {
2565         struct fib6_table *table;
2566         struct hlist_head *head;
2567         unsigned int h;
2568
2569         rcu_read_lock();
2570
2571         for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2572                 head = &net->ipv6.fib_table_hash[h];
2573                 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2574                         if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2575                                 __rt6_purge_dflt_routers(table);
2576                 }
2577         }
2578
2579         rcu_read_unlock();
2580 }
2581
2582 static void rtmsg_to_fib6_config(struct net *net,
2583                                  struct in6_rtmsg *rtmsg,
2584                                  struct fib6_config *cfg)
2585 {
2586         memset(cfg, 0, sizeof(*cfg));
2587
2588         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2589                          : RT6_TABLE_MAIN;
2590         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2591         cfg->fc_metric = rtmsg->rtmsg_metric;
2592         cfg->fc_expires = rtmsg->rtmsg_info;
2593         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2594         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2595         cfg->fc_flags = rtmsg->rtmsg_flags;
2596
2597         cfg->fc_nlinfo.nl_net = net;
2598
2599         cfg->fc_dst = rtmsg->rtmsg_dst;
2600         cfg->fc_src = rtmsg->rtmsg_src;
2601         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2602 }
2603
2604 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2605 {
2606         struct fib6_config cfg;
2607         struct in6_rtmsg rtmsg;
2608         int err;
2609
2610         switch (cmd) {
2611         case SIOCADDRT:         /* Add a route */
2612         case SIOCDELRT:         /* Delete a route */
2613                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2614                         return -EPERM;
2615                 err = copy_from_user(&rtmsg, arg,
2616                                      sizeof(struct in6_rtmsg));
2617                 if (err)
2618                         return -EFAULT;
2619
2620                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2621
2622                 rtnl_lock();
2623                 switch (cmd) {
2624                 case SIOCADDRT:
2625                         err = ip6_route_add(&cfg);
2626                         break;
2627                 case SIOCDELRT:
2628                         err = ip6_route_del(&cfg);
2629                         break;
2630                 default:
2631                         err = -EINVAL;
2632                 }
2633                 rtnl_unlock();
2634
2635                 return err;
2636         }
2637
2638         return -EINVAL;
2639 }
2640
2641 /*
2642  *      Drop the packet on the floor
2643  */
2644
2645 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2646 {
2647         int type;
2648         struct dst_entry *dst = skb_dst(skb);
2649         switch (ipstats_mib_noroutes) {
2650         case IPSTATS_MIB_INNOROUTES:
2651                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2652                 if (type == IPV6_ADDR_ANY) {
2653                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2654                                       IPSTATS_MIB_INADDRERRORS);
2655                         break;
2656                 }
2657                 /* FALLTHROUGH */
2658         case IPSTATS_MIB_OUTNOROUTES:
2659                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2660                               ipstats_mib_noroutes);
2661                 break;
2662         }
2663         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2664         kfree_skb(skb);
2665         return 0;
2666 }
2667
2668 static int ip6_pkt_discard(struct sk_buff *skb)
2669 {
2670         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2671 }
2672
2673 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2674 {
2675         skb->dev = skb_dst(skb)->dev;
2676         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2677 }
2678
2679 static int ip6_pkt_prohibit(struct sk_buff *skb)
2680 {
2681         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2682 }
2683
2684 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2685 {
2686         skb->dev = skb_dst(skb)->dev;
2687         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2688 }
2689
2690 /*
2691  *      Allocate a dst for local (unicast / anycast) address.
2692  */
2693
2694 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2695                                     const struct in6_addr *addr,
2696                                     bool anycast)
2697 {
2698         u32 tb_id;
2699         struct net *net = dev_net(idev->dev);
2700         struct net_device *dev = net->loopback_dev;
2701         struct rt6_info *rt;
2702
2703         /* use L3 Master device as loopback for host routes if device
2704          * is enslaved and address is not link local or multicast
2705          */
2706         if (!rt6_need_strict(addr))
2707                 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2708
2709         rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2710         if (!rt)
2711                 return ERR_PTR(-ENOMEM);
2712
2713         in6_dev_hold(idev);
2714
2715         rt->dst.flags |= DST_HOST;
2716         rt->dst.input = ip6_input;
2717         rt->dst.output = ip6_output;
2718         rt->rt6i_idev = idev;
2719
2720         rt->rt6i_protocol = RTPROT_KERNEL;
2721         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2722         if (anycast)
2723                 rt->rt6i_flags |= RTF_ANYCAST;
2724         else
2725                 rt->rt6i_flags |= RTF_LOCAL;
2726
2727         rt->rt6i_gateway  = *addr;
2728         rt->rt6i_dst.addr = *addr;
2729         rt->rt6i_dst.plen = 128;
2730         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2731         rt->rt6i_table = fib6_get_table(net, tb_id);
2732         rt->dst.flags |= DST_NOCACHE;
2733
2734         atomic_set(&rt->dst.__refcnt, 1);
2735
2736         return rt;
2737 }
2738
2739 /* remove deleted ip from prefsrc entries */
2740 struct arg_dev_net_ip {
2741         struct net_device *dev;
2742         struct net *net;
2743         struct in6_addr *addr;
2744 };
2745
2746 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2747 {
2748         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2749         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2750         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2751
2752         if (((void *)rt->dst.dev == dev || !dev) &&
2753             rt != net->ipv6.ip6_null_entry &&
2754             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2755                 /* remove prefsrc entry */
2756                 rt->rt6i_prefsrc.plen = 0;
2757         }
2758         return 0;
2759 }
2760
2761 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2762 {
2763         struct net *net = dev_net(ifp->idev->dev);
2764         struct arg_dev_net_ip adni = {
2765                 .dev = ifp->idev->dev,
2766                 .net = net,
2767                 .addr = &ifp->addr,
2768         };
2769         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2770 }
2771
2772 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2773 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2774
2775 /* Remove routers and update dst entries when gateway turn into host. */
2776 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2777 {
2778         struct in6_addr *gateway = (struct in6_addr *)arg;
2779
2780         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2781              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2782              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2783                 return -1;
2784         }
2785         return 0;
2786 }
2787
2788 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2789 {
2790         fib6_clean_all(net, fib6_clean_tohost, gateway);
2791 }
2792
2793 struct arg_dev_net {
2794         struct net_device *dev;
2795         struct net *net;
2796 };
2797
2798 /* called with write lock held for table with rt */
2799 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2800 {
2801         const struct arg_dev_net *adn = arg;
2802         const struct net_device *dev = adn->dev;
2803
2804         if ((rt->dst.dev == dev || !dev) &&
2805             rt != adn->net->ipv6.ip6_null_entry &&
2806             (rt->rt6i_nsiblings == 0 ||
2807              !rt->rt6i_idev->cnf.ignore_routes_with_linkdown))
2808                 return -1;
2809
2810         return 0;
2811 }
2812
2813 void rt6_ifdown(struct net *net, struct net_device *dev)
2814 {
2815         struct arg_dev_net adn = {
2816                 .dev = dev,
2817                 .net = net,
2818         };
2819
2820         fib6_clean_all(net, fib6_ifdown, &adn);
2821         icmp6_clean_all(fib6_ifdown, &adn);
2822         if (dev)
2823                 rt6_uncached_list_flush_dev(net, dev);
2824 }
2825
2826 struct rt6_mtu_change_arg {
2827         struct net_device *dev;
2828         unsigned int mtu;
2829 };
2830
2831 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2832 {
2833         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2834         struct inet6_dev *idev;
2835
2836         /* In IPv6 pmtu discovery is not optional,
2837            so that RTAX_MTU lock cannot disable it.
2838            We still use this lock to block changes
2839            caused by addrconf/ndisc.
2840         */
2841
2842         idev = __in6_dev_get(arg->dev);
2843         if (!idev)
2844                 return 0;
2845
2846         /* For administrative MTU increase, there is no way to discover
2847            IPv6 PMTU increase, so PMTU increase should be updated here.
2848            Since RFC 1981 doesn't include administrative MTU increase
2849            update PMTU increase is a MUST. (i.e. jumbo frame)
2850          */
2851         /*
2852            If new MTU is less than route PMTU, this new MTU will be the
2853            lowest MTU in the path, update the route PMTU to reflect PMTU
2854            decreases; if new MTU is greater than route PMTU, and the
2855            old MTU is the lowest MTU in the path, update the route PMTU
2856            to reflect the increase. In this case if the other nodes' MTU
2857            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2858            PMTU discovery.
2859          */
2860         if (rt->dst.dev == arg->dev &&
2861             dst_metric_raw(&rt->dst, RTAX_MTU) &&
2862             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2863                 if (rt->rt6i_flags & RTF_CACHE) {
2864                         /* For RTF_CACHE with rt6i_pmtu == 0
2865                          * (i.e. a redirected route),
2866                          * the metrics of its rt->dst.from has already
2867                          * been updated.
2868                          */
2869                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2870                                 rt->rt6i_pmtu = arg->mtu;
2871                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2872                            (dst_mtu(&rt->dst) < arg->mtu &&
2873                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2874                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2875                 }
2876         }
2877         return 0;
2878 }
2879
2880 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2881 {
2882         struct rt6_mtu_change_arg arg = {
2883                 .dev = dev,
2884                 .mtu = mtu,
2885         };
2886
2887         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2888 }
2889
2890 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2891         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2892         [RTA_OIF]               = { .type = NLA_U32 },
2893         [RTA_IIF]               = { .type = NLA_U32 },
2894         [RTA_PRIORITY]          = { .type = NLA_U32 },
2895         [RTA_METRICS]           = { .type = NLA_NESTED },
2896         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2897         [RTA_PREF]              = { .type = NLA_U8 },
2898         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2899         [RTA_ENCAP]             = { .type = NLA_NESTED },
2900         [RTA_EXPIRES]           = { .type = NLA_U32 },
2901         [RTA_UID]               = { .type = NLA_U32 },
2902         [RTA_MARK]              = { .type = NLA_U32 },
2903 };
2904
2905 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2906                               struct fib6_config *cfg)
2907 {
2908         struct rtmsg *rtm;
2909         struct nlattr *tb[RTA_MAX+1];
2910         unsigned int pref;
2911         int err;
2912
2913         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
2914                           NULL);
2915         if (err < 0)
2916                 goto errout;
2917
2918         err = -EINVAL;
2919         rtm = nlmsg_data(nlh);
2920         memset(cfg, 0, sizeof(*cfg));
2921
2922         cfg->fc_table = rtm->rtm_table;
2923         cfg->fc_dst_len = rtm->rtm_dst_len;
2924         cfg->fc_src_len = rtm->rtm_src_len;
2925         cfg->fc_flags = RTF_UP;
2926         cfg->fc_protocol = rtm->rtm_protocol;
2927         cfg->fc_type = rtm->rtm_type;
2928
2929         if (rtm->rtm_type == RTN_UNREACHABLE ||
2930             rtm->rtm_type == RTN_BLACKHOLE ||
2931             rtm->rtm_type == RTN_PROHIBIT ||
2932             rtm->rtm_type == RTN_THROW)
2933                 cfg->fc_flags |= RTF_REJECT;
2934
2935         if (rtm->rtm_type == RTN_LOCAL)
2936                 cfg->fc_flags |= RTF_LOCAL;
2937
2938         if (rtm->rtm_flags & RTM_F_CLONED)
2939                 cfg->fc_flags |= RTF_CACHE;
2940
2941         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2942         cfg->fc_nlinfo.nlh = nlh;
2943         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2944
2945         if (tb[RTA_GATEWAY]) {
2946                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2947                 cfg->fc_flags |= RTF_GATEWAY;
2948         }
2949
2950         if (tb[RTA_DST]) {
2951                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2952
2953                 if (nla_len(tb[RTA_DST]) < plen)
2954                         goto errout;
2955
2956                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2957         }
2958
2959         if (tb[RTA_SRC]) {
2960                 int plen = (rtm->rtm_src_len + 7) >> 3;
2961
2962                 if (nla_len(tb[RTA_SRC]) < plen)
2963                         goto errout;
2964
2965                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2966         }
2967
2968         if (tb[RTA_PREFSRC])
2969                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2970
2971         if (tb[RTA_OIF])
2972                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2973
2974         if (tb[RTA_PRIORITY])
2975                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2976
2977         if (tb[RTA_METRICS]) {
2978                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2979                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2980         }
2981
2982         if (tb[RTA_TABLE])
2983                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2984
2985         if (tb[RTA_MULTIPATH]) {
2986                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2987                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2988
2989                 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
2990                                                      cfg->fc_mp_len);
2991                 if (err < 0)
2992                         goto errout;
2993         }
2994
2995         if (tb[RTA_PREF]) {
2996                 pref = nla_get_u8(tb[RTA_PREF]);
2997                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2998                     pref != ICMPV6_ROUTER_PREF_HIGH)
2999                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
3000                 cfg->fc_flags |= RTF_PREF(pref);
3001         }
3002
3003         if (tb[RTA_ENCAP])
3004                 cfg->fc_encap = tb[RTA_ENCAP];
3005
3006         if (tb[RTA_ENCAP_TYPE]) {
3007                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
3008
3009                 err = lwtunnel_valid_encap_type(cfg->fc_encap_type);
3010                 if (err < 0)
3011                         goto errout;
3012         }
3013
3014         if (tb[RTA_EXPIRES]) {
3015                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
3016
3017                 if (addrconf_finite_timeout(timeout)) {
3018                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
3019                         cfg->fc_flags |= RTF_EXPIRES;
3020                 }
3021         }
3022
3023         err = 0;
3024 errout:
3025         return err;
3026 }
3027
3028 struct rt6_nh {
3029         struct rt6_info *rt6_info;
3030         struct fib6_config r_cfg;
3031         struct mx6_config mxc;
3032         struct list_head next;
3033 };
3034
3035 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
3036 {
3037         struct rt6_nh *nh;
3038
3039         list_for_each_entry(nh, rt6_nh_list, next) {
3040                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
3041                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
3042                         nh->r_cfg.fc_ifindex);
3043         }
3044 }
3045
3046 static int ip6_route_info_append(struct list_head *rt6_nh_list,
3047                                  struct rt6_info *rt, struct fib6_config *r_cfg)
3048 {
3049         struct rt6_nh *nh;
3050         struct rt6_info *rtnh;
3051         int err = -EEXIST;
3052
3053         list_for_each_entry(nh, rt6_nh_list, next) {
3054                 /* check if rt6_info already exists */
3055                 rtnh = nh->rt6_info;
3056
3057                 if (rtnh->dst.dev == rt->dst.dev &&
3058                     rtnh->rt6i_idev == rt->rt6i_idev &&
3059                     ipv6_addr_equal(&rtnh->rt6i_gateway,
3060                                     &rt->rt6i_gateway))
3061                         return err;
3062         }
3063
3064         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
3065         if (!nh)
3066                 return -ENOMEM;
3067         nh->rt6_info = rt;
3068         err = ip6_convert_metrics(&nh->mxc, r_cfg);
3069         if (err) {
3070                 kfree(nh);
3071                 return err;
3072         }
3073         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
3074         list_add_tail(&nh->next, rt6_nh_list);
3075
3076         return 0;
3077 }
3078
3079 static void ip6_route_mpath_notify(struct rt6_info *rt,
3080                                    struct rt6_info *rt_last,
3081                                    struct nl_info *info,
3082                                    __u16 nlflags)
3083 {
3084         /* if this is an APPEND route, then rt points to the first route
3085          * inserted and rt_last points to last route inserted. Userspace
3086          * wants a consistent dump of the route which starts at the first
3087          * nexthop. Since sibling routes are always added at the end of
3088          * the list, find the first sibling of the last route appended
3089          */
3090         if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->rt6i_nsiblings) {
3091                 rt = list_first_entry(&rt_last->rt6i_siblings,
3092                                       struct rt6_info,
3093                                       rt6i_siblings);
3094         }
3095
3096         if (rt)
3097                 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
3098 }
3099
3100 static int ip6_route_multipath_add(struct fib6_config *cfg)
3101 {
3102         struct rt6_info *rt_notif = NULL, *rt_last = NULL;
3103         struct nl_info *info = &cfg->fc_nlinfo;
3104         struct fib6_config r_cfg;
3105         struct rtnexthop *rtnh;
3106         struct rt6_info *rt;
3107         struct rt6_nh *err_nh;
3108         struct rt6_nh *nh, *nh_safe;
3109         __u16 nlflags;
3110         int remaining;
3111         int attrlen;
3112         int err = 1;
3113         int nhn = 0;
3114         int replace = (cfg->fc_nlinfo.nlh &&
3115                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
3116         LIST_HEAD(rt6_nh_list);
3117
3118         nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
3119         if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
3120                 nlflags |= NLM_F_APPEND;
3121
3122         remaining = cfg->fc_mp_len;
3123         rtnh = (struct rtnexthop *)cfg->fc_mp;
3124
3125         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3126          * rt6_info structs per nexthop
3127          */
3128         while (rtnh_ok(rtnh, remaining)) {
3129                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3130                 if (rtnh->rtnh_ifindex)
3131                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3132
3133                 attrlen = rtnh_attrlen(rtnh);
3134                 if (attrlen > 0) {
3135                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3136
3137                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3138                         if (nla) {
3139                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
3140                                 r_cfg.fc_flags |= RTF_GATEWAY;
3141                         }
3142                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3143                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3144                         if (nla)
3145                                 r_cfg.fc_encap_type = nla_get_u16(nla);
3146                 }
3147
3148                 rt = ip6_route_info_create(&r_cfg);
3149                 if (IS_ERR(rt)) {
3150                         err = PTR_ERR(rt);
3151                         rt = NULL;
3152                         goto cleanup;
3153                 }
3154
3155                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
3156                 if (err) {
3157                         dst_free(&rt->dst);
3158                         goto cleanup;
3159                 }
3160
3161                 rtnh = rtnh_next(rtnh, &remaining);
3162         }
3163
3164         /* for add and replace send one notification with all nexthops.
3165          * Skip the notification in fib6_add_rt2node and send one with
3166          * the full route when done
3167          */
3168         info->skip_notify = 1;
3169
3170         err_nh = NULL;
3171         list_for_each_entry(nh, &rt6_nh_list, next) {
3172                 rt_last = nh->rt6_info;
3173                 err = __ip6_ins_rt(nh->rt6_info, info, &nh->mxc);
3174                 /* save reference to first route for notification */
3175                 if (!rt_notif && !err)
3176                         rt_notif = nh->rt6_info;
3177
3178                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3179                 nh->rt6_info = NULL;
3180                 if (err) {
3181                         if (replace && nhn)
3182                                 ip6_print_replace_route_err(&rt6_nh_list);
3183                         err_nh = nh;
3184                         goto add_errout;
3185                 }
3186
3187                 /* Because each route is added like a single route we remove
3188                  * these flags after the first nexthop: if there is a collision,
3189                  * we have already failed to add the first nexthop:
3190                  * fib6_add_rt2node() has rejected it; when replacing, old
3191                  * nexthops have been replaced by first new, the rest should
3192                  * be added to it.
3193                  */
3194                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3195                                                      NLM_F_REPLACE);
3196                 nhn++;
3197         }
3198
3199         /* success ... tell user about new route */
3200         ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3201         goto cleanup;
3202
3203 add_errout:
3204         /* send notification for routes that were added so that
3205          * the delete notifications sent by ip6_route_del are
3206          * coherent
3207          */
3208         if (rt_notif)
3209                 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3210
3211         /* Delete routes that were already added */
3212         list_for_each_entry(nh, &rt6_nh_list, next) {
3213                 if (err_nh == nh)
3214                         break;
3215                 ip6_route_del(&nh->r_cfg);
3216         }
3217
3218 cleanup:
3219         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3220                 if (nh->rt6_info)
3221                         dst_free(&nh->rt6_info->dst);
3222                 kfree(nh->mxc.mx);
3223                 list_del(&nh->next);
3224                 kfree(nh);
3225         }
3226
3227         return err;
3228 }
3229
3230 static int ip6_route_multipath_del(struct fib6_config *cfg)
3231 {
3232         struct fib6_config r_cfg;
3233         struct rtnexthop *rtnh;
3234         int remaining;
3235         int attrlen;
3236         int err = 1, last_err = 0;
3237
3238         remaining = cfg->fc_mp_len;
3239         rtnh = (struct rtnexthop *)cfg->fc_mp;
3240
3241         /* Parse a Multipath Entry */
3242         while (rtnh_ok(rtnh, remaining)) {
3243                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3244                 if (rtnh->rtnh_ifindex)
3245                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3246
3247                 attrlen = rtnh_attrlen(rtnh);
3248                 if (attrlen > 0) {
3249                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3250
3251                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3252                         if (nla) {
3253                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3254                                 r_cfg.fc_flags |= RTF_GATEWAY;
3255                         }
3256                 }
3257                 err = ip6_route_del(&r_cfg);
3258                 if (err)
3259                         last_err = err;
3260
3261                 rtnh = rtnh_next(rtnh, &remaining);
3262         }
3263
3264         return last_err;
3265 }
3266
3267 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3268                               struct netlink_ext_ack *extack)
3269 {
3270         struct fib6_config cfg;
3271         int err;
3272
3273         err = rtm_to_fib6_config(skb, nlh, &cfg);
3274         if (err < 0)
3275                 return err;
3276
3277         if (cfg.fc_mp)
3278                 return ip6_route_multipath_del(&cfg);
3279         else {
3280                 cfg.fc_delete_all_nh = 1;
3281                 return ip6_route_del(&cfg);
3282         }
3283 }
3284
3285 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3286                               struct netlink_ext_ack *extack)
3287 {
3288         struct fib6_config cfg;
3289         int err;
3290
3291         err = rtm_to_fib6_config(skb, nlh, &cfg);
3292         if (err < 0)
3293                 return err;
3294
3295         if (cfg.fc_mp)
3296                 return ip6_route_multipath_add(&cfg);
3297         else
3298                 return ip6_route_add(&cfg);
3299 }
3300
3301 static size_t rt6_nlmsg_size(struct rt6_info *rt)
3302 {
3303         int nexthop_len = 0;
3304
3305         if (rt->rt6i_nsiblings) {
3306                 nexthop_len = nla_total_size(0)  /* RTA_MULTIPATH */
3307                             + NLA_ALIGN(sizeof(struct rtnexthop))
3308                             + nla_total_size(16) /* RTA_GATEWAY */
3309                             + lwtunnel_get_encap_size(rt->dst.lwtstate);
3310
3311                 nexthop_len *= rt->rt6i_nsiblings;
3312         }
3313
3314         return NLMSG_ALIGN(sizeof(struct rtmsg))
3315                + nla_total_size(16) /* RTA_SRC */
3316                + nla_total_size(16) /* RTA_DST */
3317                + nla_total_size(16) /* RTA_GATEWAY */
3318                + nla_total_size(16) /* RTA_PREFSRC */
3319                + nla_total_size(4) /* RTA_TABLE */
3320                + nla_total_size(4) /* RTA_IIF */
3321                + nla_total_size(4) /* RTA_OIF */
3322                + nla_total_size(4) /* RTA_PRIORITY */
3323                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3324                + nla_total_size(sizeof(struct rta_cacheinfo))
3325                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3326                + nla_total_size(1) /* RTA_PREF */
3327                + lwtunnel_get_encap_size(rt->dst.lwtstate)
3328                + nexthop_len;
3329 }
3330
3331 static int rt6_nexthop_info(struct sk_buff *skb, struct rt6_info *rt,
3332                             unsigned int *flags, bool skip_oif)
3333 {
3334         if (!netif_running(rt->dst.dev) || !netif_carrier_ok(rt->dst.dev)) {
3335                 *flags |= RTNH_F_LINKDOWN;
3336                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3337                         *flags |= RTNH_F_DEAD;
3338         }
3339
3340         if (rt->rt6i_flags & RTF_GATEWAY) {
3341                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3342                         goto nla_put_failure;
3343         }
3344
3345         /* not needed for multipath encoding b/c it has a rtnexthop struct */
3346         if (!skip_oif && rt->dst.dev &&
3347             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3348                 goto nla_put_failure;
3349
3350         if (rt->dst.lwtstate &&
3351             lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3352                 goto nla_put_failure;
3353
3354         return 0;
3355
3356 nla_put_failure:
3357         return -EMSGSIZE;
3358 }
3359
3360 /* add multipath next hop */
3361 static int rt6_add_nexthop(struct sk_buff *skb, struct rt6_info *rt)
3362 {
3363         struct rtnexthop *rtnh;
3364         unsigned int flags = 0;
3365
3366         rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
3367         if (!rtnh)
3368                 goto nla_put_failure;
3369
3370         rtnh->rtnh_hops = 0;
3371         rtnh->rtnh_ifindex = rt->dst.dev ? rt->dst.dev->ifindex : 0;
3372
3373         if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
3374                 goto nla_put_failure;
3375
3376         rtnh->rtnh_flags = flags;
3377
3378         /* length of rtnetlink header + attributes */
3379         rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
3380
3381         return 0;
3382
3383 nla_put_failure:
3384         return -EMSGSIZE;
3385 }
3386
3387 static int rt6_fill_node(struct net *net,
3388                          struct sk_buff *skb, struct rt6_info *rt,
3389                          struct in6_addr *dst, struct in6_addr *src,
3390                          int iif, int type, u32 portid, u32 seq,
3391                          unsigned int flags)
3392 {
3393         u32 metrics[RTAX_MAX];
3394         struct rtmsg *rtm;
3395         struct nlmsghdr *nlh;
3396         long expires;
3397         u32 table;
3398
3399         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3400         if (!nlh)
3401                 return -EMSGSIZE;
3402
3403         rtm = nlmsg_data(nlh);
3404         rtm->rtm_family = AF_INET6;
3405         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3406         rtm->rtm_src_len = rt->rt6i_src.plen;
3407         rtm->rtm_tos = 0;
3408         if (rt->rt6i_table)
3409                 table = rt->rt6i_table->tb6_id;
3410         else
3411                 table = RT6_TABLE_UNSPEC;
3412         rtm->rtm_table = table;
3413         if (nla_put_u32(skb, RTA_TABLE, table))
3414                 goto nla_put_failure;
3415         if (rt->rt6i_flags & RTF_REJECT) {
3416                 switch (rt->dst.error) {
3417                 case -EINVAL:
3418                         rtm->rtm_type = RTN_BLACKHOLE;
3419                         break;
3420                 case -EACCES:
3421                         rtm->rtm_type = RTN_PROHIBIT;
3422                         break;
3423                 case -EAGAIN:
3424                         rtm->rtm_type = RTN_THROW;
3425                         break;
3426                 default:
3427                         rtm->rtm_type = RTN_UNREACHABLE;
3428                         break;
3429                 }
3430         }
3431         else if (rt->rt6i_flags & RTF_LOCAL)
3432                 rtm->rtm_type = RTN_LOCAL;
3433         else if (rt->rt6i_flags & RTF_ANYCAST)
3434                 rtm->rtm_type = RTN_ANYCAST;
3435         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3436                 rtm->rtm_type = RTN_LOCAL;
3437         else
3438                 rtm->rtm_type = RTN_UNICAST;
3439         rtm->rtm_flags = 0;
3440         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3441         rtm->rtm_protocol = rt->rt6i_protocol;
3442         if (rt->rt6i_flags & RTF_DYNAMIC)
3443                 rtm->rtm_protocol = RTPROT_REDIRECT;
3444         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3445                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3446                         rtm->rtm_protocol = RTPROT_RA;
3447                 else
3448                         rtm->rtm_protocol = RTPROT_KERNEL;
3449         }
3450
3451         if (rt->rt6i_flags & RTF_CACHE)
3452                 rtm->rtm_flags |= RTM_F_CLONED;
3453
3454         if (dst) {
3455                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3456                         goto nla_put_failure;
3457                 rtm->rtm_dst_len = 128;
3458         } else if (rtm->rtm_dst_len)
3459                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3460                         goto nla_put_failure;
3461 #ifdef CONFIG_IPV6_SUBTREES
3462         if (src) {
3463                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3464                         goto nla_put_failure;
3465                 rtm->rtm_src_len = 128;
3466         } else if (rtm->rtm_src_len &&
3467                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3468                 goto nla_put_failure;
3469 #endif
3470         if (iif) {
3471 #ifdef CONFIG_IPV6_MROUTE
3472                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3473                         int err = ip6mr_get_route(net, skb, rtm, portid);
3474
3475                         if (err == 0)
3476                                 return 0;
3477                         if (err < 0)
3478                                 goto nla_put_failure;
3479                 } else
3480 #endif
3481                         if (nla_put_u32(skb, RTA_IIF, iif))
3482                                 goto nla_put_failure;
3483         } else if (dst) {
3484                 struct in6_addr saddr_buf;
3485                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3486                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3487                         goto nla_put_failure;
3488         }
3489
3490         if (rt->rt6i_prefsrc.plen) {
3491                 struct in6_addr saddr_buf;
3492                 saddr_buf = rt->rt6i_prefsrc.addr;
3493                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3494                         goto nla_put_failure;
3495         }
3496
3497         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3498         if (rt->rt6i_pmtu)
3499                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3500         if (rtnetlink_put_metrics(skb, metrics) < 0)
3501                 goto nla_put_failure;
3502
3503         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3504                 goto nla_put_failure;
3505
3506         /* For multipath routes, walk the siblings list and add
3507          * each as a nexthop within RTA_MULTIPATH.
3508          */
3509         if (rt->rt6i_nsiblings) {
3510                 struct rt6_info *sibling, *next_sibling;
3511                 struct nlattr *mp;
3512
3513                 mp = nla_nest_start(skb, RTA_MULTIPATH);
3514                 if (!mp)
3515                         goto nla_put_failure;
3516
3517                 if (rt6_add_nexthop(skb, rt) < 0)
3518                         goto nla_put_failure;
3519
3520                 list_for_each_entry_safe(sibling, next_sibling,
3521                                          &rt->rt6i_siblings, rt6i_siblings) {
3522                         if (rt6_add_nexthop(skb, sibling) < 0)
3523                                 goto nla_put_failure;
3524                 }
3525
3526                 nla_nest_end(skb, mp);
3527         } else {
3528                 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
3529                         goto nla_put_failure;
3530         }
3531
3532         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3533
3534         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3535                 goto nla_put_failure;
3536
3537         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3538                 goto nla_put_failure;
3539
3540
3541         nlmsg_end(skb, nlh);
3542         return 0;
3543
3544 nla_put_failure:
3545         nlmsg_cancel(skb, nlh);
3546         return -EMSGSIZE;
3547 }
3548
3549 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3550 {
3551         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3552         struct net *net = arg->net;
3553
3554         if (rt == net->ipv6.ip6_null_entry)
3555                 return 0;
3556
3557         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3558                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3559
3560                 /* user wants prefix routes only */
3561                 if (rtm->rtm_flags & RTM_F_PREFIX &&
3562                     !(rt->rt6i_flags & RTF_PREFIX_RT)) {
3563                         /* success since this is not a prefix route */
3564                         return 1;
3565                 }
3566         }
3567
3568         return rt6_fill_node(net,
3569                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3570                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3571                      NLM_F_MULTI);
3572 }
3573
3574 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3575                               struct netlink_ext_ack *extack)
3576 {
3577         struct net *net = sock_net(in_skb->sk);
3578         struct nlattr *tb[RTA_MAX+1];
3579         struct rt6_info *rt;
3580         struct sk_buff *skb;
3581         struct rtmsg *rtm;
3582         struct flowi6 fl6;
3583         int err, iif = 0, oif = 0;
3584
3585         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
3586                           extack);
3587         if (err < 0)
3588                 goto errout;
3589
3590         err = -EINVAL;
3591         memset(&fl6, 0, sizeof(fl6));
3592         rtm = nlmsg_data(nlh);
3593         fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3594
3595         if (tb[RTA_SRC]) {
3596                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3597                         goto errout;
3598
3599                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3600         }
3601
3602         if (tb[RTA_DST]) {
3603                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3604                         goto errout;
3605
3606                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3607         }
3608
3609         if (tb[RTA_IIF])
3610                 iif = nla_get_u32(tb[RTA_IIF]);
3611
3612         if (tb[RTA_OIF])
3613                 oif = nla_get_u32(tb[RTA_OIF]);
3614
3615         if (tb[RTA_MARK])
3616                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3617
3618         if (tb[RTA_UID])
3619                 fl6.flowi6_uid = make_kuid(current_user_ns(),
3620                                            nla_get_u32(tb[RTA_UID]));
3621         else
3622                 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3623
3624         if (iif) {
3625                 struct net_device *dev;
3626                 int flags = 0;
3627
3628                 dev = __dev_get_by_index(net, iif);
3629                 if (!dev) {
3630                         err = -ENODEV;
3631                         goto errout;
3632                 }
3633
3634                 fl6.flowi6_iif = iif;
3635
3636                 if (!ipv6_addr_any(&fl6.saddr))
3637                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3638
3639                 rt = (struct rt6_info *)ip6_route_input_lookup(net, dev, &fl6,
3640                                                                flags);
3641         } else {
3642                 fl6.flowi6_oif = oif;
3643
3644                 rt = (struct rt6_info *)ip6_route_output(net, NULL, &fl6);
3645         }
3646
3647         if (rt == net->ipv6.ip6_null_entry) {
3648                 err = rt->dst.error;
3649                 ip6_rt_put(rt);
3650                 goto errout;
3651         }
3652
3653         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3654         if (!skb) {
3655                 ip6_rt_put(rt);
3656                 err = -ENOBUFS;
3657                 goto errout;
3658         }
3659
3660         skb_dst_set(skb, &rt->dst);
3661
3662         err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3663                             RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3664                             nlh->nlmsg_seq, 0);
3665         if (err < 0) {
3666                 kfree_skb(skb);
3667                 goto errout;
3668         }
3669
3670         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3671 errout:
3672         return err;
3673 }
3674
3675 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3676                      unsigned int nlm_flags)
3677 {
3678         struct sk_buff *skb;
3679         struct net *net = info->nl_net;
3680         u32 seq;
3681         int err;
3682
3683         err = -ENOBUFS;
3684         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3685
3686         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3687         if (!skb)
3688                 goto errout;
3689
3690         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3691                                 event, info->portid, seq, nlm_flags);
3692         if (err < 0) {
3693                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3694                 WARN_ON(err == -EMSGSIZE);
3695                 kfree_skb(skb);
3696                 goto errout;
3697         }
3698         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3699                     info->nlh, gfp_any());
3700         return;
3701 errout:
3702         if (err < 0)
3703                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3704 }
3705
3706 static int ip6_route_dev_notify(struct notifier_block *this,
3707                                 unsigned long event, void *ptr)
3708 {
3709         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3710         struct net *net = dev_net(dev);
3711
3712         if (!(dev->flags & IFF_LOOPBACK))
3713                 return NOTIFY_OK;
3714
3715         if (event == NETDEV_REGISTER) {
3716                 net->ipv6.ip6_null_entry->dst.dev = dev;
3717                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3718 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3719                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3720                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3721                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3722                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3723 #endif
3724          } else if (event == NETDEV_UNREGISTER) {
3725                 in6_dev_put(net->ipv6.ip6_null_entry->rt6i_idev);
3726 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3727                 in6_dev_put(net->ipv6.ip6_prohibit_entry->rt6i_idev);
3728                 in6_dev_put(net->ipv6.ip6_blk_hole_entry->rt6i_idev);
3729 #endif
3730         }
3731
3732         return NOTIFY_OK;
3733 }
3734
3735 /*
3736  *      /proc
3737  */
3738
3739 #ifdef CONFIG_PROC_FS
3740
3741 static const struct file_operations ipv6_route_proc_fops = {
3742         .owner          = THIS_MODULE,
3743         .open           = ipv6_route_open,
3744         .read           = seq_read,
3745         .llseek         = seq_lseek,
3746         .release        = seq_release_net,
3747 };
3748
3749 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3750 {
3751         struct net *net = (struct net *)seq->private;
3752         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3753                    net->ipv6.rt6_stats->fib_nodes,
3754                    net->ipv6.rt6_stats->fib_route_nodes,
3755                    net->ipv6.rt6_stats->fib_rt_alloc,
3756                    net->ipv6.rt6_stats->fib_rt_entries,
3757                    net->ipv6.rt6_stats->fib_rt_cache,
3758                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3759                    net->ipv6.rt6_stats->fib_discarded_routes);
3760
3761         return 0;
3762 }
3763
3764 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3765 {
3766         return single_open_net(inode, file, rt6_stats_seq_show);
3767 }
3768
3769 static const struct file_operations rt6_stats_seq_fops = {
3770         .owner   = THIS_MODULE,
3771         .open    = rt6_stats_seq_open,
3772         .read    = seq_read,
3773         .llseek  = seq_lseek,
3774         .release = single_release_net,
3775 };
3776 #endif  /* CONFIG_PROC_FS */
3777
3778 #ifdef CONFIG_SYSCTL
3779
3780 static
3781 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3782                               void __user *buffer, size_t *lenp, loff_t *ppos)
3783 {
3784         struct net *net;
3785         int delay;
3786         if (!write)
3787                 return -EINVAL;
3788
3789         net = (struct net *)ctl->extra1;
3790         delay = net->ipv6.sysctl.flush_delay;
3791         proc_dointvec(ctl, write, buffer, lenp, ppos);
3792         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3793         return 0;
3794 }
3795
3796 struct ctl_table ipv6_route_table_template[] = {
3797         {
3798                 .procname       =       "flush",
3799                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3800                 .maxlen         =       sizeof(int),
3801                 .mode           =       0200,
3802                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3803         },
3804         {
3805                 .procname       =       "gc_thresh",
3806                 .data           =       &ip6_dst_ops_template.gc_thresh,
3807                 .maxlen         =       sizeof(int),
3808                 .mode           =       0644,
3809                 .proc_handler   =       proc_dointvec,
3810         },
3811         {
3812                 .procname       =       "max_size",
3813                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3814                 .maxlen         =       sizeof(int),
3815                 .mode           =       0644,
3816                 .proc_handler   =       proc_dointvec,
3817         },
3818         {
3819                 .procname       =       "gc_min_interval",
3820                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3821                 .maxlen         =       sizeof(int),
3822                 .mode           =       0644,
3823                 .proc_handler   =       proc_dointvec_jiffies,
3824         },
3825         {
3826                 .procname       =       "gc_timeout",
3827                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3828                 .maxlen         =       sizeof(int),
3829                 .mode           =       0644,
3830                 .proc_handler   =       proc_dointvec_jiffies,
3831         },
3832         {
3833                 .procname       =       "gc_interval",
3834                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3835                 .maxlen         =       sizeof(int),
3836                 .mode           =       0644,
3837                 .proc_handler   =       proc_dointvec_jiffies,
3838         },
3839         {
3840                 .procname       =       "gc_elasticity",
3841                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3842                 .maxlen         =       sizeof(int),
3843                 .mode           =       0644,
3844                 .proc_handler   =       proc_dointvec,
3845         },
3846         {
3847                 .procname       =       "mtu_expires",
3848                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3849                 .maxlen         =       sizeof(int),
3850                 .mode           =       0644,
3851                 .proc_handler   =       proc_dointvec_jiffies,
3852         },
3853         {
3854                 .procname       =       "min_adv_mss",
3855                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3856                 .maxlen         =       sizeof(int),
3857                 .mode           =       0644,
3858                 .proc_handler   =       proc_dointvec,
3859         },
3860         {
3861                 .procname       =       "gc_min_interval_ms",
3862                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3863                 .maxlen         =       sizeof(int),
3864                 .mode           =       0644,
3865                 .proc_handler   =       proc_dointvec_ms_jiffies,
3866         },
3867         { }
3868 };
3869
3870 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3871 {
3872         struct ctl_table *table;
3873
3874         table = kmemdup(ipv6_route_table_template,
3875                         sizeof(ipv6_route_table_template),
3876                         GFP_KERNEL);
3877
3878         if (table) {
3879                 table[0].data = &net->ipv6.sysctl.flush_delay;
3880                 table[0].extra1 = net;
3881                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3882                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3883                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3884                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3885                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3886                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3887                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3888                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3889                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3890
3891                 /* Don't export sysctls to unprivileged users */
3892                 if (net->user_ns != &init_user_ns)
3893                         table[0].procname = NULL;
3894         }
3895
3896         return table;
3897 }
3898 #endif
3899
3900 static int __net_init ip6_route_net_init(struct net *net)
3901 {
3902         int ret = -ENOMEM;
3903
3904         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3905                sizeof(net->ipv6.ip6_dst_ops));
3906
3907         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3908                 goto out_ip6_dst_ops;
3909
3910         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3911                                            sizeof(*net->ipv6.ip6_null_entry),
3912                                            GFP_KERNEL);
3913         if (!net->ipv6.ip6_null_entry)
3914                 goto out_ip6_dst_entries;
3915         net->ipv6.ip6_null_entry->dst.path =
3916                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3917         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3918         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3919                          ip6_template_metrics, true);
3920
3921 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3922         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3923                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3924                                                GFP_KERNEL);
3925         if (!net->ipv6.ip6_prohibit_entry)
3926                 goto out_ip6_null_entry;
3927         net->ipv6.ip6_prohibit_entry->dst.path =
3928                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3929         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3930         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3931                          ip6_template_metrics, true);
3932
3933         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3934                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3935                                                GFP_KERNEL);
3936         if (!net->ipv6.ip6_blk_hole_entry)
3937                 goto out_ip6_prohibit_entry;
3938         net->ipv6.ip6_blk_hole_entry->dst.path =
3939                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3940         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3941         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3942                          ip6_template_metrics, true);
3943 #endif
3944
3945         net->ipv6.sysctl.flush_delay = 0;
3946         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3947         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3948         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3949         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3950         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3951         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3952         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3953
3954         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3955
3956         ret = 0;
3957 out:
3958         return ret;
3959
3960 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3961 out_ip6_prohibit_entry:
3962         kfree(net->ipv6.ip6_prohibit_entry);
3963 out_ip6_null_entry:
3964         kfree(net->ipv6.ip6_null_entry);
3965 #endif
3966 out_ip6_dst_entries:
3967         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3968 out_ip6_dst_ops:
3969         goto out;
3970 }
3971
3972 static void __net_exit ip6_route_net_exit(struct net *net)
3973 {
3974         kfree(net->ipv6.ip6_null_entry);
3975 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3976         kfree(net->ipv6.ip6_prohibit_entry);
3977         kfree(net->ipv6.ip6_blk_hole_entry);
3978 #endif
3979         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3980 }
3981
3982 static int __net_init ip6_route_net_init_late(struct net *net)
3983 {
3984 #ifdef CONFIG_PROC_FS
3985         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3986         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3987 #endif
3988         return 0;
3989 }
3990
3991 static void __net_exit ip6_route_net_exit_late(struct net *net)
3992 {
3993 #ifdef CONFIG_PROC_FS
3994         remove_proc_entry("ipv6_route", net->proc_net);
3995         remove_proc_entry("rt6_stats", net->proc_net);
3996 #endif
3997 }
3998
3999 static struct pernet_operations ip6_route_net_ops = {
4000         .init = ip6_route_net_init,
4001         .exit = ip6_route_net_exit,
4002 };
4003
4004 static int __net_init ipv6_inetpeer_init(struct net *net)
4005 {
4006         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
4007
4008         if (!bp)
4009                 return -ENOMEM;
4010         inet_peer_base_init(bp);
4011         net->ipv6.peers = bp;
4012         return 0;
4013 }
4014
4015 static void __net_exit ipv6_inetpeer_exit(struct net *net)
4016 {
4017         struct inet_peer_base *bp = net->ipv6.peers;
4018
4019         net->ipv6.peers = NULL;
4020         inetpeer_invalidate_tree(bp);
4021         kfree(bp);
4022 }
4023
4024 static struct pernet_operations ipv6_inetpeer_ops = {
4025         .init   =       ipv6_inetpeer_init,
4026         .exit   =       ipv6_inetpeer_exit,
4027 };
4028
4029 static struct pernet_operations ip6_route_net_late_ops = {
4030         .init = ip6_route_net_init_late,
4031         .exit = ip6_route_net_exit_late,
4032 };
4033
4034 static struct notifier_block ip6_route_dev_notifier = {
4035         .notifier_call = ip6_route_dev_notify,
4036         .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
4037 };
4038
4039 void __init ip6_route_init_special_entries(void)
4040 {
4041         /* Registering of the loopback is done before this portion of code,
4042          * the loopback reference in rt6_info will not be taken, do it
4043          * manually for init_net */
4044         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
4045         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4046   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4047         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
4048         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4049         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
4050         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4051   #endif
4052 }
4053
4054 int __init ip6_route_init(void)
4055 {
4056         int ret;
4057         int cpu;
4058
4059         ret = -ENOMEM;
4060         ip6_dst_ops_template.kmem_cachep =
4061                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
4062                                   SLAB_HWCACHE_ALIGN, NULL);
4063         if (!ip6_dst_ops_template.kmem_cachep)
4064                 goto out;
4065
4066         ret = dst_entries_init(&ip6_dst_blackhole_ops);
4067         if (ret)
4068                 goto out_kmem_cache;
4069
4070         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
4071         if (ret)
4072                 goto out_dst_entries;
4073
4074         ret = register_pernet_subsys(&ip6_route_net_ops);
4075         if (ret)
4076                 goto out_register_inetpeer;
4077
4078         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
4079
4080         ret = fib6_init();
4081         if (ret)
4082                 goto out_register_subsys;
4083
4084         ret = xfrm6_init();
4085         if (ret)
4086                 goto out_fib6_init;
4087
4088         ret = fib6_rules_init();
4089         if (ret)
4090                 goto xfrm6_init;
4091
4092         ret = register_pernet_subsys(&ip6_route_net_late_ops);
4093         if (ret)
4094                 goto fib6_rules_init;
4095
4096         ret = -ENOBUFS;
4097         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
4098             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
4099             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
4100                 goto out_register_late_subsys;
4101
4102         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
4103         if (ret)
4104                 goto out_register_late_subsys;
4105
4106         for_each_possible_cpu(cpu) {
4107                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
4108
4109                 INIT_LIST_HEAD(&ul->head);
4110                 spin_lock_init(&ul->lock);
4111         }
4112
4113 out:
4114         return ret;
4115
4116 out_register_late_subsys:
4117         unregister_pernet_subsys(&ip6_route_net_late_ops);
4118 fib6_rules_init:
4119         fib6_rules_cleanup();
4120 xfrm6_init:
4121         xfrm6_fini();
4122 out_fib6_init:
4123         fib6_gc_cleanup();
4124 out_register_subsys:
4125         unregister_pernet_subsys(&ip6_route_net_ops);
4126 out_register_inetpeer:
4127         unregister_pernet_subsys(&ipv6_inetpeer_ops);
4128 out_dst_entries:
4129         dst_entries_destroy(&ip6_dst_blackhole_ops);
4130 out_kmem_cache:
4131         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4132         goto out;
4133 }
4134
4135 void ip6_route_cleanup(void)
4136 {
4137         unregister_netdevice_notifier(&ip6_route_dev_notifier);
4138         unregister_pernet_subsys(&ip6_route_net_late_ops);
4139         fib6_rules_cleanup();
4140         xfrm6_fini();
4141         fib6_gc_cleanup();
4142         unregister_pernet_subsys(&ipv6_inetpeer_ops);
4143         unregister_pernet_subsys(&ip6_route_net_ops);
4144         dst_entries_destroy(&ip6_dst_blackhole_ops);
4145         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4146 }