net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <[email protected]>
   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->rt6i_uncached_list = ul;
 132
 133         spin_lock_bh(&ul->lock);
 134         list_add_tail(&rt->rt6i_uncached, &ul->head);
 135         spin_unlock_bh(&ul->lock);
 136 }
 137
 138 static void rt6_uncached_list_del(struct rt6_info *rt)
 139 {
 140         if (!list_empty(&rt->rt6i_uncached)) {
 141                 struct uncached_list *ul = rt->rt6i_uncached_list;
 142
 143                 spin_lock_bh(&ul->lock);
 144                 list_del(&rt->rt6i_uncached);
 145                 spin_unlock_bh(&ul->lock);
 146         }
 147 }
 148
 149 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
 150 {
 151         struct net_device *loopback_dev = net->loopback_dev;
 152         int cpu;
 153
 154         if (dev == loopback_dev)
 155                 return;
 156
 157         for_each_possible_cpu(cpu) {
 158                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
 159                 struct rt6_info *rt;
 160
 161                 spin_lock_bh(&ul->lock);
 162                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
 163                         struct inet6_dev *rt_idev = rt->rt6i_idev;
 164                         struct net_device *rt_dev = rt->dst.dev;
 165
 166                         if (rt_idev->dev == dev) {
 167                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
 168                                 in6_dev_put(rt_idev);
 169                         }
 170
 171                         if (rt_dev == dev) {
 172                                 rt->dst.dev = loopback_dev;
 173                                 dev_hold(rt->dst.dev);
 174                                 dev_put(rt_dev);
 175                         }
 176                 }
 177                 spin_unlock_bh(&ul->lock);
 178         }
 179 }
 180
 181 static u32 *rt6_pcpu_cow_metrics(struct rt6_info *rt)
 182 {
 183         return dst_metrics_write_ptr(rt->dst.from);
 184 }
 185
 186 static u32 *ipv6_cow_metrics(struct dst_entry *dst, unsigned long old)
 187 {
 188         struct rt6_info *rt = (struct rt6_info *)dst;
 189
 190         if (rt->rt6i_flags & RTF_PCPU)
 191                 return rt6_pcpu_cow_metrics(rt);
 192         else if (rt->rt6i_flags & RTF_CACHE)
 193                 return NULL;
 194         else
 195                 return dst_cow_metrics_generic(dst, old);
 196 }
 197
 198 static inline const void *choose_neigh_daddr(struct rt6_info *rt,
 199                                              struct sk_buff *skb,
 200                                              const void *daddr)
 201 {
 202         struct in6_addr *p = &rt->rt6i_gateway;
 203
 204         if (!ipv6_addr_any(p))
 205                 return (const void *) p;
 206         else if (skb)
 207                 return &ipv6_hdr(skb)->daddr;
 208         return daddr;
 209 }
 210
 211 static struct neighbour *ip6_neigh_lookup(const struct dst_entry *dst,
 212                                           struct sk_buff *skb,
 213                                           const void *daddr)
 214 {
 215         struct rt6_info *rt = (struct rt6_info *) dst;
 216         struct neighbour *n;
 217
 218         daddr = choose_neigh_daddr(rt, skb, daddr);
 219         n = __ipv6_neigh_lookup(dst->dev, daddr);
 220         if (n)
 221                 return n;
 222         return neigh_create(&nd_tbl, daddr, dst->dev);
 223 }
 224
 225 static void ip6_confirm_neigh(const struct dst_entry *dst, const void *daddr)
 226 {
 227         struct net_device *dev = dst->dev;
 228         struct rt6_info *rt = (struct rt6_info *)dst;
 229
 230         daddr = choose_neigh_daddr(rt, NULL, daddr);
 231         if (!daddr)
 232                 return;
 233         if (dev->flags & (IFF_NOARP | IFF_LOOPBACK))
 234                 return;
 235         if (ipv6_addr_is_multicast((const struct in6_addr *)daddr))
 236                 return;
 237         __ipv6_confirm_neigh(dev, daddr);
 238 }
 239
 240 static struct dst_ops ip6_dst_ops_template = {
 241         .family                 =       AF_INET6,
 242         .gc                     =       ip6_dst_gc,
 243         .gc_thresh              =       1024,
 244         .check                  =       ip6_dst_check,
 245         .default_advmss         =       ip6_default_advmss,
 246         .mtu                    =       ip6_mtu,
 247         .cow_metrics            =       ipv6_cow_metrics,
 248         .destroy                =       ip6_dst_destroy,
 249         .ifdown                 =       ip6_dst_ifdown,
 250         .negative_advice        =       ip6_negative_advice,
 251         .link_failure           =       ip6_link_failure,
 252         .update_pmtu            =       ip6_rt_update_pmtu,
 253         .redirect               =       rt6_do_redirect,
 254         .local_out              =       __ip6_local_out,
 255         .neigh_lookup           =       ip6_neigh_lookup,
 256         .confirm_neigh          =       ip6_confirm_neigh,
 257 };
 258
 259 static unsigned int ip6_blackhole_mtu(const struct dst_entry *dst)
 260 {
 261         unsigned int mtu = dst_metric_raw(dst, RTAX_MTU);
 262
 263         return mtu ? : dst->dev->mtu;
 264 }
 265
 266 static void ip6_rt_blackhole_update_pmtu(struct dst_entry *dst, struct sock *sk,
 267                                          struct sk_buff *skb, u32 mtu)
 268 {
 269 }
 270
 271 static void ip6_rt_blackhole_redirect(struct dst_entry *dst, struct sock *sk,
 272                                       struct sk_buff *skb)
 273 {
 274 }
 275
 276 static struct dst_ops ip6_dst_blackhole_ops = {
 277         .family                 =       AF_INET6,
 278         .destroy                =       ip6_dst_destroy,
 279         .check                  =       ip6_dst_check,
 280         .mtu                    =       ip6_blackhole_mtu,
 281         .default_advmss         =       ip6_default_advmss,
 282         .update_pmtu            =       ip6_rt_blackhole_update_pmtu,
 283         .redirect               =       ip6_rt_blackhole_redirect,
 284         .cow_metrics            =       dst_cow_metrics_generic,
 285         .neigh_lookup           =       ip6_neigh_lookup,
 286 };
 287
 288 static const u32 ip6_template_metrics[RTAX_MAX] = {
 289         [RTAX_HOPLIMIT - 1] = 0,
 290 };
 291
 292 static const struct rt6_info ip6_null_entry_template = {
 293         .dst = {
 294                 .__refcnt       = ATOMIC_INIT(1),
 295                 .__use          = 1,
 296                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 297                 .error          = -ENETUNREACH,
 298                 .input          = ip6_pkt_discard,
 299                 .output         = ip6_pkt_discard_out,
 300         },
 301         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 302         .rt6i_protocol  = RTPROT_KERNEL,
 303         .rt6i_metric    = ~(u32) 0,
 304         .rt6i_ref       = ATOMIC_INIT(1),
 305 };
 306
 307 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
 308
 309 static const struct rt6_info ip6_prohibit_entry_template = {
 310         .dst = {
 311                 .__refcnt       = ATOMIC_INIT(1),
 312                 .__use          = 1,
 313                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 314                 .error          = -EACCES,
 315                 .input          = ip6_pkt_prohibit,
 316                 .output         = ip6_pkt_prohibit_out,
 317         },
 318         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 319         .rt6i_protocol  = RTPROT_KERNEL,
 320         .rt6i_metric    = ~(u32) 0,
 321         .rt6i_ref       = ATOMIC_INIT(1),
 322 };
 323
 324 static const struct rt6_info ip6_blk_hole_entry_template = {
 325         .dst = {
 326                 .__refcnt       = ATOMIC_INIT(1),
 327                 .__use          = 1,
 328                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
 329                 .error          = -EINVAL,
 330                 .input          = dst_discard,
 331                 .output         = dst_discard_out,
 332         },
 333         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
 334         .rt6i_protocol  = RTPROT_KERNEL,
 335         .rt6i_metric    = ~(u32) 0,
 336         .rt6i_ref       = ATOMIC_INIT(1),
 337 };
 338
 339 #endif
 340
 341 static void rt6_info_init(struct rt6_info *rt)
 342 {
 343         struct dst_entry *dst = &rt->dst;
 344
 345         memset(dst + 1, 0, sizeof(*rt) - sizeof(*dst));
 346         INIT_LIST_HEAD(&rt->rt6i_siblings);
 347         INIT_LIST_HEAD(&rt->rt6i_uncached);
 348 }
 349
 350 /* allocate dst with ip6_dst_ops */
 351 static struct rt6_info *__ip6_dst_alloc(struct net *net,
 352                                         struct net_device *dev,
 353                                         int flags)
 354 {
 355         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
 356                                         1, DST_OBSOLETE_FORCE_CHK, flags);
 357
 358         if (rt)
 359                 rt6_info_init(rt);
 360
 361         return rt;
 362 }
 363
 364 struct rt6_info *ip6_dst_alloc(struct net *net,
 365                                struct net_device *dev,
 366                                int flags)
 367 {
 368         struct rt6_info *rt = __ip6_dst_alloc(net, dev, flags);
 369
 370         if (rt) {
 371                 rt->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, GFP_ATOMIC);
 372                 if (rt->rt6i_pcpu) {
 373                         int cpu;
 374
 375                         for_each_possible_cpu(cpu) {
 376                                 struct rt6_info **p;
 377
 378                                 p = per_cpu_ptr(rt->rt6i_pcpu, cpu);
 379                                 /* no one shares rt */
 380                                 *p =  NULL;
 381                         }
 382                 } else {
 383                         dst_release_immediate(&rt->dst);
 384                         return NULL;
 385                 }
 386         }
 387
 388         return rt;
 389 }
 390 EXPORT_SYMBOL(ip6_dst_alloc);
 391
 392 static void ip6_dst_destroy(struct dst_entry *dst)
 393 {
 394         struct rt6_info *rt = (struct rt6_info *)dst;
 395         struct dst_entry *from = dst->from;
 396         struct inet6_dev *idev;
 397
 398         dst_destroy_metrics_generic(dst);
 399         free_percpu(rt->rt6i_pcpu);
 400         rt6_uncached_list_del(rt);
 401
 402         idev = rt->rt6i_idev;
 403         if (idev) {
 404                 rt->rt6i_idev = NULL;
 405                 in6_dev_put(idev);
 406         }
 407
 408         dst->from = NULL;
 409         dst_release(from);
 410 }
 411
 412 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
 413                            int how)
 414 {
 415         struct rt6_info *rt = (struct rt6_info *)dst;
 416         struct inet6_dev *idev = rt->rt6i_idev;
 417         struct net_device *loopback_dev =
 418                 dev_net(dev)->loopback_dev;
 419
 420         if (dev != loopback_dev) {
 421                 if (idev && idev->dev == dev) {
 422                         struct inet6_dev *loopback_idev =
 423                                 in6_dev_get(loopback_dev);
 424                         if (loopback_idev) {
 425                                 rt->rt6i_idev = loopback_idev;
 426                                 in6_dev_put(idev);
 427                         }
 428                 }
 429         }
 430 }
 431
 432 static bool __rt6_check_expired(const struct rt6_info *rt)
 433 {
 434         if (rt->rt6i_flags & RTF_EXPIRES)
 435                 return time_after(jiffies, rt->dst.expires);
 436         else
 437                 return false;
 438 }
 439
 440 static bool rt6_check_expired(const struct rt6_info *rt)
 441 {
 442         if (rt->rt6i_flags & RTF_EXPIRES) {
 443                 if (time_after(jiffies, rt->dst.expires))
 444                         return true;
 445         } else if (rt->dst.from) {
 446                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 447         }
 448         return false;
 449 }
 450
 451 /* Multipath route selection:
 452  *   Hash based function using packet header and flowlabel.
 453  * Adapted from fib_info_hashfn()
 454  */
 455 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 456                                const struct flowi6 *fl6)
 457 {
 458         return get_hash_from_flowi6(fl6) % candidate_count;
 459 }
 460
 461 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 462                                              struct flowi6 *fl6, int oif,
 463                                              int strict)
 464 {
 465         struct rt6_info *sibling, *next_sibling;
 466         int route_choosen;
 467
 468         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 469         /* Don't change the route, if route_choosen == 0
 470          * (siblings does not include ourself)
 471          */
 472         if (route_choosen)
 473                 list_for_each_entry_safe(sibling, next_sibling,
 474                                 &match->rt6i_siblings, rt6i_siblings) {
 475                         route_choosen--;
 476                         if (route_choosen == 0) {
 477                                 if (rt6_score_route(sibling, oif, strict) < 0)
 478                                         break;
 479                                 match = sibling;
 480                                 break;
 481                         }
 482                 }
 483         return match;
 484 }
 485
 486 /*
 487  *      Route lookup. Any table->tb6_lock is implied.
 488  */
 489
 490 static inline struct rt6_info *rt6_device_match(struct net *net,
 491                                                     struct rt6_info *rt,
 492                                                     const struct in6_addr *saddr,
 493                                                     int oif,
 494                                                     int flags)
 495 {
 496         struct rt6_info *local = NULL;
 497         struct rt6_info *sprt;
 498
 499         if (!oif && ipv6_addr_any(saddr))
 500                 goto out;
 501
 502         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 503                 struct net_device *dev = sprt->dst.dev;
 504
 505                 if (oif) {
 506                         if (dev->ifindex == oif)
 507                                 return sprt;
 508                         if (dev->flags & IFF_LOOPBACK) {
 509                                 if (!sprt->rt6i_idev ||
 510                                     sprt->rt6i_idev->dev->ifindex != oif) {
 511                                         if (flags & RT6_LOOKUP_F_IFACE)
 512                                                 continue;
 513                                         if (local &&
 514                                             local->rt6i_idev->dev->ifindex == oif)
 515                                                 continue;
 516                                 }
 517                                 local = sprt;
 518                         }
 519                 } else {
 520                         if (ipv6_chk_addr(net, saddr, dev,
 521                                           flags & RT6_LOOKUP_F_IFACE))
 522                                 return sprt;
 523                 }
 524         }
 525
 526         if (oif) {
 527                 if (local)
 528                         return local;
 529
 530                 if (flags & RT6_LOOKUP_F_IFACE)
 531                         return net->ipv6.ip6_null_entry;
 532         }
 533 out:
 534         return rt;
 535 }
 536
 537 #ifdef CONFIG_IPV6_ROUTER_PREF
 538 struct __rt6_probe_work {
 539         struct work_struct work;
 540         struct in6_addr target;
 541         struct net_device *dev;
 542 };
 543
 544 static void rt6_probe_deferred(struct work_struct *w)
 545 {
 546         struct in6_addr mcaddr;
 547         struct __rt6_probe_work *work =
 548                 container_of(w, struct __rt6_probe_work, work);
 549
 550         addrconf_addr_solict_mult(&work->target, &mcaddr);
 551         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
 552         dev_put(work->dev);
 553         kfree(work);
 554 }
 555
 556 static void rt6_probe(struct rt6_info *rt)
 557 {
 558         struct __rt6_probe_work *work;
 559         struct neighbour *neigh;
 560         /*
 561          * Okay, this does not seem to be appropriate
 562          * for now, however, we need to check if it
 563          * is really so; aka Router Reachability Probing.
 564          *
 565          * Router Reachability Probe MUST be rate-limited
 566          * to no more than one per minute.
 567          */
 568         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 569                 return;
 570         rcu_read_lock_bh();
 571         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 572         if (neigh) {
 573                 if (neigh->nud_state & NUD_VALID)
 574                         goto out;
 575
 576                 work = NULL;
 577                 write_lock(&neigh->lock);
 578                 if (!(neigh->nud_state & NUD_VALID) &&
 579                     time_after(jiffies,
 580                                neigh->updated +
 581                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 582                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 583                         if (work)
 584                                 __neigh_set_probe_once(neigh);
 585                 }
 586                 write_unlock(&neigh->lock);
 587         } else {
 588                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 589         }
 590
 591         if (work) {
 592                 INIT_WORK(&work->work, rt6_probe_deferred);
 593                 work->target = rt->rt6i_gateway;
 594                 dev_hold(rt->dst.dev);
 595                 work->dev = rt->dst.dev;
 596                 schedule_work(&work->work);
 597         }
 598
 599 out:
 600         rcu_read_unlock_bh();
 601 }
 602 #else
 603 static inline void rt6_probe(struct rt6_info *rt)
 604 {
 605 }
 606 #endif
 607
 608 /*
 609  * Default Router Selection (RFC 2461 6.3.6)
 610  */
 611 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 612 {
 613         struct net_device *dev = rt->dst.dev;
 614         if (!oif || dev->ifindex == oif)
 615                 return 2;
 616         if ((dev->flags & IFF_LOOPBACK) &&
 617             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 618                 return 1;
 619         return 0;
 620 }
 621
 622 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 623 {
 624         struct neighbour *neigh;
 625         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 626
 627         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 628             !(rt->rt6i_flags & RTF_GATEWAY))
 629                 return RT6_NUD_SUCCEED;
 630
 631         rcu_read_lock_bh();
 632         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 633         if (neigh) {
 634                 read_lock(&neigh->lock);
 635                 if (neigh->nud_state & NUD_VALID)
 636                         ret = RT6_NUD_SUCCEED;
 637 #ifdef CONFIG_IPV6_ROUTER_PREF
 638                 else if (!(neigh->nud_state & NUD_FAILED))
 639                         ret = RT6_NUD_SUCCEED;
 640                 else
 641                         ret = RT6_NUD_FAIL_PROBE;
 642 #endif
 643                 read_unlock(&neigh->lock);
 644         } else {
 645                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 646                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 647         }
 648         rcu_read_unlock_bh();
 649
 650         return ret;
 651 }
 652
 653 static int rt6_score_route(struct rt6_info *rt, int oif,
 654                            int strict)
 655 {
 656         int m;
 657
 658         m = rt6_check_dev(rt, oif);
 659         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 660                 return RT6_NUD_FAIL_HARD;
 661 #ifdef CONFIG_IPV6_ROUTER_PREF
 662         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 663 #endif
 664         if (strict & RT6_LOOKUP_F_REACHABLE) {
 665                 int n = rt6_check_neigh(rt);
 666                 if (n < 0)
 667                         return n;
 668         }
 669         return m;
 670 }
 671
 672 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 673                                    int *mpri, struct rt6_info *match,
 674                                    bool *do_rr)
 675 {
 676         int m;
 677         bool match_do_rr = false;
 678         struct inet6_dev *idev = rt->rt6i_idev;
 679         struct net_device *dev = rt->dst.dev;
 680
 681         if (dev && !netif_carrier_ok(dev) &&
 682             idev->cnf.ignore_routes_with_linkdown &&
 683             !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
 684                 goto out;
 685
 686         if (rt6_check_expired(rt))
 687                 goto out;
 688
 689         m = rt6_score_route(rt, oif, strict);
 690         if (m == RT6_NUD_FAIL_DO_RR) {
 691                 match_do_rr = true;
 692                 m = 0; /* lowest valid score */
 693         } else if (m == RT6_NUD_FAIL_HARD) {
 694                 goto out;
 695         }
 696
 697         if (strict & RT6_LOOKUP_F_REACHABLE)
 698                 rt6_probe(rt);
 699
 700         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 701         if (m > *mpri) {
 702                 *do_rr = match_do_rr;
 703                 *mpri = m;
 704                 match = rt;
 705         }
 706 out:
 707         return match;
 708 }
 709
 710 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 711                                      struct rt6_info *rr_head,
 712                                      u32 metric, int oif, int strict,
 713                                      bool *do_rr)
 714 {
 715         struct rt6_info *rt, *match, *cont;
 716         int mpri = -1;
 717
 718         match = NULL;
 719         cont = NULL;
 720         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 721                 if (rt->rt6i_metric != metric) {
 722                         cont = rt;
 723                         break;
 724                 }
 725
 726                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 727         }
 728
 729         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 730                 if (rt->rt6i_metric != metric) {
 731                         cont = rt;
 732                         break;
 733                 }
 734
 735                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 736         }
 737
 738         if (match || !cont)
 739                 return match;
 740
 741         for (rt = cont; rt; rt = rt->dst.rt6_next)
 742                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 743
 744         return match;
 745 }
 746
 747 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 748 {
 749         struct rt6_info *match, *rt0;
 750         struct net *net;
 751         bool do_rr = false;
 752
 753         rt0 = fn->rr_ptr;
 754         if (!rt0)
 755                 fn->rr_ptr = rt0 = fn->leaf;
 756
 757         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 758                              &do_rr);
 759
 760         if (do_rr) {
 761                 struct rt6_info *next = rt0->dst.rt6_next;
 762
 763                 /* no entries matched; do round-robin */
 764                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 765                         next = fn->leaf;
 766
 767                 if (next != rt0)
 768                         fn->rr_ptr = next;
 769         }
 770
 771         net = dev_net(rt0->dst.dev);
 772         return match ? match : net->ipv6.ip6_null_entry;
 773 }
 774
 775 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 776 {
 777         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 778 }
 779
 780 #ifdef CONFIG_IPV6_ROUTE_INFO
 781 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 782                   const struct in6_addr *gwaddr)
 783 {
 784         struct net *net = dev_net(dev);
 785         struct route_info *rinfo = (struct route_info *) opt;
 786         struct in6_addr prefix_buf, *prefix;
 787         unsigned int pref;
 788         unsigned long lifetime;
 789         struct rt6_info *rt;
 790
 791         if (len < sizeof(struct route_info)) {
 792                 return -EINVAL;
 793         }
 794
 795         /* Sanity check for prefix_len and length */
 796         if (rinfo->length > 3) {
 797                 return -EINVAL;
 798         } else if (rinfo->prefix_len > 128) {
 799                 return -EINVAL;
 800         } else if (rinfo->prefix_len > 64) {
 801                 if (rinfo->length < 2) {
 802                         return -EINVAL;
 803                 }
 804         } else if (rinfo->prefix_len > 0) {
 805                 if (rinfo->length < 1) {
 806                         return -EINVAL;
 807                 }
 808         }
 809
 810         pref = rinfo->route_pref;
 811         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 812                 return -EINVAL;
 813
 814         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 815
 816         if (rinfo->length == 3)
 817                 prefix = (struct in6_addr *)rinfo->prefix;
 818         else {
 819                 /* this function is safe */
 820                 ipv6_addr_prefix(&prefix_buf,
 821                                  (struct in6_addr *)rinfo->prefix,
 822                                  rinfo->prefix_len);
 823                 prefix = &prefix_buf;
 824         }
 825
 826         if (rinfo->prefix_len == 0)
 827                 rt = rt6_get_dflt_router(gwaddr, dev);
 828         else
 829                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 830                                         gwaddr, dev);
 831
 832         if (rt && !lifetime) {
 833                 ip6_del_rt(rt);
 834                 rt = NULL;
 835         }
 836
 837         if (!rt && lifetime)
 838                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 839                                         dev, pref);
 840         else if (rt)
 841                 rt->rt6i_flags = RTF_ROUTEINFO |
 842                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 843
 844         if (rt) {
 845                 if (!addrconf_finite_timeout(lifetime))
 846                         rt6_clean_expires(rt);
 847                 else
 848                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 849
 850                 ip6_rt_put(rt);
 851         }
 852         return 0;
 853 }
 854 #endif
 855
 856 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 857                                         struct in6_addr *saddr)
 858 {
 859         struct fib6_node *pn;
 860         while (1) {
 861                 if (fn->fn_flags & RTN_TL_ROOT)
 862                         return NULL;
 863                 pn = fn->parent;
 864                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 865                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 866                 else
 867                         fn = pn;
 868                 if (fn->fn_flags & RTN_RTINFO)
 869                         return fn;
 870         }
 871 }
 872
 873 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 874                                              struct fib6_table *table,
 875                                              struct flowi6 *fl6, int flags)
 876 {
 877         struct fib6_node *fn;
 878         struct rt6_info *rt;
 879
 880         read_lock_bh(&table->tb6_lock);
 881         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 882 restart:
 883         rt = fn->leaf;
 884         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 885         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 886                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 887         if (rt == net->ipv6.ip6_null_entry) {
 888                 fn = fib6_backtrack(fn, &fl6->saddr);
 889                 if (fn)
 890                         goto restart;
 891         }
 892         dst_use(&rt->dst, jiffies);
 893         read_unlock_bh(&table->tb6_lock);
 894
 895         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
 896
 897         return rt;
 898
 899 }
 900
 901 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 902                                     int flags)
 903 {
 904         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 905 }
 906 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 907
 908 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 909                             const struct in6_addr *saddr, int oif, int strict)
 910 {
 911         struct flowi6 fl6 = {
 912                 .flowi6_oif = oif,
 913                 .daddr = *daddr,
 914         };
 915         struct dst_entry *dst;
 916         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 917
 918         if (saddr) {
 919                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 920                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 921         }
 922
 923         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 924         if (dst->error == 0)
 925                 return (struct rt6_info *) dst;
 926
 927         dst_release(dst);
 928
 929         return NULL;
 930 }
 931 EXPORT_SYMBOL(rt6_lookup);
 932
 933 /* ip6_ins_rt is called with FREE table->tb6_lock.
 934  * It takes new route entry, the addition fails by any reason the
 935  * route is released.
 936  * Caller must hold dst before calling it.
 937  */
 938
 939 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 940                         struct mx6_config *mxc,
 941                         struct netlink_ext_ack *extack)
 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, extack);
 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         /* Hold dst to account for the reference from the fib6 tree */
 960         dst_hold(&rt->dst);
 961         return __ip6_ins_rt(rt, &info, &mxc, NULL);
 962 }
 963
 964 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 965                                            const struct in6_addr *daddr,
 966                                            const struct in6_addr *saddr)
 967 {
 968         struct rt6_info *rt;
 969
 970         /*
 971          *      Clone the route.
 972          */
 973
 974         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 975                 ort = (struct rt6_info *)ort->dst.from;
 976
 977         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 978
 979         if (!rt)
 980                 return NULL;
 981
 982         ip6_rt_copy_init(rt, ort);
 983         rt->rt6i_flags |= RTF_CACHE;
 984         rt->rt6i_metric = 0;
 985         rt->dst.flags |= DST_HOST;
 986         rt->rt6i_dst.addr = *daddr;
 987         rt->rt6i_dst.plen = 128;
 988
 989         if (!rt6_is_gw_or_nonexthop(ort)) {
 990                 if (ort->rt6i_dst.plen != 128 &&
 991                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 992                         rt->rt6i_flags |= RTF_ANYCAST;
 993 #ifdef CONFIG_IPV6_SUBTREES
 994                 if (rt->rt6i_src.plen && saddr) {
 995                         rt->rt6i_src.addr = *saddr;
 996                         rt->rt6i_src.plen = 128;
 997                 }
 998 #endif
 999         }
1000
1001         return rt;
1002 }
1003
1004 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
1005 {
1006         struct rt6_info *pcpu_rt;
1007
1008         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
1009                                   rt->dst.dev, rt->dst.flags);
1010
1011         if (!pcpu_rt)
1012                 return NULL;
1013         ip6_rt_copy_init(pcpu_rt, rt);
1014         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
1015         pcpu_rt->rt6i_flags |= RTF_PCPU;
1016         return pcpu_rt;
1017 }
1018
1019 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1020 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1021 {
1022         struct rt6_info *pcpu_rt, **p;
1023
1024         p = this_cpu_ptr(rt->rt6i_pcpu);
1025         pcpu_rt = *p;
1026
1027         if (pcpu_rt) {
1028                 dst_hold(&pcpu_rt->dst);
1029                 rt6_dst_from_metrics_check(pcpu_rt);
1030         }
1031         return pcpu_rt;
1032 }
1033
1034 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1035 {
1036         struct fib6_table *table = rt->rt6i_table;
1037         struct rt6_info *pcpu_rt, *prev, **p;
1038
1039         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1040         if (!pcpu_rt) {
1041                 struct net *net = dev_net(rt->dst.dev);
1042
1043                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1044                 return net->ipv6.ip6_null_entry;
1045         }
1046
1047         read_lock_bh(&table->tb6_lock);
1048         if (rt->rt6i_pcpu) {
1049                 p = this_cpu_ptr(rt->rt6i_pcpu);
1050                 prev = cmpxchg(p, NULL, pcpu_rt);
1051                 if (prev) {
1052                         /* If someone did it before us, return prev instead */
1053                         dst_release_immediate(&pcpu_rt->dst);
1054                         pcpu_rt = prev;
1055                 }
1056         } else {
1057                 /* rt has been removed from the fib6 tree
1058                  * before we have a chance to acquire the read_lock.
1059                  * In this case, don't brother to create a pcpu rt
1060                  * since rt is going away anyway.  The next
1061                  * dst_check() will trigger a re-lookup.
1062                  */
1063                 dst_release_immediate(&pcpu_rt->dst);
1064                 pcpu_rt = rt;
1065         }
1066         dst_hold(&pcpu_rt->dst);
1067         rt6_dst_from_metrics_check(pcpu_rt);
1068         read_unlock_bh(&table->tb6_lock);
1069         return pcpu_rt;
1070 }
1071
1072 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1073                                int oif, struct flowi6 *fl6, int flags)
1074 {
1075         struct fib6_node *fn, *saved_fn;
1076         struct rt6_info *rt;
1077         int strict = 0;
1078
1079         strict |= flags & RT6_LOOKUP_F_IFACE;
1080         strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1081         if (net->ipv6.devconf_all->forwarding == 0)
1082                 strict |= RT6_LOOKUP_F_REACHABLE;
1083
1084         read_lock_bh(&table->tb6_lock);
1085
1086         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1087         saved_fn = fn;
1088
1089         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1090                 oif = 0;
1091
1092 redo_rt6_select:
1093         rt = rt6_select(fn, oif, strict);
1094         if (rt->rt6i_nsiblings)
1095                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1096         if (rt == net->ipv6.ip6_null_entry) {
1097                 fn = fib6_backtrack(fn, &fl6->saddr);
1098                 if (fn)
1099                         goto redo_rt6_select;
1100                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1101                         /* also consider unreachable route */
1102                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1103                         fn = saved_fn;
1104                         goto redo_rt6_select;
1105                 }
1106         }
1107
1108
1109         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1110                 dst_use(&rt->dst, jiffies);
1111                 read_unlock_bh(&table->tb6_lock);
1112
1113                 rt6_dst_from_metrics_check(rt);
1114
1115                 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1116                 return rt;
1117         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1118                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1119                 /* Create a RTF_CACHE clone which will not be
1120                  * owned by the fib6 tree.  It is for the special case where
1121                  * the daddr in the skb during the neighbor look-up is different
1122                  * from the fl6->daddr used to look-up route here.
1123                  */
1124
1125                 struct rt6_info *uncached_rt;
1126
1127                 dst_use(&rt->dst, jiffies);
1128                 read_unlock_bh(&table->tb6_lock);
1129
1130                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1131                 dst_release(&rt->dst);
1132
1133                 if (uncached_rt) {
1134                         /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
1135                          * No need for another dst_hold()
1136                          */
1137                         rt6_uncached_list_add(uncached_rt);
1138                 } else {
1139                         uncached_rt = net->ipv6.ip6_null_entry;
1140                         dst_hold(&uncached_rt->dst);
1141                 }
1142
1143                 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1144                 return uncached_rt;
1145
1146         } else {
1147                 /* Get a percpu copy */
1148
1149                 struct rt6_info *pcpu_rt;
1150
1151                 rt->dst.lastuse = jiffies;
1152                 rt->dst.__use++;
1153                 pcpu_rt = rt6_get_pcpu_route(rt);
1154
1155                 if (pcpu_rt) {
1156                         read_unlock_bh(&table->tb6_lock);
1157                 } else {
1158                         /* We have to do the read_unlock first
1159                          * because rt6_make_pcpu_route() may trigger
1160                          * ip6_dst_gc() which will take the write_lock.
1161                          */
1162                         dst_hold(&rt->dst);
1163                         read_unlock_bh(&table->tb6_lock);
1164                         pcpu_rt = rt6_make_pcpu_route(rt);
1165                         dst_release(&rt->dst);
1166                 }
1167
1168                 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1169                 return pcpu_rt;
1170
1171         }
1172 }
1173 EXPORT_SYMBOL_GPL(ip6_pol_route);
1174
1175 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1176                                             struct flowi6 *fl6, int flags)
1177 {
1178         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1179 }
1180
1181 struct dst_entry *ip6_route_input_lookup(struct net *net,
1182                                          struct net_device *dev,
1183                                          struct flowi6 *fl6, int flags)
1184 {
1185         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1186                 flags |= RT6_LOOKUP_F_IFACE;
1187
1188         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1189 }
1190 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1191
1192 void ip6_route_input(struct sk_buff *skb)
1193 {
1194         const struct ipv6hdr *iph = ipv6_hdr(skb);
1195         struct net *net = dev_net(skb->dev);
1196         int flags = RT6_LOOKUP_F_HAS_SADDR;
1197         struct ip_tunnel_info *tun_info;
1198         struct flowi6 fl6 = {
1199                 .flowi6_iif = skb->dev->ifindex,
1200                 .daddr = iph->daddr,
1201                 .saddr = iph->saddr,
1202                 .flowlabel = ip6_flowinfo(iph),
1203                 .flowi6_mark = skb->mark,
1204                 .flowi6_proto = iph->nexthdr,
1205         };
1206
1207         tun_info = skb_tunnel_info(skb);
1208         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1209                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1210         skb_dst_drop(skb);
1211         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1212 }
1213
1214 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1215                                              struct flowi6 *fl6, int flags)
1216 {
1217         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1218 }
1219
1220 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1221                                          struct flowi6 *fl6, int flags)
1222 {
1223         bool any_src;
1224
1225         if (rt6_need_strict(&fl6->daddr)) {
1226                 struct dst_entry *dst;
1227
1228                 dst = l3mdev_link_scope_lookup(net, fl6);
1229                 if (dst)
1230                         return dst;
1231         }
1232
1233         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1234
1235         any_src = ipv6_addr_any(&fl6->saddr);
1236         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1237             (fl6->flowi6_oif && any_src))
1238                 flags |= RT6_LOOKUP_F_IFACE;
1239
1240         if (!any_src)
1241                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1242         else if (sk)
1243                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1244
1245         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1246 }
1247 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1248
1249 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1250 {
1251         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1252         struct net_device *loopback_dev = net->loopback_dev;
1253         struct dst_entry *new = NULL;
1254
1255         rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
1256                        DST_OBSOLETE_NONE, 0);
1257         if (rt) {
1258                 rt6_info_init(rt);
1259
1260                 new = &rt->dst;
1261                 new->__use = 1;
1262                 new->input = dst_discard;
1263                 new->output = dst_discard_out;
1264
1265                 dst_copy_metrics(new, &ort->dst);
1266
1267                 rt->rt6i_idev = in6_dev_get(loopback_dev);
1268                 rt->rt6i_gateway = ort->rt6i_gateway;
1269                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1270                 rt->rt6i_metric = 0;
1271
1272                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1273 #ifdef CONFIG_IPV6_SUBTREES
1274                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1275 #endif
1276         }
1277
1278         dst_release(dst_orig);
1279         return new ? new : ERR_PTR(-ENOMEM);
1280 }
1281
1282 /*
1283  *      Destination cache support functions
1284  */
1285
1286 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1287 {
1288         if (rt->dst.from &&
1289             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1290                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1291 }
1292
1293 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1294 {
1295         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1296                 return NULL;
1297
1298         if (rt6_check_expired(rt))
1299                 return NULL;
1300
1301         return &rt->dst;
1302 }
1303
1304 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1305 {
1306         if (!__rt6_check_expired(rt) &&
1307             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1308             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1309                 return &rt->dst;
1310         else
1311                 return NULL;
1312 }
1313
1314 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1315 {
1316         struct rt6_info *rt;
1317
1318         rt = (struct rt6_info *) dst;
1319
1320         /* All IPV6 dsts are created with ->obsolete set to the value
1321          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1322          * into this function always.
1323          */
1324
1325         rt6_dst_from_metrics_check(rt);
1326
1327         if (rt->rt6i_flags & RTF_PCPU ||
1328             (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->dst.from))
1329                 return rt6_dst_from_check(rt, cookie);
1330         else
1331                 return rt6_check(rt, cookie);
1332 }
1333
1334 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1335 {
1336         struct rt6_info *rt = (struct rt6_info *) dst;
1337
1338         if (rt) {
1339                 if (rt->rt6i_flags & RTF_CACHE) {
1340                         if (rt6_check_expired(rt)) {
1341                                 ip6_del_rt(rt);
1342                                 dst = NULL;
1343                         }
1344                 } else {
1345                         dst_release(dst);
1346                         dst = NULL;
1347                 }
1348         }
1349         return dst;
1350 }
1351
1352 static void ip6_link_failure(struct sk_buff *skb)
1353 {
1354         struct rt6_info *rt;
1355
1356         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1357
1358         rt = (struct rt6_info *) skb_dst(skb);
1359         if (rt) {
1360                 if (rt->rt6i_flags & RTF_CACHE) {
1361                         if (dst_hold_safe(&rt->dst))
1362                                 ip6_del_rt(rt);
1363                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1364                         rt->rt6i_node->fn_sernum = -1;
1365                 }
1366         }
1367 }
1368
1369 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1370 {
1371         struct net *net = dev_net(rt->dst.dev);
1372
1373         rt->rt6i_flags |= RTF_MODIFIED;
1374         rt->rt6i_pmtu = mtu;
1375         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1376 }
1377
1378 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1379 {
1380         return !(rt->rt6i_flags & RTF_CACHE) &&
1381                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1382 }
1383
1384 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1385                                  const struct ipv6hdr *iph, u32 mtu)
1386 {
1387         const struct in6_addr *daddr, *saddr;
1388         struct rt6_info *rt6 = (struct rt6_info *)dst;
1389
1390         if (rt6->rt6i_flags & RTF_LOCAL)
1391                 return;
1392
1393         if (dst_metric_locked(dst, RTAX_MTU))
1394                 return;
1395
1396         if (iph) {
1397                 daddr = &iph->daddr;
1398                 saddr = &iph->saddr;
1399         } else if (sk) {
1400                 daddr = &sk->sk_v6_daddr;
1401                 saddr = &inet6_sk(sk)->saddr;
1402         } else {
1403                 daddr = NULL;
1404                 saddr = NULL;
1405         }
1406         dst_confirm_neigh(dst, daddr);
1407         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1408         if (mtu >= dst_mtu(dst))
1409                 return;
1410
1411         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1412                 rt6_do_update_pmtu(rt6, mtu);
1413         } else if (daddr) {
1414                 struct rt6_info *nrt6;
1415
1416                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1417                 if (nrt6) {
1418                         rt6_do_update_pmtu(nrt6, mtu);
1419
1420                         /* ip6_ins_rt(nrt6) will bump the
1421                          * rt6->rt6i_node->fn_sernum
1422                          * which will fail the next rt6_check() and
1423                          * invalidate the sk->sk_dst_cache.
1424                          */
1425                         ip6_ins_rt(nrt6);
1426                         /* Release the reference taken in
1427                          * ip6_rt_cache_alloc()
1428                          */
1429                         dst_release(&nrt6->dst);
1430                 }
1431         }
1432 }
1433
1434 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1435                                struct sk_buff *skb, u32 mtu)
1436 {
1437         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1438 }
1439
1440 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1441                      int oif, u32 mark, kuid_t uid)
1442 {
1443         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1444         struct dst_entry *dst;
1445         struct flowi6 fl6;
1446
1447         memset(&fl6, 0, sizeof(fl6));
1448         fl6.flowi6_oif = oif;
1449         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1450         fl6.daddr = iph->daddr;
1451         fl6.saddr = iph->saddr;
1452         fl6.flowlabel = ip6_flowinfo(iph);
1453         fl6.flowi6_uid = uid;
1454
1455         dst = ip6_route_output(net, NULL, &fl6);
1456         if (!dst->error)
1457                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1458         dst_release(dst);
1459 }
1460 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1461
1462 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1463 {
1464         struct dst_entry *dst;
1465
1466         ip6_update_pmtu(skb, sock_net(sk), mtu,
1467                         sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
1468
1469         dst = __sk_dst_get(sk);
1470         if (!dst || !dst->obsolete ||
1471             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1472                 return;
1473
1474         bh_lock_sock(sk);
1475         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1476                 ip6_datagram_dst_update(sk, false);
1477         bh_unlock_sock(sk);
1478 }
1479 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1480
1481 /* Handle redirects */
1482 struct ip6rd_flowi {
1483         struct flowi6 fl6;
1484         struct in6_addr gateway;
1485 };
1486
1487 static struct rt6_info *__ip6_route_redirect(struct net *net,
1488                                              struct fib6_table *table,
1489                                              struct flowi6 *fl6,
1490                                              int flags)
1491 {
1492         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1493         struct rt6_info *rt;
1494         struct fib6_node *fn;
1495
1496         /* Get the "current" route for this destination and
1497          * check if the redirect has come from appropriate router.
1498          *
1499          * RFC 4861 specifies that redirects should only be
1500          * accepted if they come from the nexthop to the target.
1501          * Due to the way the routes are chosen, this notion
1502          * is a bit fuzzy and one might need to check all possible
1503          * routes.
1504          */
1505
1506         read_lock_bh(&table->tb6_lock);
1507         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1508 restart:
1509         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1510                 if (rt6_check_expired(rt))
1511                         continue;
1512                 if (rt->dst.error)
1513                         break;
1514                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1515                         continue;
1516                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1517                         continue;
1518                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1519                         continue;
1520                 break;
1521         }
1522
1523         if (!rt)
1524                 rt = net->ipv6.ip6_null_entry;
1525         else if (rt->dst.error) {
1526                 rt = net->ipv6.ip6_null_entry;
1527                 goto out;
1528         }
1529
1530         if (rt == net->ipv6.ip6_null_entry) {
1531                 fn = fib6_backtrack(fn, &fl6->saddr);
1532                 if (fn)
1533                         goto restart;
1534         }
1535
1536 out:
1537         dst_hold(&rt->dst);
1538
1539         read_unlock_bh(&table->tb6_lock);
1540
1541         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1542         return rt;
1543 };
1544
1545 static struct dst_entry *ip6_route_redirect(struct net *net,
1546                                         const struct flowi6 *fl6,
1547                                         const struct in6_addr *gateway)
1548 {
1549         int flags = RT6_LOOKUP_F_HAS_SADDR;
1550         struct ip6rd_flowi rdfl;
1551
1552         rdfl.fl6 = *fl6;
1553         rdfl.gateway = *gateway;
1554
1555         return fib6_rule_lookup(net, &rdfl.fl6,
1556                                 flags, __ip6_route_redirect);
1557 }
1558
1559 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1560                   kuid_t uid)
1561 {
1562         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1563         struct dst_entry *dst;
1564         struct flowi6 fl6;
1565
1566         memset(&fl6, 0, sizeof(fl6));
1567         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1568         fl6.flowi6_oif = oif;
1569         fl6.flowi6_mark = mark;
1570         fl6.daddr = iph->daddr;
1571         fl6.saddr = iph->saddr;
1572         fl6.flowlabel = ip6_flowinfo(iph);
1573         fl6.flowi6_uid = uid;
1574
1575         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1576         rt6_do_redirect(dst, NULL, skb);
1577         dst_release(dst);
1578 }
1579 EXPORT_SYMBOL_GPL(ip6_redirect);
1580
1581 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1582                             u32 mark)
1583 {
1584         const struct ipv6hdr *iph = ipv6_hdr(skb);
1585         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1586         struct dst_entry *dst;
1587         struct flowi6 fl6;
1588
1589         memset(&fl6, 0, sizeof(fl6));
1590         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1591         fl6.flowi6_oif = oif;
1592         fl6.flowi6_mark = mark;
1593         fl6.daddr = msg->dest;
1594         fl6.saddr = iph->daddr;
1595         fl6.flowi6_uid = sock_net_uid(net, NULL);
1596
1597         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1598         rt6_do_redirect(dst, NULL, skb);
1599         dst_release(dst);
1600 }
1601
1602 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1603 {
1604         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1605                      sk->sk_uid);
1606 }
1607 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1608
1609 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1610 {
1611         struct net_device *dev = dst->dev;
1612         unsigned int mtu = dst_mtu(dst);
1613         struct net *net = dev_net(dev);
1614
1615         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1616
1617         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1618                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1619
1620         /*
1621          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1622          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1623          * IPV6_MAXPLEN is also valid and means: "any MSS,
1624          * rely only on pmtu discovery"
1625          */
1626         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1627                 mtu = IPV6_MAXPLEN;
1628         return mtu;
1629 }
1630
1631 static unsigned int ip6_mtu(const struct dst_entry *dst)
1632 {
1633         const struct rt6_info *rt = (const struct rt6_info *)dst;
1634         unsigned int mtu = rt->rt6i_pmtu;
1635         struct inet6_dev *idev;
1636
1637         if (mtu)
1638                 goto out;
1639
1640         mtu = dst_metric_raw(dst, RTAX_MTU);
1641         if (mtu)
1642                 goto out;
1643
1644         mtu = IPV6_MIN_MTU;
1645
1646         rcu_read_lock();
1647         idev = __in6_dev_get(dst->dev);
1648         if (idev)
1649                 mtu = idev->cnf.mtu6;
1650         rcu_read_unlock();
1651
1652 out:
1653         mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1654
1655         return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1656 }
1657
1658 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1659                                   struct flowi6 *fl6)
1660 {
1661         struct dst_entry *dst;
1662         struct rt6_info *rt;
1663         struct inet6_dev *idev = in6_dev_get(dev);
1664         struct net *net = dev_net(dev);
1665
1666         if (unlikely(!idev))
1667                 return ERR_PTR(-ENODEV);
1668
1669         rt = ip6_dst_alloc(net, dev, 0);
1670         if (unlikely(!rt)) {
1671                 in6_dev_put(idev);
1672                 dst = ERR_PTR(-ENOMEM);
1673                 goto out;
1674         }
1675
1676         rt->dst.flags |= DST_HOST;
1677         rt->dst.output  = ip6_output;
1678         rt->rt6i_gateway  = fl6->daddr;
1679         rt->rt6i_dst.addr = fl6->daddr;
1680         rt->rt6i_dst.plen = 128;
1681         rt->rt6i_idev     = idev;
1682         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1683
1684         /* Add this dst into uncached_list so that rt6_ifdown() can
1685          * do proper release of the net_device
1686          */
1687         rt6_uncached_list_add(rt);
1688
1689         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1690
1691 out:
1692         return dst;
1693 }
1694
1695 static int ip6_dst_gc(struct dst_ops *ops)
1696 {
1697         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1698         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1699         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1700         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1701         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1702         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1703         int entries;
1704
1705         entries = dst_entries_get_fast(ops);
1706         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1707             entries <= rt_max_size)
1708                 goto out;
1709
1710         net->ipv6.ip6_rt_gc_expire++;
1711         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1712         entries = dst_entries_get_slow(ops);
1713         if (entries < ops->gc_thresh)
1714                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1715 out:
1716         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1717         return entries > rt_max_size;
1718 }
1719
1720 static int ip6_convert_metrics(struct mx6_config *mxc,
1721                                const struct fib6_config *cfg)
1722 {
1723         bool ecn_ca = false;
1724         struct nlattr *nla;
1725         int remaining;
1726         u32 *mp;
1727
1728         if (!cfg->fc_mx)
1729                 return 0;
1730
1731         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1732         if (unlikely(!mp))
1733                 return -ENOMEM;
1734
1735         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1736                 int type = nla_type(nla);
1737                 u32 val;
1738
1739                 if (!type)
1740                         continue;
1741                 if (unlikely(type > RTAX_MAX))
1742                         goto err;
1743
1744                 if (type == RTAX_CC_ALGO) {
1745                         char tmp[TCP_CA_NAME_MAX];
1746
1747                         nla_strlcpy(tmp, nla, sizeof(tmp));
1748                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1749                         if (val == TCP_CA_UNSPEC)
1750                                 goto err;
1751                 } else {
1752                         val = nla_get_u32(nla);
1753                 }
1754                 if (type == RTAX_HOPLIMIT && val > 255)
1755                         val = 255;
1756                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1757                         goto err;
1758
1759                 mp[type - 1] = val;
1760                 __set_bit(type - 1, mxc->mx_valid);
1761         }
1762
1763         if (ecn_ca) {
1764                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1765                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1766         }
1767
1768         mxc->mx = mp;
1769         return 0;
1770  err:
1771         kfree(mp);
1772         return -EINVAL;
1773 }
1774
1775 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1776                                             struct fib6_config *cfg,
1777                                             const struct in6_addr *gw_addr)
1778 {
1779         struct flowi6 fl6 = {
1780                 .flowi6_oif = cfg->fc_ifindex,
1781                 .daddr = *gw_addr,
1782                 .saddr = cfg->fc_prefsrc,
1783         };
1784         struct fib6_table *table;
1785         struct rt6_info *rt;
1786         int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
1787
1788         table = fib6_get_table(net, cfg->fc_table);
1789         if (!table)
1790                 return NULL;
1791
1792         if (!ipv6_addr_any(&cfg->fc_prefsrc))
1793                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1794
1795         rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1796
1797         /* if table lookup failed, fall back to full lookup */
1798         if (rt == net->ipv6.ip6_null_entry) {
1799                 ip6_rt_put(rt);
1800                 rt = NULL;
1801         }
1802
1803         return rt;
1804 }
1805
1806 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg,
1807                                               struct netlink_ext_ack *extack)
1808 {
1809         struct net *net = cfg->fc_nlinfo.nl_net;
1810         struct rt6_info *rt = NULL;
1811         struct net_device *dev = NULL;
1812         struct inet6_dev *idev = NULL;
1813         struct fib6_table *table;
1814         int addr_type;
1815         int err = -EINVAL;
1816
1817         /* RTF_PCPU is an internal flag; can not be set by userspace */
1818         if (cfg->fc_flags & RTF_PCPU) {
1819                 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
1820                 goto out;
1821         }
1822
1823         if (cfg->fc_dst_len > 128) {
1824                 NL_SET_ERR_MSG(extack, "Invalid prefix length");
1825                 goto out;
1826         }
1827         if (cfg->fc_src_len > 128) {
1828                 NL_SET_ERR_MSG(extack, "Invalid source address length");
1829                 goto out;
1830         }
1831 #ifndef CONFIG_IPV6_SUBTREES
1832         if (cfg->fc_src_len) {
1833                 NL_SET_ERR_MSG(extack,
1834                                "Specifying source address requires IPV6_SUBTREES to be enabled");
1835                 goto out;
1836         }
1837 #endif
1838         if (cfg->fc_ifindex) {
1839                 err = -ENODEV;
1840                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1841                 if (!dev)
1842                         goto out;
1843                 idev = in6_dev_get(dev);
1844                 if (!idev)
1845                         goto out;
1846         }
1847
1848         if (cfg->fc_metric == 0)
1849                 cfg->fc_metric = IP6_RT_PRIO_USER;
1850
1851         err = -ENOBUFS;
1852         if (cfg->fc_nlinfo.nlh &&
1853             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1854                 table = fib6_get_table(net, cfg->fc_table);
1855                 if (!table) {
1856                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1857                         table = fib6_new_table(net, cfg->fc_table);
1858                 }
1859         } else {
1860                 table = fib6_new_table(net, cfg->fc_table);
1861         }
1862
1863         if (!table)
1864                 goto out;
1865
1866         rt = ip6_dst_alloc(net, NULL,
1867                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1868
1869         if (!rt) {
1870                 err = -ENOMEM;
1871                 goto out;
1872         }
1873
1874         if (cfg->fc_flags & RTF_EXPIRES)
1875                 rt6_set_expires(rt, jiffies +
1876                                 clock_t_to_jiffies(cfg->fc_expires));
1877         else
1878                 rt6_clean_expires(rt);
1879
1880         if (cfg->fc_protocol == RTPROT_UNSPEC)
1881                 cfg->fc_protocol = RTPROT_BOOT;
1882         rt->rt6i_protocol = cfg->fc_protocol;
1883
1884         addr_type = ipv6_addr_type(&cfg->fc_dst);
1885
1886         if (addr_type & IPV6_ADDR_MULTICAST)
1887                 rt->dst.input = ip6_mc_input;
1888         else if (cfg->fc_flags & RTF_LOCAL)
1889                 rt->dst.input = ip6_input;
1890         else
1891                 rt->dst.input = ip6_forward;
1892
1893         rt->dst.output = ip6_output;
1894
1895         if (cfg->fc_encap) {
1896                 struct lwtunnel_state *lwtstate;
1897
1898                 err = lwtunnel_build_state(cfg->fc_encap_type,
1899                                            cfg->fc_encap, AF_INET6, cfg,
1900                                            &lwtstate, extack);
1901                 if (err)
1902                         goto out;
1903                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1904                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1905                         rt->dst.lwtstate->orig_output = rt->dst.output;
1906                         rt->dst.output = lwtunnel_output;
1907                 }
1908                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1909                         rt->dst.lwtstate->orig_input = rt->dst.input;
1910                         rt->dst.input = lwtunnel_input;
1911                 }
1912         }
1913
1914         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1915         rt->rt6i_dst.plen = cfg->fc_dst_len;
1916         if (rt->rt6i_dst.plen == 128)
1917                 rt->dst.flags |= DST_HOST;
1918
1919 #ifdef CONFIG_IPV6_SUBTREES
1920         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1921         rt->rt6i_src.plen = cfg->fc_src_len;
1922 #endif
1923
1924         rt->rt6i_metric = cfg->fc_metric;
1925
1926         /* We cannot add true routes via loopback here,
1927            they would result in kernel looping; promote them to reject routes
1928          */
1929         if ((cfg->fc_flags & RTF_REJECT) ||
1930             (dev && (dev->flags & IFF_LOOPBACK) &&
1931              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1932              !(cfg->fc_flags & RTF_LOCAL))) {
1933                 /* hold loopback dev/idev if we haven't done so. */
1934                 if (dev != net->loopback_dev) {
1935                         if (dev) {
1936                                 dev_put(dev);
1937                                 in6_dev_put(idev);
1938                         }
1939                         dev = net->loopback_dev;
1940                         dev_hold(dev);
1941                         idev = in6_dev_get(dev);
1942                         if (!idev) {
1943                                 err = -ENODEV;
1944                                 goto out;
1945                         }
1946                 }
1947                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1948                 switch (cfg->fc_type) {
1949                 case RTN_BLACKHOLE:
1950                         rt->dst.error = -EINVAL;
1951                         rt->dst.output = dst_discard_out;
1952                         rt->dst.input = dst_discard;
1953                         break;
1954                 case RTN_PROHIBIT:
1955                         rt->dst.error = -EACCES;
1956                         rt->dst.output = ip6_pkt_prohibit_out;
1957                         rt->dst.input = ip6_pkt_prohibit;
1958                         break;
1959                 case RTN_THROW:
1960                 case RTN_UNREACHABLE:
1961                 default:
1962                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1963                                         : (cfg->fc_type == RTN_UNREACHABLE)
1964                                         ? -EHOSTUNREACH : -ENETUNREACH;
1965                         rt->dst.output = ip6_pkt_discard_out;
1966                         rt->dst.input = ip6_pkt_discard;
1967                         break;
1968                 }
1969                 goto install_route;
1970         }
1971
1972         if (cfg->fc_flags & RTF_GATEWAY) {
1973                 const struct in6_addr *gw_addr;
1974                 int gwa_type;
1975
1976                 gw_addr = &cfg->fc_gateway;
1977                 gwa_type = ipv6_addr_type(gw_addr);
1978
1979                 /* if gw_addr is local we will fail to detect this in case
1980                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1981                  * will return already-added prefix route via interface that
1982                  * prefix route was assigned to, which might be non-loopback.
1983                  */
1984                 err = -EINVAL;
1985                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1986                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1987                                             dev : NULL, 0, 0)) {
1988                         NL_SET_ERR_MSG(extack, "Invalid gateway address");
1989                         goto out;
1990                 }
1991                 rt->rt6i_gateway = *gw_addr;
1992
1993                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1994                         struct rt6_info *grt = NULL;
1995
1996                         /* IPv6 strictly inhibits using not link-local
1997                            addresses as nexthop address.
1998                            Otherwise, router will not able to send redirects.
1999                            It is very good, but in some (rare!) circumstances
2000                            (SIT, PtP, NBMA NOARP links) it is handy to allow
2001                            some exceptions. --ANK
2002                            We allow IPv4-mapped nexthops to support RFC4798-type
2003                            addressing
2004                          */
2005                         if (!(gwa_type & (IPV6_ADDR_UNICAST |
2006                                           IPV6_ADDR_MAPPED))) {
2007                                 NL_SET_ERR_MSG(extack,
2008                                                "Invalid gateway address");
2009                                 goto out;
2010                         }
2011
2012                         if (cfg->fc_table) {
2013                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2014
2015                                 if (grt) {
2016                                         if (grt->rt6i_flags & RTF_GATEWAY ||
2017                                             (dev && dev != grt->dst.dev)) {
2018                                                 ip6_rt_put(grt);
2019                                                 grt = NULL;
2020                                         }
2021                                 }
2022                         }
2023
2024                         if (!grt)
2025                                 grt = rt6_lookup(net, gw_addr, NULL,
2026                                                  cfg->fc_ifindex, 1);
2027
2028                         err = -EHOSTUNREACH;
2029                         if (!grt)
2030                                 goto out;
2031                         if (dev) {
2032                                 if (dev != grt->dst.dev) {
2033                                         ip6_rt_put(grt);
2034                                         goto out;
2035                                 }
2036                         } else {
2037                                 dev = grt->dst.dev;
2038                                 idev = grt->rt6i_idev;
2039                                 dev_hold(dev);
2040                                 in6_dev_hold(grt->rt6i_idev);
2041                         }
2042                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2043                                 err = 0;
2044                         ip6_rt_put(grt);
2045
2046                         if (err)
2047                                 goto out;
2048                 }
2049                 err = -EINVAL;
2050                 if (!dev) {
2051                         NL_SET_ERR_MSG(extack, "Egress device not specified");
2052                         goto out;
2053                 } else if (dev->flags & IFF_LOOPBACK) {
2054                         NL_SET_ERR_MSG(extack,
2055                                        "Egress device can not be loopback device for this route");
2056                         goto out;
2057                 }
2058         }
2059
2060         err = -ENODEV;
2061         if (!dev)
2062                 goto out;
2063
2064         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2065                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2066                         NL_SET_ERR_MSG(extack, "Invalid source address");
2067                         err = -EINVAL;
2068                         goto out;
2069                 }
2070                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2071                 rt->rt6i_prefsrc.plen = 128;
2072         } else
2073                 rt->rt6i_prefsrc.plen = 0;
2074
2075         rt->rt6i_flags = cfg->fc_flags;
2076
2077 install_route:
2078         rt->dst.dev = dev;
2079         rt->rt6i_idev = idev;
2080         rt->rt6i_table = table;
2081
2082         cfg->fc_nlinfo.nl_net = dev_net(dev);
2083
2084         return rt;
2085 out:
2086         if (dev)
2087                 dev_put(dev);
2088         if (idev)
2089                 in6_dev_put(idev);
2090         if (rt)
2091                 dst_release_immediate(&rt->dst);
2092
2093         return ERR_PTR(err);
2094 }
2095
2096 int ip6_route_add(struct fib6_config *cfg,
2097                   struct netlink_ext_ack *extack)
2098 {
2099         struct mx6_config mxc = { .mx = NULL, };
2100         struct rt6_info *rt;
2101         int err;
2102
2103         rt = ip6_route_info_create(cfg, extack);
2104         if (IS_ERR(rt)) {
2105                 err = PTR_ERR(rt);
2106                 rt = NULL;
2107                 goto out;
2108         }
2109
2110         err = ip6_convert_metrics(&mxc, cfg);
2111         if (err)
2112                 goto out;
2113
2114         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc, extack);
2115
2116         kfree(mxc.mx);
2117
2118         return err;
2119 out:
2120         if (rt)
2121                 dst_release_immediate(&rt->dst);
2122
2123         return err;
2124 }
2125
2126 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2127 {
2128         int err;
2129         struct fib6_table *table;
2130         struct net *net = dev_net(rt->dst.dev);
2131
2132         if (rt == net->ipv6.ip6_null_entry) {
2133                 err = -ENOENT;
2134                 goto out;
2135         }
2136
2137         table = rt->rt6i_table;
2138         write_lock_bh(&table->tb6_lock);
2139         err = fib6_del(rt, info);
2140         write_unlock_bh(&table->tb6_lock);
2141
2142 out:
2143         ip6_rt_put(rt);
2144         return err;
2145 }
2146
2147 int ip6_del_rt(struct rt6_info *rt)
2148 {
2149         struct nl_info info = {
2150                 .nl_net = dev_net(rt->dst.dev),
2151         };
2152         return __ip6_del_rt(rt, &info);
2153 }
2154
2155 static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
2156 {
2157         struct nl_info *info = &cfg->fc_nlinfo;
2158         struct net *net = info->nl_net;
2159         struct sk_buff *skb = NULL;
2160         struct fib6_table *table;
2161         int err = -ENOENT;
2162
2163         if (rt == net->ipv6.ip6_null_entry)
2164                 goto out_put;
2165         table = rt->rt6i_table;
2166         write_lock_bh(&table->tb6_lock);
2167
2168         if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
2169                 struct rt6_info *sibling, *next_sibling;
2170
2171                 /* prefer to send a single notification with all hops */
2172                 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
2173                 if (skb) {
2174                         u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2175
2176                         if (rt6_fill_node(net, skb, rt,
2177                                           NULL, NULL, 0, RTM_DELROUTE,
2178                                           info->portid, seq, 0) < 0) {
2179                                 kfree_skb(skb);
2180                                 skb = NULL;
2181                         } else
2182                                 info->skip_notify = 1;
2183                 }
2184
2185                 list_for_each_entry_safe(sibling, next_sibling,
2186                                          &rt->rt6i_siblings,
2187                                          rt6i_siblings) {
2188                         err = fib6_del(sibling, info);
2189                         if (err)
2190                                 goto out_unlock;
2191                 }
2192         }
2193
2194         err = fib6_del(rt, info);
2195 out_unlock:
2196         write_unlock_bh(&table->tb6_lock);
2197 out_put:
2198         ip6_rt_put(rt);
2199
2200         if (skb) {
2201                 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2202                             info->nlh, gfp_any());
2203         }
2204         return err;
2205 }
2206
2207 static int ip6_route_del(struct fib6_config *cfg,
2208                          struct netlink_ext_ack *extack)
2209 {
2210         struct fib6_table *table;
2211         struct fib6_node *fn;
2212         struct rt6_info *rt;
2213         int err = -ESRCH;
2214
2215         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2216         if (!table) {
2217                 NL_SET_ERR_MSG(extack, "FIB table does not exist");
2218                 return err;
2219         }
2220
2221         read_lock_bh(&table->tb6_lock);
2222
2223         fn = fib6_locate(&table->tb6_root,
2224                          &cfg->fc_dst, cfg->fc_dst_len,
2225                          &cfg->fc_src, cfg->fc_src_len);
2226
2227         if (fn) {
2228                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2229                         if ((rt->rt6i_flags & RTF_CACHE) &&
2230                             !(cfg->fc_flags & RTF_CACHE))
2231                                 continue;
2232                         if (cfg->fc_ifindex &&
2233                             (!rt->dst.dev ||
2234                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2235                                 continue;
2236                         if (cfg->fc_flags & RTF_GATEWAY &&
2237                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2238                                 continue;
2239                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2240                                 continue;
2241                         if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2242                                 continue;
2243                         dst_hold(&rt->dst);
2244                         read_unlock_bh(&table->tb6_lock);
2245
2246                         /* if gateway was specified only delete the one hop */
2247                         if (cfg->fc_flags & RTF_GATEWAY)
2248                                 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2249
2250                         return __ip6_del_rt_siblings(rt, cfg);
2251                 }
2252         }
2253         read_unlock_bh(&table->tb6_lock);
2254
2255         return err;
2256 }
2257
2258 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2259 {
2260         struct netevent_redirect netevent;
2261         struct rt6_info *rt, *nrt = NULL;
2262         struct ndisc_options ndopts;
2263         struct inet6_dev *in6_dev;
2264         struct neighbour *neigh;
2265         struct rd_msg *msg;
2266         int optlen, on_link;
2267         u8 *lladdr;
2268
2269         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2270         optlen -= sizeof(*msg);
2271
2272         if (optlen < 0) {
2273                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2274                 return;
2275         }
2276
2277         msg = (struct rd_msg *)icmp6_hdr(skb);
2278
2279         if (ipv6_addr_is_multicast(&msg->dest)) {
2280                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2281                 return;
2282         }
2283
2284         on_link = 0;
2285         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2286                 on_link = 1;
2287         } else if (ipv6_addr_type(&msg->target) !=
2288                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2289                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2290                 return;
2291         }
2292
2293         in6_dev = __in6_dev_get(skb->dev);
2294         if (!in6_dev)
2295                 return;
2296         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2297                 return;
2298
2299         /* RFC2461 8.1:
2300          *      The IP source address of the Redirect MUST be the same as the current
2301          *      first-hop router for the specified ICMP Destination Address.
2302          */
2303
2304         if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2305                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2306                 return;
2307         }
2308
2309         lladdr = NULL;
2310         if (ndopts.nd_opts_tgt_lladdr) {
2311                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2312                                              skb->dev);
2313                 if (!lladdr) {
2314                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2315                         return;
2316                 }
2317         }
2318
2319         rt = (struct rt6_info *) dst;
2320         if (rt->rt6i_flags & RTF_REJECT) {
2321                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2322                 return;
2323         }
2324
2325         /* Redirect received -> path was valid.
2326          * Look, redirects are sent only in response to data packets,
2327          * so that this nexthop apparently is reachable. --ANK
2328          */
2329         dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
2330
2331         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2332         if (!neigh)
2333                 return;
2334
2335         /*
2336          *      We have finally decided to accept it.
2337          */
2338
2339         ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2340                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2341                      NEIGH_UPDATE_F_OVERRIDE|
2342                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2343                                      NEIGH_UPDATE_F_ISROUTER)),
2344                      NDISC_REDIRECT, &ndopts);
2345
2346         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2347         if (!nrt)
2348                 goto out;
2349
2350         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2351         if (on_link)
2352                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2353
2354         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2355
2356         if (ip6_ins_rt(nrt))
2357                 goto out_release;
2358
2359         netevent.old = &rt->dst;
2360         netevent.new = &nrt->dst;
2361         netevent.daddr = &msg->dest;
2362         netevent.neigh = neigh;
2363         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2364
2365         if (rt->rt6i_flags & RTF_CACHE) {
2366                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2367                 ip6_del_rt(rt);
2368         }
2369
2370 out_release:
2371         /* Release the reference taken in
2372          * ip6_rt_cache_alloc()
2373          */
2374         dst_release(&nrt->dst);
2375
2376 out:
2377         neigh_release(neigh);
2378 }
2379
2380 /*
2381  *      Misc support functions
2382  */
2383
2384 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2385 {
2386         BUG_ON(from->dst.from);
2387
2388         rt->rt6i_flags &= ~RTF_EXPIRES;
2389         dst_hold(&from->dst);
2390         rt->dst.from = &from->dst;
2391         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2392 }
2393
2394 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2395 {
2396         rt->dst.input = ort->dst.input;
2397         rt->dst.output = ort->dst.output;
2398         rt->rt6i_dst = ort->rt6i_dst;
2399         rt->dst.error = ort->dst.error;
2400         rt->rt6i_idev = ort->rt6i_idev;
2401         if (rt->rt6i_idev)
2402                 in6_dev_hold(rt->rt6i_idev);
2403         rt->dst.lastuse = jiffies;
2404         rt->rt6i_gateway = ort->rt6i_gateway;
2405         rt->rt6i_flags = ort->rt6i_flags;
2406         rt6_set_from(rt, ort);
2407         rt->rt6i_metric = ort->rt6i_metric;
2408 #ifdef CONFIG_IPV6_SUBTREES
2409         rt->rt6i_src = ort->rt6i_src;
2410 #endif
2411         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2412         rt->rt6i_table = ort->rt6i_table;
2413         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2414 }
2415
2416 #ifdef CONFIG_IPV6_ROUTE_INFO
2417 static struct rt6_info *rt6_get_route_info(struct net *net,
2418                                            const struct in6_addr *prefix, int prefixlen,
2419                                            const struct in6_addr *gwaddr,
2420                                            struct net_device *dev)
2421 {
2422         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2423         int ifindex = dev->ifindex;
2424         struct fib6_node *fn;
2425         struct rt6_info *rt = NULL;
2426         struct fib6_table *table;
2427
2428         table = fib6_get_table(net, tb_id);
2429         if (!table)
2430                 return NULL;
2431
2432         read_lock_bh(&table->tb6_lock);
2433         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2434         if (!fn)
2435                 goto out;
2436
2437         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2438                 if (rt->dst.dev->ifindex != ifindex)
2439                         continue;
2440                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2441                         continue;
2442                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2443                         continue;
2444                 dst_hold(&rt->dst);
2445                 break;
2446         }
2447 out:
2448         read_unlock_bh(&table->tb6_lock);
2449         return rt;
2450 }
2451
2452 static struct rt6_info *rt6_add_route_info(struct net *net,
2453                                            const struct in6_addr *prefix, int prefixlen,
2454                                            const struct in6_addr *gwaddr,
2455                                            struct net_device *dev,
2456                                            unsigned int pref)
2457 {
2458         struct fib6_config cfg = {
2459                 .fc_metric      = IP6_RT_PRIO_USER,
2460                 .fc_ifindex     = dev->ifindex,
2461                 .fc_dst_len     = prefixlen,
2462                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2463                                   RTF_UP | RTF_PREF(pref),
2464                 .fc_nlinfo.portid = 0,
2465                 .fc_nlinfo.nlh = NULL,
2466                 .fc_nlinfo.nl_net = net,
2467         };
2468
2469         cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2470         cfg.fc_dst = *prefix;
2471         cfg.fc_gateway = *gwaddr;
2472
2473         /* We should treat it as a default route if prefix length is 0. */
2474         if (!prefixlen)
2475                 cfg.fc_flags |= RTF_DEFAULT;
2476
2477         ip6_route_add(&cfg, NULL);
2478
2479         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2480 }
2481 #endif
2482
2483 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2484 {
2485         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2486         struct rt6_info *rt;
2487         struct fib6_table *table;
2488
2489         table = fib6_get_table(dev_net(dev), tb_id);
2490         if (!table)
2491                 return NULL;
2492
2493         read_lock_bh(&table->tb6_lock);
2494         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2495                 if (dev == rt->dst.dev &&
2496                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2497                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2498                         break;
2499         }
2500         if (rt)
2501                 dst_hold(&rt->dst);
2502         read_unlock_bh(&table->tb6_lock);
2503         return rt;
2504 }
2505
2506 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2507                                      struct net_device *dev,
2508                                      unsigned int pref)
2509 {
2510         struct fib6_config cfg = {
2511                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2512                 .fc_metric      = IP6_RT_PRIO_USER,
2513                 .fc_ifindex     = dev->ifindex,
2514                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2515                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2516                 .fc_nlinfo.portid = 0,
2517                 .fc_nlinfo.nlh = NULL,
2518                 .fc_nlinfo.nl_net = dev_net(dev),
2519         };
2520
2521         cfg.fc_gateway = *gwaddr;
2522
2523         if (!ip6_route_add(&cfg, NULL)) {
2524                 struct fib6_table *table;
2525
2526                 table = fib6_get_table(dev_net(dev), cfg.fc_table);
2527                 if (table)
2528                         table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
2529         }
2530
2531         return rt6_get_dflt_router(gwaddr, dev);
2532 }
2533
2534 static void __rt6_purge_dflt_routers(struct fib6_table *table)
2535 {
2536         struct rt6_info *rt;
2537
2538 restart:
2539         read_lock_bh(&table->tb6_lock);
2540         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2541                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
2542                     (!rt->rt6i_idev || rt->rt6i_idev->cnf.accept_ra != 2)) {
2543                         dst_hold(&rt->dst);
2544                         read_unlock_bh(&table->tb6_lock);
2545                         ip6_del_rt(rt);
2546                         goto restart;
2547                 }
2548         }
2549         read_unlock_bh(&table->tb6_lock);
2550
2551         table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
2552 }
2553
2554 void rt6_purge_dflt_routers(struct net *net)
2555 {
2556         struct fib6_table *table;
2557         struct hlist_head *head;
2558         unsigned int h;
2559
2560         rcu_read_lock();
2561
2562         for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
2563                 head = &net->ipv6.fib_table_hash[h];
2564                 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
2565                         if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
2566                                 __rt6_purge_dflt_routers(table);
2567                 }
2568         }
2569
2570         rcu_read_unlock();
2571 }
2572
2573 static void rtmsg_to_fib6_config(struct net *net,
2574                                  struct in6_rtmsg *rtmsg,
2575                                  struct fib6_config *cfg)
2576 {
2577         memset(cfg, 0, sizeof(*cfg));
2578
2579         cfg->fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
2580                          : RT6_TABLE_MAIN;
2581         cfg->fc_ifindex = rtmsg->rtmsg_ifindex;
2582         cfg->fc_metric = rtmsg->rtmsg_metric;
2583         cfg->fc_expires = rtmsg->rtmsg_info;
2584         cfg->fc_dst_len = rtmsg->rtmsg_dst_len;
2585         cfg->fc_src_len = rtmsg->rtmsg_src_len;
2586         cfg->fc_flags = rtmsg->rtmsg_flags;
2587
2588         cfg->fc_nlinfo.nl_net = net;
2589
2590         cfg->fc_dst = rtmsg->rtmsg_dst;
2591         cfg->fc_src = rtmsg->rtmsg_src;
2592         cfg->fc_gateway = rtmsg->rtmsg_gateway;
2593 }
2594
2595 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
2596 {
2597         struct fib6_config cfg;
2598         struct in6_rtmsg rtmsg;
2599         int err;
2600
2601         switch (cmd) {
2602         case SIOCADDRT:         /* Add a route */
2603         case SIOCDELRT:         /* Delete a route */
2604                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
2605                         return -EPERM;
2606                 err = copy_from_user(&rtmsg, arg,
2607                                      sizeof(struct in6_rtmsg));
2608                 if (err)
2609                         return -EFAULT;
2610
2611                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
2612
2613                 rtnl_lock();
2614                 switch (cmd) {
2615                 case SIOCADDRT:
2616                         err = ip6_route_add(&cfg, NULL);
2617                         break;
2618                 case SIOCDELRT:
2619                         err = ip6_route_del(&cfg, NULL);
2620                         break;
2621                 default:
2622                         err = -EINVAL;
2623                 }
2624                 rtnl_unlock();
2625
2626                 return err;
2627         }
2628
2629         return -EINVAL;
2630 }
2631
2632 /*
2633  *      Drop the packet on the floor
2634  */
2635
2636 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
2637 {
2638         int type;
2639         struct dst_entry *dst = skb_dst(skb);
2640         switch (ipstats_mib_noroutes) {
2641         case IPSTATS_MIB_INNOROUTES:
2642                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
2643                 if (type == IPV6_ADDR_ANY) {
2644                         IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2645                                       IPSTATS_MIB_INADDRERRORS);
2646                         break;
2647                 }
2648                 /* FALLTHROUGH */
2649         case IPSTATS_MIB_OUTNOROUTES:
2650                 IP6_INC_STATS(dev_net(dst->dev), ip6_dst_idev(dst),
2651                               ipstats_mib_noroutes);
2652                 break;
2653         }
2654         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
2655         kfree_skb(skb);
2656         return 0;
2657 }
2658
2659 static int ip6_pkt_discard(struct sk_buff *skb)
2660 {
2661         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
2662 }
2663
2664 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2665 {
2666         skb->dev = skb_dst(skb)->dev;
2667         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
2668 }
2669
2670 static int ip6_pkt_prohibit(struct sk_buff *skb)
2671 {
2672         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
2673 }
2674
2675 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
2676 {
2677         skb->dev = skb_dst(skb)->dev;
2678         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
2679 }
2680
2681 /*
2682  *      Allocate a dst for local (unicast / anycast) address.
2683  */
2684
2685 struct rt6_info *addrconf_dst_alloc(struct inet6_dev *idev,
2686                                     const struct in6_addr *addr,
2687                                     bool anycast)
2688 {
2689         u32 tb_id;
2690         struct net *net = dev_net(idev->dev);
2691         struct net_device *dev = net->loopback_dev;
2692         struct rt6_info *rt;
2693
2694         /* use L3 Master device as loopback for host routes if device
2695          * is enslaved and address is not link local or multicast
2696          */
2697         if (!rt6_need_strict(addr))
2698                 dev = l3mdev_master_dev_rcu(idev->dev) ? : dev;
2699
2700         rt = ip6_dst_alloc(net, dev, DST_NOCOUNT);
2701         if (!rt)
2702                 return ERR_PTR(-ENOMEM);
2703
2704         in6_dev_hold(idev);
2705
2706         rt->dst.flags |= DST_HOST;
2707         rt->dst.input = ip6_input;
2708         rt->dst.output = ip6_output;
2709         rt->rt6i_idev = idev;
2710
2711         rt->rt6i_protocol = RTPROT_KERNEL;
2712         rt->rt6i_flags = RTF_UP | RTF_NONEXTHOP;
2713         if (anycast)
2714                 rt->rt6i_flags |= RTF_ANYCAST;
2715         else
2716                 rt->rt6i_flags |= RTF_LOCAL;
2717
2718         rt->rt6i_gateway  = *addr;
2719         rt->rt6i_dst.addr = *addr;
2720         rt->rt6i_dst.plen = 128;
2721         tb_id = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL;
2722         rt->rt6i_table = fib6_get_table(net, tb_id);
2723
2724         return rt;
2725 }
2726
2727 /* remove deleted ip from prefsrc entries */
2728 struct arg_dev_net_ip {
2729         struct net_device *dev;
2730         struct net *net;
2731         struct in6_addr *addr;
2732 };
2733
2734 static int fib6_remove_prefsrc(struct rt6_info *rt, void *arg)
2735 {
2736         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
2737         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
2738         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
2739
2740         if (((void *)rt->dst.dev == dev || !dev) &&
2741             rt != net->ipv6.ip6_null_entry &&
2742             ipv6_addr_equal(addr, &rt->rt6i_prefsrc.addr)) {
2743                 /* remove prefsrc entry */
2744                 rt->rt6i_prefsrc.plen = 0;
2745         }
2746         return 0;
2747 }
2748
2749 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
2750 {
2751         struct net *net = dev_net(ifp->idev->dev);
2752         struct arg_dev_net_ip adni = {
2753                 .dev = ifp->idev->dev,
2754                 .net = net,
2755                 .addr = &ifp->addr,
2756         };
2757         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
2758 }
2759
2760 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2761 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2762
2763 /* Remove routers and update dst entries when gateway turn into host. */
2764 static int fib6_clean_tohost(struct rt6_info *rt, void *arg)
2765 {
2766         struct in6_addr *gateway = (struct in6_addr *)arg;
2767
2768         if ((((rt->rt6i_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) ||
2769              ((rt->rt6i_flags & RTF_CACHE_GATEWAY) == RTF_CACHE_GATEWAY)) &&
2770              ipv6_addr_equal(gateway, &rt->rt6i_gateway)) {
2771                 return -1;
2772         }
2773         return 0;
2774 }
2775
2776 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
2777 {
2778         fib6_clean_all(net, fib6_clean_tohost, gateway);
2779 }
2780
2781 struct arg_dev_net {
2782         struct net_device *dev;
2783         struct net *net;
2784 };
2785
2786 /* called with write lock held for table with rt */
2787 static int fib6_ifdown(struct rt6_info *rt, void *arg)
2788 {
2789         const struct arg_dev_net *adn = arg;
2790         const struct net_device *dev = adn->dev;
2791
2792         if ((rt->dst.dev == dev || !dev) &&
2793             rt != adn->net->ipv6.ip6_null_entry &&
2794             (rt->rt6i_nsiblings == 0 ||
2795              (dev && netdev_unregistering(dev)) ||
2796              !rt->rt6i_idev->cnf.ignore_routes_with_linkdown))
2797                 return -1;
2798
2799         return 0;
2800 }
2801
2802 void rt6_ifdown(struct net *net, struct net_device *dev)
2803 {
2804         struct arg_dev_net adn = {
2805                 .dev = dev,
2806                 .net = net,
2807         };
2808
2809         fib6_clean_all(net, fib6_ifdown, &adn);
2810         if (dev)
2811                 rt6_uncached_list_flush_dev(net, dev);
2812 }
2813
2814 struct rt6_mtu_change_arg {
2815         struct net_device *dev;
2816         unsigned int mtu;
2817 };
2818
2819 static int rt6_mtu_change_route(struct rt6_info *rt, void *p_arg)
2820 {
2821         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
2822         struct inet6_dev *idev;
2823
2824         /* In IPv6 pmtu discovery is not optional,
2825            so that RTAX_MTU lock cannot disable it.
2826            We still use this lock to block changes
2827            caused by addrconf/ndisc.
2828         */
2829
2830         idev = __in6_dev_get(arg->dev);
2831         if (!idev)
2832                 return 0;
2833
2834         /* For administrative MTU increase, there is no way to discover
2835            IPv6 PMTU increase, so PMTU increase should be updated here.
2836            Since RFC 1981 doesn't include administrative MTU increase
2837            update PMTU increase is a MUST. (i.e. jumbo frame)
2838          */
2839         /*
2840            If new MTU is less than route PMTU, this new MTU will be the
2841            lowest MTU in the path, update the route PMTU to reflect PMTU
2842            decreases; if new MTU is greater than route PMTU, and the
2843            old MTU is the lowest MTU in the path, update the route PMTU
2844            to reflect the increase. In this case if the other nodes' MTU
2845            also have the lowest MTU, TOO BIG MESSAGE will be lead to
2846            PMTU discovery.
2847          */
2848         if (rt->dst.dev == arg->dev &&
2849             dst_metric_raw(&rt->dst, RTAX_MTU) &&
2850             !dst_metric_locked(&rt->dst, RTAX_MTU)) {
2851                 if (rt->rt6i_flags & RTF_CACHE) {
2852                         /* For RTF_CACHE with rt6i_pmtu == 0
2853                          * (i.e. a redirected route),
2854                          * the metrics of its rt->dst.from has already
2855                          * been updated.
2856                          */
2857                         if (rt->rt6i_pmtu && rt->rt6i_pmtu > arg->mtu)
2858                                 rt->rt6i_pmtu = arg->mtu;
2859                 } else if (dst_mtu(&rt->dst) >= arg->mtu ||
2860                            (dst_mtu(&rt->dst) < arg->mtu &&
2861                             dst_mtu(&rt->dst) == idev->cnf.mtu6)) {
2862                         dst_metric_set(&rt->dst, RTAX_MTU, arg->mtu);
2863                 }
2864         }
2865         return 0;
2866 }
2867
2868 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
2869 {
2870         struct rt6_mtu_change_arg arg = {
2871                 .dev = dev,
2872                 .mtu = mtu,
2873         };
2874
2875         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
2876 }
2877
2878 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
2879         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
2880         [RTA_OIF]               = { .type = NLA_U32 },
2881         [RTA_IIF]               = { .type = NLA_U32 },
2882         [RTA_PRIORITY]          = { .type = NLA_U32 },
2883         [RTA_METRICS]           = { .type = NLA_NESTED },
2884         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
2885         [RTA_PREF]              = { .type = NLA_U8 },
2886         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
2887         [RTA_ENCAP]             = { .type = NLA_NESTED },
2888         [RTA_EXPIRES]           = { .type = NLA_U32 },
2889         [RTA_UID]               = { .type = NLA_U32 },
2890         [RTA_MARK]              = { .type = NLA_U32 },
2891 };
2892
2893 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
2894                               struct fib6_config *cfg,
2895                               struct netlink_ext_ack *extack)
2896 {
2897         struct rtmsg *rtm;
2898         struct nlattr *tb[RTA_MAX+1];
2899         unsigned int pref;
2900         int err;
2901
2902         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
2903                           NULL);
2904         if (err < 0)
2905                 goto errout;
2906
2907         err = -EINVAL;
2908         rtm = nlmsg_data(nlh);
2909         memset(cfg, 0, sizeof(*cfg));
2910
2911         cfg->fc_table = rtm->rtm_table;
2912         cfg->fc_dst_len = rtm->rtm_dst_len;
2913         cfg->fc_src_len = rtm->rtm_src_len;
2914         cfg->fc_flags = RTF_UP;
2915         cfg->fc_protocol = rtm->rtm_protocol;
2916         cfg->fc_type = rtm->rtm_type;
2917
2918         if (rtm->rtm_type == RTN_UNREACHABLE ||
2919             rtm->rtm_type == RTN_BLACKHOLE ||
2920             rtm->rtm_type == RTN_PROHIBIT ||
2921             rtm->rtm_type == RTN_THROW)
2922                 cfg->fc_flags |= RTF_REJECT;
2923
2924         if (rtm->rtm_type == RTN_LOCAL)
2925                 cfg->fc_flags |= RTF_LOCAL;
2926
2927         if (rtm->rtm_flags & RTM_F_CLONED)
2928                 cfg->fc_flags |= RTF_CACHE;
2929
2930         cfg->fc_nlinfo.portid = NETLINK_CB(skb).portid;
2931         cfg->fc_nlinfo.nlh = nlh;
2932         cfg->fc_nlinfo.nl_net = sock_net(skb->sk);
2933
2934         if (tb[RTA_GATEWAY]) {
2935                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
2936                 cfg->fc_flags |= RTF_GATEWAY;
2937         }
2938
2939         if (tb[RTA_DST]) {
2940                 int plen = (rtm->rtm_dst_len + 7) >> 3;
2941
2942                 if (nla_len(tb[RTA_DST]) < plen)
2943                         goto errout;
2944
2945                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
2946         }
2947
2948         if (tb[RTA_SRC]) {
2949                 int plen = (rtm->rtm_src_len + 7) >> 3;
2950
2951                 if (nla_len(tb[RTA_SRC]) < plen)
2952                         goto errout;
2953
2954                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
2955         }
2956
2957         if (tb[RTA_PREFSRC])
2958                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
2959
2960         if (tb[RTA_OIF])
2961                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
2962
2963         if (tb[RTA_PRIORITY])
2964                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
2965
2966         if (tb[RTA_METRICS]) {
2967                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
2968                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
2969         }
2970
2971         if (tb[RTA_TABLE])
2972                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
2973
2974         if (tb[RTA_MULTIPATH]) {
2975                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
2976                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
2977
2978                 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
2979                                                      cfg->fc_mp_len, extack);
2980                 if (err < 0)
2981                         goto errout;
2982         }
2983
2984         if (tb[RTA_PREF]) {
2985                 pref = nla_get_u8(tb[RTA_PREF]);
2986                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
2987                     pref != ICMPV6_ROUTER_PREF_HIGH)
2988                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
2989                 cfg->fc_flags |= RTF_PREF(pref);
2990         }
2991
2992         if (tb[RTA_ENCAP])
2993                 cfg->fc_encap = tb[RTA_ENCAP];
2994
2995         if (tb[RTA_ENCAP_TYPE]) {
2996                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
2997
2998                 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
2999                 if (err < 0)
3000                         goto errout;
3001         }
3002
3003         if (tb[RTA_EXPIRES]) {
3004                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
3005
3006                 if (addrconf_finite_timeout(timeout)) {
3007                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
3008                         cfg->fc_flags |= RTF_EXPIRES;
3009                 }
3010         }
3011
3012         err = 0;
3013 errout:
3014         return err;
3015 }
3016
3017 struct rt6_nh {
3018         struct rt6_info *rt6_info;
3019         struct fib6_config r_cfg;
3020         struct mx6_config mxc;
3021         struct list_head next;
3022 };
3023
3024 static void ip6_print_replace_route_err(struct list_head *rt6_nh_list)
3025 {
3026         struct rt6_nh *nh;
3027
3028         list_for_each_entry(nh, rt6_nh_list, next) {
3029                 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
3030                         &nh->r_cfg.fc_dst, &nh->r_cfg.fc_gateway,
3031                         nh->r_cfg.fc_ifindex);
3032         }
3033 }
3034
3035 static int ip6_route_info_append(struct list_head *rt6_nh_list,
3036                                  struct rt6_info *rt, struct fib6_config *r_cfg)
3037 {
3038         struct rt6_nh *nh;
3039         int err = -EEXIST;
3040
3041         list_for_each_entry(nh, rt6_nh_list, next) {
3042                 /* check if rt6_info already exists */
3043                 if (rt6_duplicate_nexthop(nh->rt6_info, rt))
3044                         return err;
3045         }
3046
3047         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
3048         if (!nh)
3049                 return -ENOMEM;
3050         nh->rt6_info = rt;
3051         err = ip6_convert_metrics(&nh->mxc, r_cfg);
3052         if (err) {
3053                 kfree(nh);
3054                 return err;
3055         }
3056         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
3057         list_add_tail(&nh->next, rt6_nh_list);
3058
3059         return 0;
3060 }
3061
3062 static void ip6_route_mpath_notify(struct rt6_info *rt,
3063                                    struct rt6_info *rt_last,
3064                                    struct nl_info *info,
3065                                    __u16 nlflags)
3066 {
3067         /* if this is an APPEND route, then rt points to the first route
3068          * inserted and rt_last points to last route inserted. Userspace
3069          * wants a consistent dump of the route which starts at the first
3070          * nexthop. Since sibling routes are always added at the end of
3071          * the list, find the first sibling of the last route appended
3072          */
3073         if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->rt6i_nsiblings) {
3074                 rt = list_first_entry(&rt_last->rt6i_siblings,
3075                                       struct rt6_info,
3076                                       rt6i_siblings);
3077         }
3078
3079         if (rt)
3080                 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
3081 }
3082
3083 static int ip6_route_multipath_add(struct fib6_config *cfg,
3084                                    struct netlink_ext_ack *extack)
3085 {
3086         struct rt6_info *rt_notif = NULL, *rt_last = NULL;
3087         struct nl_info *info = &cfg->fc_nlinfo;
3088         struct fib6_config r_cfg;
3089         struct rtnexthop *rtnh;
3090         struct rt6_info *rt;
3091         struct rt6_nh *err_nh;
3092         struct rt6_nh *nh, *nh_safe;
3093         __u16 nlflags;
3094         int remaining;
3095         int attrlen;
3096         int err = 1;
3097         int nhn = 0;
3098         int replace = (cfg->fc_nlinfo.nlh &&
3099                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
3100         LIST_HEAD(rt6_nh_list);
3101
3102         nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
3103         if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
3104                 nlflags |= NLM_F_APPEND;
3105
3106         remaining = cfg->fc_mp_len;
3107         rtnh = (struct rtnexthop *)cfg->fc_mp;
3108
3109         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3110          * rt6_info structs per nexthop
3111          */
3112         while (rtnh_ok(rtnh, remaining)) {
3113                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3114                 if (rtnh->rtnh_ifindex)
3115                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3116
3117                 attrlen = rtnh_attrlen(rtnh);
3118                 if (attrlen > 0) {
3119                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3120
3121                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3122                         if (nla) {
3123                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
3124                                 r_cfg.fc_flags |= RTF_GATEWAY;
3125                         }
3126                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
3127                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
3128                         if (nla)
3129                                 r_cfg.fc_encap_type = nla_get_u16(nla);
3130                 }
3131
3132                 rt = ip6_route_info_create(&r_cfg, extack);
3133                 if (IS_ERR(rt)) {
3134                         err = PTR_ERR(rt);
3135                         rt = NULL;
3136                         goto cleanup;
3137                 }
3138
3139                 err = ip6_route_info_append(&rt6_nh_list, rt, &r_cfg);
3140                 if (err) {
3141                         dst_release_immediate(&rt->dst);
3142                         goto cleanup;
3143                 }
3144
3145                 rtnh = rtnh_next(rtnh, &remaining);
3146         }
3147
3148         /* for add and replace send one notification with all nexthops.
3149          * Skip the notification in fib6_add_rt2node and send one with
3150          * the full route when done
3151          */
3152         info->skip_notify = 1;
3153
3154         err_nh = NULL;
3155         list_for_each_entry(nh, &rt6_nh_list, next) {
3156                 rt_last = nh->rt6_info;
3157                 err = __ip6_ins_rt(nh->rt6_info, info, &nh->mxc, extack);
3158                 /* save reference to first route for notification */
3159                 if (!rt_notif && !err)
3160                         rt_notif = nh->rt6_info;
3161
3162                 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3163                 nh->rt6_info = NULL;
3164                 if (err) {
3165                         if (replace && nhn)
3166                                 ip6_print_replace_route_err(&rt6_nh_list);
3167                         err_nh = nh;
3168                         goto add_errout;
3169                 }
3170
3171                 /* Because each route is added like a single route we remove
3172                  * these flags after the first nexthop: if there is a collision,
3173                  * we have already failed to add the first nexthop:
3174                  * fib6_add_rt2node() has rejected it; when replacing, old
3175                  * nexthops have been replaced by first new, the rest should
3176                  * be added to it.
3177                  */
3178                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
3179                                                      NLM_F_REPLACE);
3180                 nhn++;
3181         }
3182
3183         /* success ... tell user about new route */
3184         ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3185         goto cleanup;
3186
3187 add_errout:
3188         /* send notification for routes that were added so that
3189          * the delete notifications sent by ip6_route_del are
3190          * coherent
3191          */
3192         if (rt_notif)
3193                 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
3194
3195         /* Delete routes that were already added */
3196         list_for_each_entry(nh, &rt6_nh_list, next) {
3197                 if (err_nh == nh)
3198                         break;
3199                 ip6_route_del(&nh->r_cfg, extack);
3200         }
3201
3202 cleanup:
3203         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
3204                 if (nh->rt6_info)
3205                         dst_release_immediate(&nh->rt6_info->dst);
3206                 kfree(nh->mxc.mx);
3207                 list_del(&nh->next);
3208                 kfree(nh);
3209         }
3210
3211         return err;
3212 }
3213
3214 static int ip6_route_multipath_del(struct fib6_config *cfg,
3215                                    struct netlink_ext_ack *extack)
3216 {
3217         struct fib6_config r_cfg;
3218         struct rtnexthop *rtnh;
3219         int remaining;
3220         int attrlen;
3221         int err = 1, last_err = 0;
3222
3223         remaining = cfg->fc_mp_len;
3224         rtnh = (struct rtnexthop *)cfg->fc_mp;
3225
3226         /* Parse a Multipath Entry */
3227         while (rtnh_ok(rtnh, remaining)) {
3228                 memcpy(&r_cfg, cfg, sizeof(*cfg));
3229                 if (rtnh->rtnh_ifindex)
3230                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
3231
3232                 attrlen = rtnh_attrlen(rtnh);
3233                 if (attrlen > 0) {
3234                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
3235
3236                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
3237                         if (nla) {
3238                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
3239                                 r_cfg.fc_flags |= RTF_GATEWAY;
3240                         }
3241                 }
3242                 err = ip6_route_del(&r_cfg, extack);
3243                 if (err)
3244                         last_err = err;
3245
3246                 rtnh = rtnh_next(rtnh, &remaining);
3247         }
3248
3249         return last_err;
3250 }
3251
3252 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3253                               struct netlink_ext_ack *extack)
3254 {
3255         struct fib6_config cfg;
3256         int err;
3257
3258         err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
3259         if (err < 0)
3260                 return err;
3261
3262         if (cfg.fc_mp)
3263                 return ip6_route_multipath_del(&cfg, extack);
3264         else {
3265                 cfg.fc_delete_all_nh = 1;
3266                 return ip6_route_del(&cfg, extack);
3267         }
3268 }
3269
3270 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
3271                               struct netlink_ext_ack *extack)
3272 {
3273         struct fib6_config cfg;
3274         int err;
3275
3276         err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
3277         if (err < 0)
3278                 return err;
3279
3280         if (cfg.fc_mp)
3281                 return ip6_route_multipath_add(&cfg, extack);
3282         else
3283                 return ip6_route_add(&cfg, extack);
3284 }
3285
3286 static size_t rt6_nlmsg_size(struct rt6_info *rt)
3287 {
3288         int nexthop_len = 0;
3289
3290         if (rt->rt6i_nsiblings) {
3291                 nexthop_len = nla_total_size(0)  /* RTA_MULTIPATH */
3292                             + NLA_ALIGN(sizeof(struct rtnexthop))
3293                             + nla_total_size(16) /* RTA_GATEWAY */
3294                             + lwtunnel_get_encap_size(rt->dst.lwtstate);
3295
3296                 nexthop_len *= rt->rt6i_nsiblings;
3297         }
3298
3299         return NLMSG_ALIGN(sizeof(struct rtmsg))
3300                + nla_total_size(16) /* RTA_SRC */
3301                + nla_total_size(16) /* RTA_DST */
3302                + nla_total_size(16) /* RTA_GATEWAY */
3303                + nla_total_size(16) /* RTA_PREFSRC */
3304                + nla_total_size(4) /* RTA_TABLE */
3305                + nla_total_size(4) /* RTA_IIF */
3306                + nla_total_size(4) /* RTA_OIF */
3307                + nla_total_size(4) /* RTA_PRIORITY */
3308                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
3309                + nla_total_size(sizeof(struct rta_cacheinfo))
3310                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
3311                + nla_total_size(1) /* RTA_PREF */
3312                + lwtunnel_get_encap_size(rt->dst.lwtstate)
3313                + nexthop_len;
3314 }
3315
3316 static int rt6_nexthop_info(struct sk_buff *skb, struct rt6_info *rt,
3317                             unsigned int *flags, bool skip_oif)
3318 {
3319         if (!netif_running(rt->dst.dev) || !netif_carrier_ok(rt->dst.dev)) {
3320                 *flags |= RTNH_F_LINKDOWN;
3321                 if (rt->rt6i_idev->cnf.ignore_routes_with_linkdown)
3322                         *flags |= RTNH_F_DEAD;
3323         }
3324
3325         if (rt->rt6i_flags & RTF_GATEWAY) {
3326                 if (nla_put_in6_addr(skb, RTA_GATEWAY, &rt->rt6i_gateway) < 0)
3327                         goto nla_put_failure;
3328         }
3329
3330         /* not needed for multipath encoding b/c it has a rtnexthop struct */
3331         if (!skip_oif && rt->dst.dev &&
3332             nla_put_u32(skb, RTA_OIF, rt->dst.dev->ifindex))
3333                 goto nla_put_failure;
3334
3335         if (rt->dst.lwtstate &&
3336             lwtunnel_fill_encap(skb, rt->dst.lwtstate) < 0)
3337                 goto nla_put_failure;
3338
3339         return 0;
3340
3341 nla_put_failure:
3342         return -EMSGSIZE;
3343 }
3344
3345 /* add multipath next hop */
3346 static int rt6_add_nexthop(struct sk_buff *skb, struct rt6_info *rt)
3347 {
3348         struct rtnexthop *rtnh;
3349         unsigned int flags = 0;
3350
3351         rtnh = nla_reserve_nohdr(skb, sizeof(*rtnh));
3352         if (!rtnh)
3353                 goto nla_put_failure;
3354
3355         rtnh->rtnh_hops = 0;
3356         rtnh->rtnh_ifindex = rt->dst.dev ? rt->dst.dev->ifindex : 0;
3357
3358         if (rt6_nexthop_info(skb, rt, &flags, true) < 0)
3359                 goto nla_put_failure;
3360
3361         rtnh->rtnh_flags = flags;
3362
3363         /* length of rtnetlink header + attributes */
3364         rtnh->rtnh_len = nlmsg_get_pos(skb) - (void *)rtnh;
3365
3366         return 0;
3367
3368 nla_put_failure:
3369         return -EMSGSIZE;
3370 }
3371
3372 static int rt6_fill_node(struct net *net,
3373                          struct sk_buff *skb, struct rt6_info *rt,
3374                          struct in6_addr *dst, struct in6_addr *src,
3375                          int iif, int type, u32 portid, u32 seq,
3376                          unsigned int flags)
3377 {
3378         u32 metrics[RTAX_MAX];
3379         struct rtmsg *rtm;
3380         struct nlmsghdr *nlh;
3381         long expires;
3382         u32 table;
3383
3384         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
3385         if (!nlh)
3386                 return -EMSGSIZE;
3387
3388         rtm = nlmsg_data(nlh);
3389         rtm->rtm_family = AF_INET6;
3390         rtm->rtm_dst_len = rt->rt6i_dst.plen;
3391         rtm->rtm_src_len = rt->rt6i_src.plen;
3392         rtm->rtm_tos = 0;
3393         if (rt->rt6i_table)
3394                 table = rt->rt6i_table->tb6_id;
3395         else
3396                 table = RT6_TABLE_UNSPEC;
3397         rtm->rtm_table = table;
3398         if (nla_put_u32(skb, RTA_TABLE, table))
3399                 goto nla_put_failure;
3400         if (rt->rt6i_flags & RTF_REJECT) {
3401                 switch (rt->dst.error) {
3402                 case -EINVAL:
3403                         rtm->rtm_type = RTN_BLACKHOLE;
3404                         break;
3405                 case -EACCES:
3406                         rtm->rtm_type = RTN_PROHIBIT;
3407                         break;
3408                 case -EAGAIN:
3409                         rtm->rtm_type = RTN_THROW;
3410                         break;
3411                 default:
3412                         rtm->rtm_type = RTN_UNREACHABLE;
3413                         break;
3414                 }
3415         }
3416         else if (rt->rt6i_flags & RTF_LOCAL)
3417                 rtm->rtm_type = RTN_LOCAL;
3418         else if (rt->rt6i_flags & RTF_ANYCAST)
3419                 rtm->rtm_type = RTN_ANYCAST;
3420         else if (rt->dst.dev && (rt->dst.dev->flags & IFF_LOOPBACK))
3421                 rtm->rtm_type = RTN_LOCAL;
3422         else
3423                 rtm->rtm_type = RTN_UNICAST;
3424         rtm->rtm_flags = 0;
3425         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
3426         rtm->rtm_protocol = rt->rt6i_protocol;
3427         if (rt->rt6i_flags & RTF_DYNAMIC)
3428                 rtm->rtm_protocol = RTPROT_REDIRECT;
3429         else if (rt->rt6i_flags & RTF_ADDRCONF) {
3430                 if (rt->rt6i_flags & (RTF_DEFAULT | RTF_ROUTEINFO))
3431                         rtm->rtm_protocol = RTPROT_RA;
3432                 else
3433                         rtm->rtm_protocol = RTPROT_KERNEL;
3434         }
3435
3436         if (rt->rt6i_flags & RTF_CACHE)
3437                 rtm->rtm_flags |= RTM_F_CLONED;
3438
3439         if (dst) {
3440                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3441                         goto nla_put_failure;
3442                 rtm->rtm_dst_len = 128;
3443         } else if (rtm->rtm_dst_len)
3444                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3445                         goto nla_put_failure;
3446 #ifdef CONFIG_IPV6_SUBTREES
3447         if (src) {
3448                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3449                         goto nla_put_failure;
3450                 rtm->rtm_src_len = 128;
3451         } else if (rtm->rtm_src_len &&
3452                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3453                 goto nla_put_failure;
3454 #endif
3455         if (iif) {
3456 #ifdef CONFIG_IPV6_MROUTE
3457                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3458                         int err = ip6mr_get_route(net, skb, rtm, portid);
3459
3460                         if (err == 0)
3461                                 return 0;
3462                         if (err < 0)
3463                                 goto nla_put_failure;
3464                 } else
3465 #endif
3466                         if (nla_put_u32(skb, RTA_IIF, iif))
3467                                 goto nla_put_failure;
3468         } else if (dst) {
3469                 struct in6_addr saddr_buf;
3470                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3471                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3472                         goto nla_put_failure;
3473         }
3474
3475         if (rt->rt6i_prefsrc.plen) {
3476                 struct in6_addr saddr_buf;
3477                 saddr_buf = rt->rt6i_prefsrc.addr;
3478                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3479                         goto nla_put_failure;
3480         }
3481
3482         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3483         if (rt->rt6i_pmtu)
3484                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3485         if (rtnetlink_put_metrics(skb, metrics) < 0)
3486                 goto nla_put_failure;
3487
3488         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3489                 goto nla_put_failure;
3490
3491         /* For multipath routes, walk the siblings list and add
3492          * each as a nexthop within RTA_MULTIPATH.
3493          */
3494         if (rt->rt6i_nsiblings) {
3495                 struct rt6_info *sibling, *next_sibling;
3496                 struct nlattr *mp;
3497
3498                 mp = nla_nest_start(skb, RTA_MULTIPATH);
3499                 if (!mp)
3500                         goto nla_put_failure;
3501
3502                 if (rt6_add_nexthop(skb, rt) < 0)
3503                         goto nla_put_failure;
3504
3505                 list_for_each_entry_safe(sibling, next_sibling,
3506                                          &rt->rt6i_siblings, rt6i_siblings) {
3507                         if (rt6_add_nexthop(skb, sibling) < 0)
3508                                 goto nla_put_failure;
3509                 }
3510
3511                 nla_nest_end(skb, mp);
3512         } else {
3513                 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
3514                         goto nla_put_failure;
3515         }
3516
3517         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3518
3519         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3520                 goto nla_put_failure;
3521
3522         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3523                 goto nla_put_failure;
3524
3525
3526         nlmsg_end(skb, nlh);
3527         return 0;
3528
3529 nla_put_failure:
3530         nlmsg_cancel(skb, nlh);
3531         return -EMSGSIZE;
3532 }
3533
3534 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3535 {
3536         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3537         struct net *net = arg->net;
3538
3539         if (rt == net->ipv6.ip6_null_entry)
3540                 return 0;
3541
3542         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3543                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3544
3545                 /* user wants prefix routes only */
3546                 if (rtm->rtm_flags & RTM_F_PREFIX &&
3547                     !(rt->rt6i_flags & RTF_PREFIX_RT)) {
3548                         /* success since this is not a prefix route */
3549                         return 1;
3550                 }
3551         }
3552
3553         return rt6_fill_node(net,
3554                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3555                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3556                      NLM_F_MULTI);
3557 }
3558
3559 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3560                               struct netlink_ext_ack *extack)
3561 {
3562         struct net *net = sock_net(in_skb->sk);
3563         struct nlattr *tb[RTA_MAX+1];
3564         int err, iif = 0, oif = 0;
3565         struct dst_entry *dst;
3566         struct rt6_info *rt;
3567         struct sk_buff *skb;
3568         struct rtmsg *rtm;
3569         struct flowi6 fl6;
3570         bool fibmatch;
3571
3572         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
3573                           extack);
3574         if (err < 0)
3575                 goto errout;
3576
3577         err = -EINVAL;
3578         memset(&fl6, 0, sizeof(fl6));
3579         rtm = nlmsg_data(nlh);
3580         fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3581         fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
3582
3583         if (tb[RTA_SRC]) {
3584                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3585                         goto errout;
3586
3587                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3588         }
3589
3590         if (tb[RTA_DST]) {
3591                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3592                         goto errout;
3593
3594                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3595         }
3596
3597         if (tb[RTA_IIF])
3598                 iif = nla_get_u32(tb[RTA_IIF]);
3599
3600         if (tb[RTA_OIF])
3601                 oif = nla_get_u32(tb[RTA_OIF]);
3602
3603         if (tb[RTA_MARK])
3604                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3605
3606         if (tb[RTA_UID])
3607                 fl6.flowi6_uid = make_kuid(current_user_ns(),
3608                                            nla_get_u32(tb[RTA_UID]));
3609         else
3610                 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3611
3612         if (iif) {
3613                 struct net_device *dev;
3614                 int flags = 0;
3615
3616                 dev = __dev_get_by_index(net, iif);
3617                 if (!dev) {
3618                         err = -ENODEV;
3619                         goto errout;
3620                 }
3621
3622                 fl6.flowi6_iif = iif;
3623
3624                 if (!ipv6_addr_any(&fl6.saddr))
3625                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3626
3627                 if (!fibmatch)
3628                         dst = ip6_route_input_lookup(net, dev, &fl6, flags);
3629         } else {
3630                 fl6.flowi6_oif = oif;
3631
3632                 if (!fibmatch)
3633                         dst = ip6_route_output(net, NULL, &fl6);
3634         }
3635
3636         if (fibmatch)
3637                 dst = ip6_route_lookup(net, &fl6, 0);
3638
3639         rt = container_of(dst, struct rt6_info, dst);
3640         if (rt->dst.error) {
3641                 err = rt->dst.error;
3642                 ip6_rt_put(rt);
3643                 goto errout;
3644         }
3645
3646         if (rt == net->ipv6.ip6_null_entry) {
3647                 err = rt->dst.error;
3648                 ip6_rt_put(rt);
3649                 goto errout;
3650         }
3651
3652         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3653         if (!skb) {
3654                 ip6_rt_put(rt);
3655                 err = -ENOBUFS;
3656                 goto errout;
3657         }
3658
3659         skb_dst_set(skb, &rt->dst);
3660         if (fibmatch)
3661                 err = rt6_fill_node(net, skb, rt, NULL, NULL, iif,
3662                                     RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3663                                     nlh->nlmsg_seq, 0);
3664         else
3665                 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3666                                     RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3667                                     nlh->nlmsg_seq, 0);
3668         if (err < 0) {
3669                 kfree_skb(skb);
3670                 goto errout;
3671         }
3672
3673         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3674 errout:
3675         return err;
3676 }
3677
3678 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3679                      unsigned int nlm_flags)
3680 {
3681         struct sk_buff *skb;
3682         struct net *net = info->nl_net;
3683         u32 seq;
3684         int err;
3685
3686         err = -ENOBUFS;
3687         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3688
3689         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3690         if (!skb)
3691                 goto errout;
3692
3693         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3694                                 event, info->portid, seq, nlm_flags);
3695         if (err < 0) {
3696                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3697                 WARN_ON(err == -EMSGSIZE);
3698                 kfree_skb(skb);
3699                 goto errout;
3700         }
3701         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3702                     info->nlh, gfp_any());
3703         return;
3704 errout:
3705         if (err < 0)
3706                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3707 }
3708
3709 static int ip6_route_dev_notify(struct notifier_block *this,
3710                                 unsigned long event, void *ptr)
3711 {
3712         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3713         struct net *net = dev_net(dev);
3714
3715         if (!(dev->flags & IFF_LOOPBACK))
3716                 return NOTIFY_OK;
3717
3718         if (event == NETDEV_REGISTER) {
3719                 net->ipv6.ip6_null_entry->dst.dev = dev;
3720                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3721 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3722                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3723                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3724                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3725                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3726 #endif
3727          } else if (event == NETDEV_UNREGISTER &&
3728                     dev->reg_state != NETREG_UNREGISTERED) {
3729                 /* NETDEV_UNREGISTER could be fired for multiple times by
3730                  * netdev_wait_allrefs(). Make sure we only call this once.
3731                  */
3732                 in6_dev_put(net->ipv6.ip6_null_entry->rt6i_idev);
3733 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3734                 in6_dev_put(net->ipv6.ip6_prohibit_entry->rt6i_idev);
3735                 in6_dev_put(net->ipv6.ip6_blk_hole_entry->rt6i_idev);
3736 #endif
3737         }
3738
3739         return NOTIFY_OK;
3740 }
3741
3742 /*
3743  *      /proc
3744  */
3745
3746 #ifdef CONFIG_PROC_FS
3747
3748 static const struct file_operations ipv6_route_proc_fops = {
3749         .owner          = THIS_MODULE,
3750         .open           = ipv6_route_open,
3751         .read           = seq_read,
3752         .llseek         = seq_lseek,
3753         .release        = seq_release_net,
3754 };
3755
3756 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3757 {
3758         struct net *net = (struct net *)seq->private;
3759         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3760                    net->ipv6.rt6_stats->fib_nodes,
3761                    net->ipv6.rt6_stats->fib_route_nodes,
3762                    net->ipv6.rt6_stats->fib_rt_alloc,
3763                    net->ipv6.rt6_stats->fib_rt_entries,
3764                    net->ipv6.rt6_stats->fib_rt_cache,
3765                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3766                    net->ipv6.rt6_stats->fib_discarded_routes);
3767
3768         return 0;
3769 }
3770
3771 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3772 {
3773         return single_open_net(inode, file, rt6_stats_seq_show);
3774 }
3775
3776 static const struct file_operations rt6_stats_seq_fops = {
3777         .owner   = THIS_MODULE,
3778         .open    = rt6_stats_seq_open,
3779         .read    = seq_read,
3780         .llseek  = seq_lseek,
3781         .release = single_release_net,
3782 };
3783 #endif  /* CONFIG_PROC_FS */
3784
3785 #ifdef CONFIG_SYSCTL
3786
3787 static
3788 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3789                               void __user *buffer, size_t *lenp, loff_t *ppos)
3790 {
3791         struct net *net;
3792         int delay;
3793         if (!write)
3794                 return -EINVAL;
3795
3796         net = (struct net *)ctl->extra1;
3797         delay = net->ipv6.sysctl.flush_delay;
3798         proc_dointvec(ctl, write, buffer, lenp, ppos);
3799         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3800         return 0;
3801 }
3802
3803 struct ctl_table ipv6_route_table_template[] = {
3804         {
3805                 .procname       =       "flush",
3806                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3807                 .maxlen         =       sizeof(int),
3808                 .mode           =       0200,
3809                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3810         },
3811         {
3812                 .procname       =       "gc_thresh",
3813                 .data           =       &ip6_dst_ops_template.gc_thresh,
3814                 .maxlen         =       sizeof(int),
3815                 .mode           =       0644,
3816                 .proc_handler   =       proc_dointvec,
3817         },
3818         {
3819                 .procname       =       "max_size",
3820                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3821                 .maxlen         =       sizeof(int),
3822                 .mode           =       0644,
3823                 .proc_handler   =       proc_dointvec,
3824         },
3825         {
3826                 .procname       =       "gc_min_interval",
3827                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3828                 .maxlen         =       sizeof(int),
3829                 .mode           =       0644,
3830                 .proc_handler   =       proc_dointvec_jiffies,
3831         },
3832         {
3833                 .procname       =       "gc_timeout",
3834                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3835                 .maxlen         =       sizeof(int),
3836                 .mode           =       0644,
3837                 .proc_handler   =       proc_dointvec_jiffies,
3838         },
3839         {
3840                 .procname       =       "gc_interval",
3841                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3842                 .maxlen         =       sizeof(int),
3843                 .mode           =       0644,
3844                 .proc_handler   =       proc_dointvec_jiffies,
3845         },
3846         {
3847                 .procname       =       "gc_elasticity",
3848                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3849                 .maxlen         =       sizeof(int),
3850                 .mode           =       0644,
3851                 .proc_handler   =       proc_dointvec,
3852         },
3853         {
3854                 .procname       =       "mtu_expires",
3855                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3856                 .maxlen         =       sizeof(int),
3857                 .mode           =       0644,
3858                 .proc_handler   =       proc_dointvec_jiffies,
3859         },
3860         {
3861                 .procname       =       "min_adv_mss",
3862                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3863                 .maxlen         =       sizeof(int),
3864                 .mode           =       0644,
3865                 .proc_handler   =       proc_dointvec,
3866         },
3867         {
3868                 .procname       =       "gc_min_interval_ms",
3869                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3870                 .maxlen         =       sizeof(int),
3871                 .mode           =       0644,
3872                 .proc_handler   =       proc_dointvec_ms_jiffies,
3873         },
3874         { }
3875 };
3876
3877 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3878 {
3879         struct ctl_table *table;
3880
3881         table = kmemdup(ipv6_route_table_template,
3882                         sizeof(ipv6_route_table_template),
3883                         GFP_KERNEL);
3884
3885         if (table) {
3886                 table[0].data = &net->ipv6.sysctl.flush_delay;
3887                 table[0].extra1 = net;
3888                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3889                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3890                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3891                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3892                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3893                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3894                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3895                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3896                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3897
3898                 /* Don't export sysctls to unprivileged users */
3899                 if (net->user_ns != &init_user_ns)
3900                         table[0].procname = NULL;
3901         }
3902
3903         return table;
3904 }
3905 #endif
3906
3907 static int __net_init ip6_route_net_init(struct net *net)
3908 {
3909         int ret = -ENOMEM;
3910
3911         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3912                sizeof(net->ipv6.ip6_dst_ops));
3913
3914         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3915                 goto out_ip6_dst_ops;
3916
3917         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3918                                            sizeof(*net->ipv6.ip6_null_entry),
3919                                            GFP_KERNEL);
3920         if (!net->ipv6.ip6_null_entry)
3921                 goto out_ip6_dst_entries;
3922         net->ipv6.ip6_null_entry->dst.path =
3923                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3924         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3925         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3926                          ip6_template_metrics, true);
3927
3928 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3929         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3930                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3931                                                GFP_KERNEL);
3932         if (!net->ipv6.ip6_prohibit_entry)
3933                 goto out_ip6_null_entry;
3934         net->ipv6.ip6_prohibit_entry->dst.path =
3935                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3936         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3937         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3938                          ip6_template_metrics, true);
3939
3940         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3941                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3942                                                GFP_KERNEL);
3943         if (!net->ipv6.ip6_blk_hole_entry)
3944                 goto out_ip6_prohibit_entry;
3945         net->ipv6.ip6_blk_hole_entry->dst.path =
3946                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3947         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3948         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3949                          ip6_template_metrics, true);
3950 #endif
3951
3952         net->ipv6.sysctl.flush_delay = 0;
3953         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3954         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3955         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3956         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3957         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3958         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3959         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3960
3961         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3962
3963         ret = 0;
3964 out:
3965         return ret;
3966
3967 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3968 out_ip6_prohibit_entry:
3969         kfree(net->ipv6.ip6_prohibit_entry);
3970 out_ip6_null_entry:
3971         kfree(net->ipv6.ip6_null_entry);
3972 #endif
3973 out_ip6_dst_entries:
3974         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3975 out_ip6_dst_ops:
3976         goto out;
3977 }
3978
3979 static void __net_exit ip6_route_net_exit(struct net *net)
3980 {
3981         kfree(net->ipv6.ip6_null_entry);
3982 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3983         kfree(net->ipv6.ip6_prohibit_entry);
3984         kfree(net->ipv6.ip6_blk_hole_entry);
3985 #endif
3986         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3987 }
3988
3989 static int __net_init ip6_route_net_init_late(struct net *net)
3990 {
3991 #ifdef CONFIG_PROC_FS
3992         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3993         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3994 #endif
3995         return 0;
3996 }
3997
3998 static void __net_exit ip6_route_net_exit_late(struct net *net)
3999 {
4000 #ifdef CONFIG_PROC_FS
4001         remove_proc_entry("ipv6_route", net->proc_net);
4002         remove_proc_entry("rt6_stats", net->proc_net);
4003 #endif
4004 }
4005
4006 static struct pernet_operations ip6_route_net_ops = {
4007         .init = ip6_route_net_init,
4008         .exit = ip6_route_net_exit,
4009 };
4010
4011 static int __net_init ipv6_inetpeer_init(struct net *net)
4012 {
4013         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
4014
4015         if (!bp)
4016                 return -ENOMEM;
4017         inet_peer_base_init(bp);
4018         net->ipv6.peers = bp;
4019         return 0;
4020 }
4021
4022 static void __net_exit ipv6_inetpeer_exit(struct net *net)
4023 {
4024         struct inet_peer_base *bp = net->ipv6.peers;
4025
4026         net->ipv6.peers = NULL;
4027         inetpeer_invalidate_tree(bp);
4028         kfree(bp);
4029 }
4030
4031 static struct pernet_operations ipv6_inetpeer_ops = {
4032         .init   =       ipv6_inetpeer_init,
4033         .exit   =       ipv6_inetpeer_exit,
4034 };
4035
4036 static struct pernet_operations ip6_route_net_late_ops = {
4037         .init = ip6_route_net_init_late,
4038         .exit = ip6_route_net_exit_late,
4039 };
4040
4041 static struct notifier_block ip6_route_dev_notifier = {
4042         .notifier_call = ip6_route_dev_notify,
4043         .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
4044 };
4045
4046 void __init ip6_route_init_special_entries(void)
4047 {
4048         /* Registering of the loopback is done before this portion of code,
4049          * the loopback reference in rt6_info will not be taken, do it
4050          * manually for init_net */
4051         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
4052         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4053   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4054         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
4055         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4056         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
4057         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4058   #endif
4059 }
4060
4061 int __init ip6_route_init(void)
4062 {
4063         int ret;
4064         int cpu;
4065
4066         ret = -ENOMEM;
4067         ip6_dst_ops_template.kmem_cachep =
4068                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
4069                                   SLAB_HWCACHE_ALIGN, NULL);
4070         if (!ip6_dst_ops_template.kmem_cachep)
4071                 goto out;
4072
4073         ret = dst_entries_init(&ip6_dst_blackhole_ops);
4074         if (ret)
4075                 goto out_kmem_cache;
4076
4077         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
4078         if (ret)
4079                 goto out_dst_entries;
4080
4081         ret = register_pernet_subsys(&ip6_route_net_ops);
4082         if (ret)
4083                 goto out_register_inetpeer;
4084
4085         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
4086
4087         ret = fib6_init();
4088         if (ret)
4089                 goto out_register_subsys;
4090
4091         ret = xfrm6_init();
4092         if (ret)
4093                 goto out_fib6_init;
4094
4095         ret = fib6_rules_init();
4096         if (ret)
4097                 goto xfrm6_init;
4098
4099         ret = register_pernet_subsys(&ip6_route_net_late_ops);
4100         if (ret)
4101                 goto fib6_rules_init;
4102
4103         ret = -ENOBUFS;
4104         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
4105             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
4106             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
4107                 goto out_register_late_subsys;
4108
4109         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
4110         if (ret)
4111                 goto out_register_late_subsys;
4112
4113         for_each_possible_cpu(cpu) {
4114                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
4115
4116                 INIT_LIST_HEAD(&ul->head);
4117                 spin_lock_init(&ul->lock);
4118         }
4119
4120 out:
4121         return ret;
4122
4123 out_register_late_subsys:
4124         unregister_pernet_subsys(&ip6_route_net_late_ops);
4125 fib6_rules_init:
4126         fib6_rules_cleanup();
4127 xfrm6_init:
4128         xfrm6_fini();
4129 out_fib6_init:
4130         fib6_gc_cleanup();
4131 out_register_subsys:
4132         unregister_pernet_subsys(&ip6_route_net_ops);
4133 out_register_inetpeer:
4134         unregister_pernet_subsys(&ipv6_inetpeer_ops);
4135 out_dst_entries:
4136         dst_entries_destroy(&ip6_dst_blackhole_ops);
4137 out_kmem_cache:
4138         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4139         goto out;
4140 }
4141
4142 void ip6_route_cleanup(void)
4143 {
4144         unregister_netdevice_notifier(&ip6_route_dev_notifier);
4145         unregister_pernet_subsys(&ip6_route_net_late_ops);
4146         fib6_rules_cleanup();
4147         xfrm6_fini();
4148         fib6_gc_cleanup();
4149         unregister_pernet_subsys(&ipv6_inetpeer_ops);
4150         unregister_pernet_subsys(&ip6_route_net_ops);
4151         dst_entries_destroy(&ip6_dst_blackhole_ops);
4152         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4153 }