net/ipv6/route.c

   1 /*
   2  *      Linux INET6 implementation
   3  *      FIB front-end.
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *
   8  *      This program is free software; you can redistribute it and/or
   9  *      modify it under the terms of the GNU General Public License
  10  *      as published by the Free Software Foundation; either version
  11  *      2 of the License, or (at your option) any later version.
  12  */
  13
  14 /*      Changes:
  15  *
  16  *      YOSHIFUJI Hideaki @USAGI
  17  *              reworked default router selection.
  18  *              - respect outgoing interface
  19  *              - select from (probably) reachable routers (i.e.
  20  *              routers in REACHABLE, STALE, DELAY or PROBE states).
  21  *              - always select the same router if it is (probably)
  22  *              reachable.  otherwise, round-robin the list.
  23  *      Ville Nuorvala
  24  *              Fixed routing subtrees.
  25  */
  26
  27 #define pr_fmt(fmt) "IPv6: " fmt
  28
  29 #include <linux/capability.h>
  30 #include <linux/errno.h>
  31 #include <linux/export.h>
  32 #include <linux/types.h>
  33 #include <linux/times.h>
  34 #include <linux/socket.h>
  35 #include <linux/sockios.h>
  36 #include <linux/net.h>
  37 #include <linux/route.h>
  38 #include <linux/netdevice.h>
  39 #include <linux/in6.h>
  40 #include <linux/mroute6.h>
  41 #include <linux/init.h>
  42 #include <linux/if_arp.h>
  43 #include <linux/proc_fs.h>
  44 #include <linux/seq_file.h>
  45 #include <linux/nsproxy.h>
  46 #include <linux/slab.h>
  47 #include <net/net_namespace.h>
  48 #include <net/snmp.h>
  49 #include <net/ipv6.h>
  50 #include <net/ip6_fib.h>
  51 #include <net/ip6_route.h>
  52 #include <net/ndisc.h>
  53 #include <net/addrconf.h>
  54 #include <net/tcp.h>
  55 #include <linux/rtnetlink.h>
  56 #include <net/dst.h>
  57 #include <net/dst_metadata.h>
  58 #include <net/xfrm.h>
  59 #include <net/netevent.h>
  60 #include <net/netlink.h>
  61 #include <net/nexthop.h>
  62 #include <net/lwtunnel.h>
  63 #include <net/ip_tunnels.h>
  64 #include <net/l3mdev.h>
  65 #include <trace/events/fib6.h>
  66
  67 #include <linux/uaccess.h>
  68
  69 #ifdef CONFIG_SYSCTL
  70 #include <linux/sysctl.h>
  71 #endif
  72
  73 enum rt6_nud_state {
  74         RT6_NUD_FAIL_HARD = -3,
  75         RT6_NUD_FAIL_PROBE = -2,
  76         RT6_NUD_FAIL_DO_RR = -1,
  77         RT6_NUD_SUCCEED = 1
  78 };
  79
  80 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort);
  81 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
  82 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
  83 static unsigned int      ip6_mtu(const struct dst_entry *dst);
  84 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
  85 static void             ip6_dst_destroy(struct dst_entry *);
  86 static void             ip6_dst_ifdown(struct dst_entry *,
  87                                        struct net_device *dev, int how);
  88 static int               ip6_dst_gc(struct dst_ops *ops);
  89
  90 static int              ip6_pkt_discard(struct sk_buff *skb);
  91 static int              ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  92 static int              ip6_pkt_prohibit(struct sk_buff *skb);
  93 static int              ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
  94 static void             ip6_link_failure(struct sk_buff *skb);
  95 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
  96                                            struct sk_buff *skb, u32 mtu);
  97 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
  98                                         struct sk_buff *skb);
  99 static void             rt6_dst_from_metrics_check(struct rt6_info *rt);
 100 static int rt6_score_route(struct rt6_info *rt, int oif, int strict);
 101 static size_t rt6_nlmsg_size(struct rt6_info *rt);
 102 static int rt6_fill_node(struct net *net,
 103                          struct sk_buff *skb, struct rt6_info *rt,
 104                          struct in6_addr *dst, struct in6_addr *src,
 105                          int iif, int type, u32 portid, u32 seq,
 106                          unsigned int flags);
 107
 108 #ifdef CONFIG_IPV6_ROUTE_INFO
 109 static struct rt6_info *rt6_add_route_info(struct net *net,
 110                                            const struct in6_addr *prefix, int prefixlen,
 111                                            const struct in6_addr *gwaddr,
 112                                            struct net_device *dev,
 113                                            unsigned int pref);
 114 static struct rt6_info *rt6_get_route_info(struct net *net,
 115                                            const struct in6_addr *prefix, int prefixlen,
 116                                            const struct in6_addr *gwaddr,
 117                                            struct net_device *dev);
 118 #endif
 119
 120 struct uncached_list {
 121         spinlock_t              lock;
 122         struct list_head        head;
 123 };
 124
 125 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
 126
 127 static void rt6_uncached_list_add(struct rt6_info *rt)
 128 {
 129         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
 130
 131         rt->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 (idev && idev->dev != loopback_dev) {
 421                 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
 422                 if (loopback_idev) {
 423                         rt->rt6i_idev = loopback_idev;
 424                         in6_dev_put(idev);
 425                 }
 426         }
 427 }
 428
 429 static bool __rt6_check_expired(const struct rt6_info *rt)
 430 {
 431         if (rt->rt6i_flags & RTF_EXPIRES)
 432                 return time_after(jiffies, rt->dst.expires);
 433         else
 434                 return false;
 435 }
 436
 437 static bool rt6_check_expired(const struct rt6_info *rt)
 438 {
 439         if (rt->rt6i_flags & RTF_EXPIRES) {
 440                 if (time_after(jiffies, rt->dst.expires))
 441                         return true;
 442         } else if (rt->dst.from) {
 443                 return rt6_check_expired((struct rt6_info *) rt->dst.from);
 444         }
 445         return false;
 446 }
 447
 448 /* Multipath route selection:
 449  *   Hash based function using packet header and flowlabel.
 450  * Adapted from fib_info_hashfn()
 451  */
 452 static int rt6_info_hash_nhsfn(unsigned int candidate_count,
 453                                const struct flowi6 *fl6)
 454 {
 455         return get_hash_from_flowi6(fl6) % candidate_count;
 456 }
 457
 458 static struct rt6_info *rt6_multipath_select(struct rt6_info *match,
 459                                              struct flowi6 *fl6, int oif,
 460                                              int strict)
 461 {
 462         struct rt6_info *sibling, *next_sibling;
 463         int route_choosen;
 464
 465         route_choosen = rt6_info_hash_nhsfn(match->rt6i_nsiblings + 1, fl6);
 466         /* Don't change the route, if route_choosen == 0
 467          * (siblings does not include ourself)
 468          */
 469         if (route_choosen)
 470                 list_for_each_entry_safe(sibling, next_sibling,
 471                                 &match->rt6i_siblings, rt6i_siblings) {
 472                         route_choosen--;
 473                         if (route_choosen == 0) {
 474                                 if (rt6_score_route(sibling, oif, strict) < 0)
 475                                         break;
 476                                 match = sibling;
 477                                 break;
 478                         }
 479                 }
 480         return match;
 481 }
 482
 483 /*
 484  *      Route lookup. Any table->tb6_lock is implied.
 485  */
 486
 487 static inline struct rt6_info *rt6_device_match(struct net *net,
 488                                                     struct rt6_info *rt,
 489                                                     const struct in6_addr *saddr,
 490                                                     int oif,
 491                                                     int flags)
 492 {
 493         struct rt6_info *local = NULL;
 494         struct rt6_info *sprt;
 495
 496         if (!oif && ipv6_addr_any(saddr))
 497                 goto out;
 498
 499         for (sprt = rt; sprt; sprt = sprt->dst.rt6_next) {
 500                 struct net_device *dev = sprt->dst.dev;
 501
 502                 if (oif) {
 503                         if (dev->ifindex == oif)
 504                                 return sprt;
 505                         if (dev->flags & IFF_LOOPBACK) {
 506                                 if (!sprt->rt6i_idev ||
 507                                     sprt->rt6i_idev->dev->ifindex != oif) {
 508                                         if (flags & RT6_LOOKUP_F_IFACE)
 509                                                 continue;
 510                                         if (local &&
 511                                             local->rt6i_idev->dev->ifindex == oif)
 512                                                 continue;
 513                                 }
 514                                 local = sprt;
 515                         }
 516                 } else {
 517                         if (ipv6_chk_addr(net, saddr, dev,
 518                                           flags & RT6_LOOKUP_F_IFACE))
 519                                 return sprt;
 520                 }
 521         }
 522
 523         if (oif) {
 524                 if (local)
 525                         return local;
 526
 527                 if (flags & RT6_LOOKUP_F_IFACE)
 528                         return net->ipv6.ip6_null_entry;
 529         }
 530 out:
 531         return rt;
 532 }
 533
 534 #ifdef CONFIG_IPV6_ROUTER_PREF
 535 struct __rt6_probe_work {
 536         struct work_struct work;
 537         struct in6_addr target;
 538         struct net_device *dev;
 539 };
 540
 541 static void rt6_probe_deferred(struct work_struct *w)
 542 {
 543         struct in6_addr mcaddr;
 544         struct __rt6_probe_work *work =
 545                 container_of(w, struct __rt6_probe_work, work);
 546
 547         addrconf_addr_solict_mult(&work->target, &mcaddr);
 548         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
 549         dev_put(work->dev);
 550         kfree(work);
 551 }
 552
 553 static void rt6_probe(struct rt6_info *rt)
 554 {
 555         struct __rt6_probe_work *work;
 556         struct neighbour *neigh;
 557         /*
 558          * Okay, this does not seem to be appropriate
 559          * for now, however, we need to check if it
 560          * is really so; aka Router Reachability Probing.
 561          *
 562          * Router Reachability Probe MUST be rate-limited
 563          * to no more than one per minute.
 564          */
 565         if (!rt || !(rt->rt6i_flags & RTF_GATEWAY))
 566                 return;
 567         rcu_read_lock_bh();
 568         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 569         if (neigh) {
 570                 if (neigh->nud_state & NUD_VALID)
 571                         goto out;
 572
 573                 work = NULL;
 574                 write_lock(&neigh->lock);
 575                 if (!(neigh->nud_state & NUD_VALID) &&
 576                     time_after(jiffies,
 577                                neigh->updated +
 578                                rt->rt6i_idev->cnf.rtr_probe_interval)) {
 579                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
 580                         if (work)
 581                                 __neigh_set_probe_once(neigh);
 582                 }
 583                 write_unlock(&neigh->lock);
 584         } else {
 585                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
 586         }
 587
 588         if (work) {
 589                 INIT_WORK(&work->work, rt6_probe_deferred);
 590                 work->target = rt->rt6i_gateway;
 591                 dev_hold(rt->dst.dev);
 592                 work->dev = rt->dst.dev;
 593                 schedule_work(&work->work);
 594         }
 595
 596 out:
 597         rcu_read_unlock_bh();
 598 }
 599 #else
 600 static inline void rt6_probe(struct rt6_info *rt)
 601 {
 602 }
 603 #endif
 604
 605 /*
 606  * Default Router Selection (RFC 2461 6.3.6)
 607  */
 608 static inline int rt6_check_dev(struct rt6_info *rt, int oif)
 609 {
 610         struct net_device *dev = rt->dst.dev;
 611         if (!oif || dev->ifindex == oif)
 612                 return 2;
 613         if ((dev->flags & IFF_LOOPBACK) &&
 614             rt->rt6i_idev && rt->rt6i_idev->dev->ifindex == oif)
 615                 return 1;
 616         return 0;
 617 }
 618
 619 static inline enum rt6_nud_state rt6_check_neigh(struct rt6_info *rt)
 620 {
 621         struct neighbour *neigh;
 622         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
 623
 624         if (rt->rt6i_flags & RTF_NONEXTHOP ||
 625             !(rt->rt6i_flags & RTF_GATEWAY))
 626                 return RT6_NUD_SUCCEED;
 627
 628         rcu_read_lock_bh();
 629         neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
 630         if (neigh) {
 631                 read_lock(&neigh->lock);
 632                 if (neigh->nud_state & NUD_VALID)
 633                         ret = RT6_NUD_SUCCEED;
 634 #ifdef CONFIG_IPV6_ROUTER_PREF
 635                 else if (!(neigh->nud_state & NUD_FAILED))
 636                         ret = RT6_NUD_SUCCEED;
 637                 else
 638                         ret = RT6_NUD_FAIL_PROBE;
 639 #endif
 640                 read_unlock(&neigh->lock);
 641         } else {
 642                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
 643                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
 644         }
 645         rcu_read_unlock_bh();
 646
 647         return ret;
 648 }
 649
 650 static int rt6_score_route(struct rt6_info *rt, int oif,
 651                            int strict)
 652 {
 653         int m;
 654
 655         m = rt6_check_dev(rt, oif);
 656         if (!m && (strict & RT6_LOOKUP_F_IFACE))
 657                 return RT6_NUD_FAIL_HARD;
 658 #ifdef CONFIG_IPV6_ROUTER_PREF
 659         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt->rt6i_flags)) << 2;
 660 #endif
 661         if (strict & RT6_LOOKUP_F_REACHABLE) {
 662                 int n = rt6_check_neigh(rt);
 663                 if (n < 0)
 664                         return n;
 665         }
 666         return m;
 667 }
 668
 669 static struct rt6_info *find_match(struct rt6_info *rt, int oif, int strict,
 670                                    int *mpri, struct rt6_info *match,
 671                                    bool *do_rr)
 672 {
 673         int m;
 674         bool match_do_rr = false;
 675         struct inet6_dev *idev = rt->rt6i_idev;
 676         struct net_device *dev = rt->dst.dev;
 677
 678         if (dev && !netif_carrier_ok(dev) &&
 679             idev->cnf.ignore_routes_with_linkdown &&
 680             !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
 681                 goto out;
 682
 683         if (rt6_check_expired(rt))
 684                 goto out;
 685
 686         m = rt6_score_route(rt, oif, strict);
 687         if (m == RT6_NUD_FAIL_DO_RR) {
 688                 match_do_rr = true;
 689                 m = 0; /* lowest valid score */
 690         } else if (m == RT6_NUD_FAIL_HARD) {
 691                 goto out;
 692         }
 693
 694         if (strict & RT6_LOOKUP_F_REACHABLE)
 695                 rt6_probe(rt);
 696
 697         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
 698         if (m > *mpri) {
 699                 *do_rr = match_do_rr;
 700                 *mpri = m;
 701                 match = rt;
 702         }
 703 out:
 704         return match;
 705 }
 706
 707 static struct rt6_info *find_rr_leaf(struct fib6_node *fn,
 708                                      struct rt6_info *rr_head,
 709                                      u32 metric, int oif, int strict,
 710                                      bool *do_rr)
 711 {
 712         struct rt6_info *rt, *match, *cont;
 713         int mpri = -1;
 714
 715         match = NULL;
 716         cont = NULL;
 717         for (rt = rr_head; rt; rt = rt->dst.rt6_next) {
 718                 if (rt->rt6i_metric != metric) {
 719                         cont = rt;
 720                         break;
 721                 }
 722
 723                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 724         }
 725
 726         for (rt = fn->leaf; rt && rt != rr_head; rt = rt->dst.rt6_next) {
 727                 if (rt->rt6i_metric != metric) {
 728                         cont = rt;
 729                         break;
 730                 }
 731
 732                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 733         }
 734
 735         if (match || !cont)
 736                 return match;
 737
 738         for (rt = cont; rt; rt = rt->dst.rt6_next)
 739                 match = find_match(rt, oif, strict, &mpri, match, do_rr);
 740
 741         return match;
 742 }
 743
 744 static struct rt6_info *rt6_select(struct fib6_node *fn, int oif, int strict)
 745 {
 746         struct rt6_info *match, *rt0;
 747         struct net *net;
 748         bool do_rr = false;
 749
 750         rt0 = fn->rr_ptr;
 751         if (!rt0)
 752                 fn->rr_ptr = rt0 = fn->leaf;
 753
 754         match = find_rr_leaf(fn, rt0, rt0->rt6i_metric, oif, strict,
 755                              &do_rr);
 756
 757         if (do_rr) {
 758                 struct rt6_info *next = rt0->dst.rt6_next;
 759
 760                 /* no entries matched; do round-robin */
 761                 if (!next || next->rt6i_metric != rt0->rt6i_metric)
 762                         next = fn->leaf;
 763
 764                 if (next != rt0)
 765                         fn->rr_ptr = next;
 766         }
 767
 768         net = dev_net(rt0->dst.dev);
 769         return match ? match : net->ipv6.ip6_null_entry;
 770 }
 771
 772 static bool rt6_is_gw_or_nonexthop(const struct rt6_info *rt)
 773 {
 774         return (rt->rt6i_flags & (RTF_NONEXTHOP | RTF_GATEWAY));
 775 }
 776
 777 #ifdef CONFIG_IPV6_ROUTE_INFO
 778 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
 779                   const struct in6_addr *gwaddr)
 780 {
 781         struct net *net = dev_net(dev);
 782         struct route_info *rinfo = (struct route_info *) opt;
 783         struct in6_addr prefix_buf, *prefix;
 784         unsigned int pref;
 785         unsigned long lifetime;
 786         struct rt6_info *rt;
 787
 788         if (len < sizeof(struct route_info)) {
 789                 return -EINVAL;
 790         }
 791
 792         /* Sanity check for prefix_len and length */
 793         if (rinfo->length > 3) {
 794                 return -EINVAL;
 795         } else if (rinfo->prefix_len > 128) {
 796                 return -EINVAL;
 797         } else if (rinfo->prefix_len > 64) {
 798                 if (rinfo->length < 2) {
 799                         return -EINVAL;
 800                 }
 801         } else if (rinfo->prefix_len > 0) {
 802                 if (rinfo->length < 1) {
 803                         return -EINVAL;
 804                 }
 805         }
 806
 807         pref = rinfo->route_pref;
 808         if (pref == ICMPV6_ROUTER_PREF_INVALID)
 809                 return -EINVAL;
 810
 811         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
 812
 813         if (rinfo->length == 3)
 814                 prefix = (struct in6_addr *)rinfo->prefix;
 815         else {
 816                 /* this function is safe */
 817                 ipv6_addr_prefix(&prefix_buf,
 818                                  (struct in6_addr *)rinfo->prefix,
 819                                  rinfo->prefix_len);
 820                 prefix = &prefix_buf;
 821         }
 822
 823         if (rinfo->prefix_len == 0)
 824                 rt = rt6_get_dflt_router(gwaddr, dev);
 825         else
 826                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
 827                                         gwaddr, dev);
 828
 829         if (rt && !lifetime) {
 830                 ip6_del_rt(rt);
 831                 rt = NULL;
 832         }
 833
 834         if (!rt && lifetime)
 835                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
 836                                         dev, pref);
 837         else if (rt)
 838                 rt->rt6i_flags = RTF_ROUTEINFO |
 839                                  (rt->rt6i_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
 840
 841         if (rt) {
 842                 if (!addrconf_finite_timeout(lifetime))
 843                         rt6_clean_expires(rt);
 844                 else
 845                         rt6_set_expires(rt, jiffies + HZ * lifetime);
 846
 847                 ip6_rt_put(rt);
 848         }
 849         return 0;
 850 }
 851 #endif
 852
 853 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
 854                                         struct in6_addr *saddr)
 855 {
 856         struct fib6_node *pn;
 857         while (1) {
 858                 if (fn->fn_flags & RTN_TL_ROOT)
 859                         return NULL;
 860                 pn = fn->parent;
 861                 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn)
 862                         fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr);
 863                 else
 864                         fn = pn;
 865                 if (fn->fn_flags & RTN_RTINFO)
 866                         return fn;
 867         }
 868 }
 869
 870 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
 871                                              struct fib6_table *table,
 872                                              struct flowi6 *fl6, int flags)
 873 {
 874         struct fib6_node *fn;
 875         struct rt6_info *rt;
 876
 877         read_lock_bh(&table->tb6_lock);
 878         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
 879 restart:
 880         rt = fn->leaf;
 881         rt = rt6_device_match(net, rt, &fl6->saddr, fl6->flowi6_oif, flags);
 882         if (rt->rt6i_nsiblings && fl6->flowi6_oif == 0)
 883                 rt = rt6_multipath_select(rt, fl6, fl6->flowi6_oif, flags);
 884         if (rt == net->ipv6.ip6_null_entry) {
 885                 fn = fib6_backtrack(fn, &fl6->saddr);
 886                 if (fn)
 887                         goto restart;
 888         }
 889         dst_use(&rt->dst, jiffies);
 890         read_unlock_bh(&table->tb6_lock);
 891
 892         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
 893
 894         return rt;
 895
 896 }
 897
 898 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
 899                                     int flags)
 900 {
 901         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_lookup);
 902 }
 903 EXPORT_SYMBOL_GPL(ip6_route_lookup);
 904
 905 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
 906                             const struct in6_addr *saddr, int oif, int strict)
 907 {
 908         struct flowi6 fl6 = {
 909                 .flowi6_oif = oif,
 910                 .daddr = *daddr,
 911         };
 912         struct dst_entry *dst;
 913         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
 914
 915         if (saddr) {
 916                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
 917                 flags |= RT6_LOOKUP_F_HAS_SADDR;
 918         }
 919
 920         dst = fib6_rule_lookup(net, &fl6, flags, ip6_pol_route_lookup);
 921         if (dst->error == 0)
 922                 return (struct rt6_info *) dst;
 923
 924         dst_release(dst);
 925
 926         return NULL;
 927 }
 928 EXPORT_SYMBOL(rt6_lookup);
 929
 930 /* ip6_ins_rt is called with FREE table->tb6_lock.
 931  * It takes new route entry, the addition fails by any reason the
 932  * route is released.
 933  * Caller must hold dst before calling it.
 934  */
 935
 936 static int __ip6_ins_rt(struct rt6_info *rt, struct nl_info *info,
 937                         struct mx6_config *mxc,
 938                         struct netlink_ext_ack *extack)
 939 {
 940         int err;
 941         struct fib6_table *table;
 942
 943         table = rt->rt6i_table;
 944         write_lock_bh(&table->tb6_lock);
 945         err = fib6_add(&table->tb6_root, rt, info, mxc, extack);
 946         write_unlock_bh(&table->tb6_lock);
 947
 948         return err;
 949 }
 950
 951 int ip6_ins_rt(struct rt6_info *rt)
 952 {
 953         struct nl_info info = { .nl_net = dev_net(rt->dst.dev), };
 954         struct mx6_config mxc = { .mx = NULL, };
 955
 956         /* Hold dst to account for the reference from the fib6 tree */
 957         dst_hold(&rt->dst);
 958         return __ip6_ins_rt(rt, &info, &mxc, NULL);
 959 }
 960
 961 static struct rt6_info *ip6_rt_cache_alloc(struct rt6_info *ort,
 962                                            const struct in6_addr *daddr,
 963                                            const struct in6_addr *saddr)
 964 {
 965         struct rt6_info *rt;
 966
 967         /*
 968          *      Clone the route.
 969          */
 970
 971         if (ort->rt6i_flags & (RTF_CACHE | RTF_PCPU))
 972                 ort = (struct rt6_info *)ort->dst.from;
 973
 974         rt = __ip6_dst_alloc(dev_net(ort->dst.dev), ort->dst.dev, 0);
 975
 976         if (!rt)
 977                 return NULL;
 978
 979         ip6_rt_copy_init(rt, ort);
 980         rt->rt6i_flags |= RTF_CACHE;
 981         rt->rt6i_metric = 0;
 982         rt->dst.flags |= DST_HOST;
 983         rt->rt6i_dst.addr = *daddr;
 984         rt->rt6i_dst.plen = 128;
 985
 986         if (!rt6_is_gw_or_nonexthop(ort)) {
 987                 if (ort->rt6i_dst.plen != 128 &&
 988                     ipv6_addr_equal(&ort->rt6i_dst.addr, daddr))
 989                         rt->rt6i_flags |= RTF_ANYCAST;
 990 #ifdef CONFIG_IPV6_SUBTREES
 991                 if (rt->rt6i_src.plen && saddr) {
 992                         rt->rt6i_src.addr = *saddr;
 993                         rt->rt6i_src.plen = 128;
 994                 }
 995 #endif
 996         }
 997
 998         return rt;
 999 }
1000
1001 static struct rt6_info *ip6_rt_pcpu_alloc(struct rt6_info *rt)
1002 {
1003         struct rt6_info *pcpu_rt;
1004
1005         pcpu_rt = __ip6_dst_alloc(dev_net(rt->dst.dev),
1006                                   rt->dst.dev, rt->dst.flags);
1007
1008         if (!pcpu_rt)
1009                 return NULL;
1010         ip6_rt_copy_init(pcpu_rt, rt);
1011         pcpu_rt->rt6i_protocol = rt->rt6i_protocol;
1012         pcpu_rt->rt6i_flags |= RTF_PCPU;
1013         return pcpu_rt;
1014 }
1015
1016 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1017 static struct rt6_info *rt6_get_pcpu_route(struct rt6_info *rt)
1018 {
1019         struct rt6_info *pcpu_rt, **p;
1020
1021         p = this_cpu_ptr(rt->rt6i_pcpu);
1022         pcpu_rt = *p;
1023
1024         if (pcpu_rt) {
1025                 dst_hold(&pcpu_rt->dst);
1026                 rt6_dst_from_metrics_check(pcpu_rt);
1027         }
1028         return pcpu_rt;
1029 }
1030
1031 static struct rt6_info *rt6_make_pcpu_route(struct rt6_info *rt)
1032 {
1033         struct fib6_table *table = rt->rt6i_table;
1034         struct rt6_info *pcpu_rt, *prev, **p;
1035
1036         pcpu_rt = ip6_rt_pcpu_alloc(rt);
1037         if (!pcpu_rt) {
1038                 struct net *net = dev_net(rt->dst.dev);
1039
1040                 dst_hold(&net->ipv6.ip6_null_entry->dst);
1041                 return net->ipv6.ip6_null_entry;
1042         }
1043
1044         read_lock_bh(&table->tb6_lock);
1045         if (rt->rt6i_pcpu) {
1046                 p = this_cpu_ptr(rt->rt6i_pcpu);
1047                 prev = cmpxchg(p, NULL, pcpu_rt);
1048                 if (prev) {
1049                         /* If someone did it before us, return prev instead */
1050                         dst_release_immediate(&pcpu_rt->dst);
1051                         pcpu_rt = prev;
1052                 }
1053         } else {
1054                 /* rt has been removed from the fib6 tree
1055                  * before we have a chance to acquire the read_lock.
1056                  * In this case, don't brother to create a pcpu rt
1057                  * since rt is going away anyway.  The next
1058                  * dst_check() will trigger a re-lookup.
1059                  */
1060                 dst_release_immediate(&pcpu_rt->dst);
1061                 pcpu_rt = rt;
1062         }
1063         dst_hold(&pcpu_rt->dst);
1064         rt6_dst_from_metrics_check(pcpu_rt);
1065         read_unlock_bh(&table->tb6_lock);
1066         return pcpu_rt;
1067 }
1068
1069 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
1070                                int oif, struct flowi6 *fl6, int flags)
1071 {
1072         struct fib6_node *fn, *saved_fn;
1073         struct rt6_info *rt;
1074         int strict = 0;
1075
1076         strict |= flags & RT6_LOOKUP_F_IFACE;
1077         strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
1078         if (net->ipv6.devconf_all->forwarding == 0)
1079                 strict |= RT6_LOOKUP_F_REACHABLE;
1080
1081         read_lock_bh(&table->tb6_lock);
1082
1083         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1084         saved_fn = fn;
1085
1086         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1087                 oif = 0;
1088
1089 redo_rt6_select:
1090         rt = rt6_select(fn, oif, strict);
1091         if (rt->rt6i_nsiblings)
1092                 rt = rt6_multipath_select(rt, fl6, oif, strict);
1093         if (rt == net->ipv6.ip6_null_entry) {
1094                 fn = fib6_backtrack(fn, &fl6->saddr);
1095                 if (fn)
1096                         goto redo_rt6_select;
1097                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
1098                         /* also consider unreachable route */
1099                         strict &= ~RT6_LOOKUP_F_REACHABLE;
1100                         fn = saved_fn;
1101                         goto redo_rt6_select;
1102                 }
1103         }
1104
1105
1106         if (rt == net->ipv6.ip6_null_entry || (rt->rt6i_flags & RTF_CACHE)) {
1107                 dst_use(&rt->dst, jiffies);
1108                 read_unlock_bh(&table->tb6_lock);
1109
1110                 rt6_dst_from_metrics_check(rt);
1111
1112                 trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1113                 return rt;
1114         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
1115                             !(rt->rt6i_flags & RTF_GATEWAY))) {
1116                 /* Create a RTF_CACHE clone which will not be
1117                  * owned by the fib6 tree.  It is for the special case where
1118                  * the daddr in the skb during the neighbor look-up is different
1119                  * from the fl6->daddr used to look-up route here.
1120                  */
1121
1122                 struct rt6_info *uncached_rt;
1123
1124                 dst_use(&rt->dst, jiffies);
1125                 read_unlock_bh(&table->tb6_lock);
1126
1127                 uncached_rt = ip6_rt_cache_alloc(rt, &fl6->daddr, NULL);
1128                 dst_release(&rt->dst);
1129
1130                 if (uncached_rt) {
1131                         /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
1132                          * No need for another dst_hold()
1133                          */
1134                         rt6_uncached_list_add(uncached_rt);
1135                 } else {
1136                         uncached_rt = net->ipv6.ip6_null_entry;
1137                         dst_hold(&uncached_rt->dst);
1138                 }
1139
1140                 trace_fib6_table_lookup(net, uncached_rt, table->tb6_id, fl6);
1141                 return uncached_rt;
1142
1143         } else {
1144                 /* Get a percpu copy */
1145
1146                 struct rt6_info *pcpu_rt;
1147
1148                 rt->dst.lastuse = jiffies;
1149                 rt->dst.__use++;
1150                 pcpu_rt = rt6_get_pcpu_route(rt);
1151
1152                 if (pcpu_rt) {
1153                         read_unlock_bh(&table->tb6_lock);
1154                 } else {
1155                         /* We have to do the read_unlock first
1156                          * because rt6_make_pcpu_route() may trigger
1157                          * ip6_dst_gc() which will take the write_lock.
1158                          */
1159                         dst_hold(&rt->dst);
1160                         read_unlock_bh(&table->tb6_lock);
1161                         pcpu_rt = rt6_make_pcpu_route(rt);
1162                         dst_release(&rt->dst);
1163                 }
1164
1165                 trace_fib6_table_lookup(net, pcpu_rt, table->tb6_id, fl6);
1166                 return pcpu_rt;
1167
1168         }
1169 }
1170 EXPORT_SYMBOL_GPL(ip6_pol_route);
1171
1172 static struct rt6_info *ip6_pol_route_input(struct net *net, struct fib6_table *table,
1173                                             struct flowi6 *fl6, int flags)
1174 {
1175         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, flags);
1176 }
1177
1178 struct dst_entry *ip6_route_input_lookup(struct net *net,
1179                                          struct net_device *dev,
1180                                          struct flowi6 *fl6, int flags)
1181 {
1182         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
1183                 flags |= RT6_LOOKUP_F_IFACE;
1184
1185         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_input);
1186 }
1187 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
1188
1189 void ip6_route_input(struct sk_buff *skb)
1190 {
1191         const struct ipv6hdr *iph = ipv6_hdr(skb);
1192         struct net *net = dev_net(skb->dev);
1193         int flags = RT6_LOOKUP_F_HAS_SADDR;
1194         struct ip_tunnel_info *tun_info;
1195         struct flowi6 fl6 = {
1196                 .flowi6_iif = skb->dev->ifindex,
1197                 .daddr = iph->daddr,
1198                 .saddr = iph->saddr,
1199                 .flowlabel = ip6_flowinfo(iph),
1200                 .flowi6_mark = skb->mark,
1201                 .flowi6_proto = iph->nexthdr,
1202         };
1203
1204         tun_info = skb_tunnel_info(skb);
1205         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
1206                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
1207         skb_dst_drop(skb);
1208         skb_dst_set(skb, ip6_route_input_lookup(net, skb->dev, &fl6, flags));
1209 }
1210
1211 static struct rt6_info *ip6_pol_route_output(struct net *net, struct fib6_table *table,
1212                                              struct flowi6 *fl6, int flags)
1213 {
1214         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, flags);
1215 }
1216
1217 struct dst_entry *ip6_route_output_flags(struct net *net, const struct sock *sk,
1218                                          struct flowi6 *fl6, int flags)
1219 {
1220         bool any_src;
1221
1222         if (rt6_need_strict(&fl6->daddr)) {
1223                 struct dst_entry *dst;
1224
1225                 dst = l3mdev_link_scope_lookup(net, fl6);
1226                 if (dst)
1227                         return dst;
1228         }
1229
1230         fl6->flowi6_iif = LOOPBACK_IFINDEX;
1231
1232         any_src = ipv6_addr_any(&fl6->saddr);
1233         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
1234             (fl6->flowi6_oif && any_src))
1235                 flags |= RT6_LOOKUP_F_IFACE;
1236
1237         if (!any_src)
1238                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1239         else if (sk)
1240                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
1241
1242         return fib6_rule_lookup(net, fl6, flags, ip6_pol_route_output);
1243 }
1244 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
1245
1246 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
1247 {
1248         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
1249         struct net_device *loopback_dev = net->loopback_dev;
1250         struct dst_entry *new = NULL;
1251
1252         rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
1253                        DST_OBSOLETE_NONE, 0);
1254         if (rt) {
1255                 rt6_info_init(rt);
1256
1257                 new = &rt->dst;
1258                 new->__use = 1;
1259                 new->input = dst_discard;
1260                 new->output = dst_discard_out;
1261
1262                 dst_copy_metrics(new, &ort->dst);
1263
1264                 rt->rt6i_idev = in6_dev_get(loopback_dev);
1265                 rt->rt6i_gateway = ort->rt6i_gateway;
1266                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
1267                 rt->rt6i_metric = 0;
1268
1269                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
1270 #ifdef CONFIG_IPV6_SUBTREES
1271                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
1272 #endif
1273         }
1274
1275         dst_release(dst_orig);
1276         return new ? new : ERR_PTR(-ENOMEM);
1277 }
1278
1279 /*
1280  *      Destination cache support functions
1281  */
1282
1283 static void rt6_dst_from_metrics_check(struct rt6_info *rt)
1284 {
1285         if (rt->dst.from &&
1286             dst_metrics_ptr(&rt->dst) != dst_metrics_ptr(rt->dst.from))
1287                 dst_init_metrics(&rt->dst, dst_metrics_ptr(rt->dst.from), true);
1288 }
1289
1290 static struct dst_entry *rt6_check(struct rt6_info *rt, u32 cookie)
1291 {
1292         if (!rt->rt6i_node || (rt->rt6i_node->fn_sernum != cookie))
1293                 return NULL;
1294
1295         if (rt6_check_expired(rt))
1296                 return NULL;
1297
1298         return &rt->dst;
1299 }
1300
1301 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt, u32 cookie)
1302 {
1303         if (!__rt6_check_expired(rt) &&
1304             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
1305             rt6_check((struct rt6_info *)(rt->dst.from), cookie))
1306                 return &rt->dst;
1307         else
1308                 return NULL;
1309 }
1310
1311 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
1312 {
1313         struct rt6_info *rt;
1314
1315         rt = (struct rt6_info *) dst;
1316
1317         /* All IPV6 dsts are created with ->obsolete set to the value
1318          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1319          * into this function always.
1320          */
1321
1322         rt6_dst_from_metrics_check(rt);
1323
1324         if (rt->rt6i_flags & RTF_PCPU ||
1325             (unlikely(!list_empty(&rt->rt6i_uncached)) && rt->dst.from))
1326                 return rt6_dst_from_check(rt, cookie);
1327         else
1328                 return rt6_check(rt, cookie);
1329 }
1330
1331 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
1332 {
1333         struct rt6_info *rt = (struct rt6_info *) dst;
1334
1335         if (rt) {
1336                 if (rt->rt6i_flags & RTF_CACHE) {
1337                         if (rt6_check_expired(rt)) {
1338                                 ip6_del_rt(rt);
1339                                 dst = NULL;
1340                         }
1341                 } else {
1342                         dst_release(dst);
1343                         dst = NULL;
1344                 }
1345         }
1346         return dst;
1347 }
1348
1349 static void ip6_link_failure(struct sk_buff *skb)
1350 {
1351         struct rt6_info *rt;
1352
1353         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
1354
1355         rt = (struct rt6_info *) skb_dst(skb);
1356         if (rt) {
1357                 if (rt->rt6i_flags & RTF_CACHE) {
1358                         if (dst_hold_safe(&rt->dst))
1359                                 ip6_del_rt(rt);
1360                 } else if (rt->rt6i_node && (rt->rt6i_flags & RTF_DEFAULT)) {
1361                         rt->rt6i_node->fn_sernum = -1;
1362                 }
1363         }
1364 }
1365
1366 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
1367 {
1368         struct net *net = dev_net(rt->dst.dev);
1369
1370         rt->rt6i_flags |= RTF_MODIFIED;
1371         rt->rt6i_pmtu = mtu;
1372         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
1373 }
1374
1375 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
1376 {
1377         return !(rt->rt6i_flags & RTF_CACHE) &&
1378                 (rt->rt6i_flags & RTF_PCPU || rt->rt6i_node);
1379 }
1380
1381 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
1382                                  const struct ipv6hdr *iph, u32 mtu)
1383 {
1384         const struct in6_addr *daddr, *saddr;
1385         struct rt6_info *rt6 = (struct rt6_info *)dst;
1386
1387         if (rt6->rt6i_flags & RTF_LOCAL)
1388                 return;
1389
1390         if (dst_metric_locked(dst, RTAX_MTU))
1391                 return;
1392
1393         if (iph) {
1394                 daddr = &iph->daddr;
1395                 saddr = &iph->saddr;
1396         } else if (sk) {
1397                 daddr = &sk->sk_v6_daddr;
1398                 saddr = &inet6_sk(sk)->saddr;
1399         } else {
1400                 daddr = NULL;
1401                 saddr = NULL;
1402         }
1403         dst_confirm_neigh(dst, daddr);
1404         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
1405         if (mtu >= dst_mtu(dst))
1406                 return;
1407
1408         if (!rt6_cache_allowed_for_pmtu(rt6)) {
1409                 rt6_do_update_pmtu(rt6, mtu);
1410         } else if (daddr) {
1411                 struct rt6_info *nrt6;
1412
1413                 nrt6 = ip6_rt_cache_alloc(rt6, daddr, saddr);
1414                 if (nrt6) {
1415                         rt6_do_update_pmtu(nrt6, mtu);
1416
1417                         /* ip6_ins_rt(nrt6) will bump the
1418                          * rt6->rt6i_node->fn_sernum
1419                          * which will fail the next rt6_check() and
1420                          * invalidate the sk->sk_dst_cache.
1421                          */
1422                         ip6_ins_rt(nrt6);
1423                         /* Release the reference taken in
1424                          * ip6_rt_cache_alloc()
1425                          */
1426                         dst_release(&nrt6->dst);
1427                 }
1428         }
1429 }
1430
1431 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
1432                                struct sk_buff *skb, u32 mtu)
1433 {
1434         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
1435 }
1436
1437 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
1438                      int oif, u32 mark, kuid_t uid)
1439 {
1440         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1441         struct dst_entry *dst;
1442         struct flowi6 fl6;
1443
1444         memset(&fl6, 0, sizeof(fl6));
1445         fl6.flowi6_oif = oif;
1446         fl6.flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark);
1447         fl6.daddr = iph->daddr;
1448         fl6.saddr = iph->saddr;
1449         fl6.flowlabel = ip6_flowinfo(iph);
1450         fl6.flowi6_uid = uid;
1451
1452         dst = ip6_route_output(net, NULL, &fl6);
1453         if (!dst->error)
1454                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
1455         dst_release(dst);
1456 }
1457 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
1458
1459 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
1460 {
1461         struct dst_entry *dst;
1462
1463         ip6_update_pmtu(skb, sock_net(sk), mtu,
1464                         sk->sk_bound_dev_if, sk->sk_mark, sk->sk_uid);
1465
1466         dst = __sk_dst_get(sk);
1467         if (!dst || !dst->obsolete ||
1468             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
1469                 return;
1470
1471         bh_lock_sock(sk);
1472         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
1473                 ip6_datagram_dst_update(sk, false);
1474         bh_unlock_sock(sk);
1475 }
1476 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
1477
1478 /* Handle redirects */
1479 struct ip6rd_flowi {
1480         struct flowi6 fl6;
1481         struct in6_addr gateway;
1482 };
1483
1484 static struct rt6_info *__ip6_route_redirect(struct net *net,
1485                                              struct fib6_table *table,
1486                                              struct flowi6 *fl6,
1487                                              int flags)
1488 {
1489         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
1490         struct rt6_info *rt;
1491         struct fib6_node *fn;
1492
1493         /* Get the "current" route for this destination and
1494          * check if the redirect has come from appropriate router.
1495          *
1496          * RFC 4861 specifies that redirects should only be
1497          * accepted if they come from the nexthop to the target.
1498          * Due to the way the routes are chosen, this notion
1499          * is a bit fuzzy and one might need to check all possible
1500          * routes.
1501          */
1502
1503         read_lock_bh(&table->tb6_lock);
1504         fn = fib6_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1505 restart:
1506         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
1507                 if (rt6_check_expired(rt))
1508                         continue;
1509                 if (rt->dst.error)
1510                         break;
1511                 if (!(rt->rt6i_flags & RTF_GATEWAY))
1512                         continue;
1513                 if (fl6->flowi6_oif != rt->dst.dev->ifindex)
1514                         continue;
1515                 if (!ipv6_addr_equal(&rdfl->gateway, &rt->rt6i_gateway))
1516                         continue;
1517                 break;
1518         }
1519
1520         if (!rt)
1521                 rt = net->ipv6.ip6_null_entry;
1522         else if (rt->dst.error) {
1523                 rt = net->ipv6.ip6_null_entry;
1524                 goto out;
1525         }
1526
1527         if (rt == net->ipv6.ip6_null_entry) {
1528                 fn = fib6_backtrack(fn, &fl6->saddr);
1529                 if (fn)
1530                         goto restart;
1531         }
1532
1533 out:
1534         dst_hold(&rt->dst);
1535
1536         read_unlock_bh(&table->tb6_lock);
1537
1538         trace_fib6_table_lookup(net, rt, table->tb6_id, fl6);
1539         return rt;
1540 };
1541
1542 static struct dst_entry *ip6_route_redirect(struct net *net,
1543                                         const struct flowi6 *fl6,
1544                                         const struct in6_addr *gateway)
1545 {
1546         int flags = RT6_LOOKUP_F_HAS_SADDR;
1547         struct ip6rd_flowi rdfl;
1548
1549         rdfl.fl6 = *fl6;
1550         rdfl.gateway = *gateway;
1551
1552         return fib6_rule_lookup(net, &rdfl.fl6,
1553                                 flags, __ip6_route_redirect);
1554 }
1555
1556 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
1557                   kuid_t uid)
1558 {
1559         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
1560         struct dst_entry *dst;
1561         struct flowi6 fl6;
1562
1563         memset(&fl6, 0, sizeof(fl6));
1564         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1565         fl6.flowi6_oif = oif;
1566         fl6.flowi6_mark = mark;
1567         fl6.daddr = iph->daddr;
1568         fl6.saddr = iph->saddr;
1569         fl6.flowlabel = ip6_flowinfo(iph);
1570         fl6.flowi6_uid = uid;
1571
1572         dst = ip6_route_redirect(net, &fl6, &ipv6_hdr(skb)->saddr);
1573         rt6_do_redirect(dst, NULL, skb);
1574         dst_release(dst);
1575 }
1576 EXPORT_SYMBOL_GPL(ip6_redirect);
1577
1578 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif,
1579                             u32 mark)
1580 {
1581         const struct ipv6hdr *iph = ipv6_hdr(skb);
1582         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
1583         struct dst_entry *dst;
1584         struct flowi6 fl6;
1585
1586         memset(&fl6, 0, sizeof(fl6));
1587         fl6.flowi6_iif = LOOPBACK_IFINDEX;
1588         fl6.flowi6_oif = oif;
1589         fl6.flowi6_mark = mark;
1590         fl6.daddr = msg->dest;
1591         fl6.saddr = iph->daddr;
1592         fl6.flowi6_uid = sock_net_uid(net, NULL);
1593
1594         dst = ip6_route_redirect(net, &fl6, &iph->saddr);
1595         rt6_do_redirect(dst, NULL, skb);
1596         dst_release(dst);
1597 }
1598
1599 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
1600 {
1601         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
1602                      sk->sk_uid);
1603 }
1604 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
1605
1606 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
1607 {
1608         struct net_device *dev = dst->dev;
1609         unsigned int mtu = dst_mtu(dst);
1610         struct net *net = dev_net(dev);
1611
1612         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
1613
1614         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
1615                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
1616
1617         /*
1618          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1619          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1620          * IPV6_MAXPLEN is also valid and means: "any MSS,
1621          * rely only on pmtu discovery"
1622          */
1623         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
1624                 mtu = IPV6_MAXPLEN;
1625         return mtu;
1626 }
1627
1628 static unsigned int ip6_mtu(const struct dst_entry *dst)
1629 {
1630         const struct rt6_info *rt = (const struct rt6_info *)dst;
1631         unsigned int mtu = rt->rt6i_pmtu;
1632         struct inet6_dev *idev;
1633
1634         if (mtu)
1635                 goto out;
1636
1637         mtu = dst_metric_raw(dst, RTAX_MTU);
1638         if (mtu)
1639                 goto out;
1640
1641         mtu = IPV6_MIN_MTU;
1642
1643         rcu_read_lock();
1644         idev = __in6_dev_get(dst->dev);
1645         if (idev)
1646                 mtu = idev->cnf.mtu6;
1647         rcu_read_unlock();
1648
1649 out:
1650         mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1651
1652         return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
1653 }
1654
1655 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
1656                                   struct flowi6 *fl6)
1657 {
1658         struct dst_entry *dst;
1659         struct rt6_info *rt;
1660         struct inet6_dev *idev = in6_dev_get(dev);
1661         struct net *net = dev_net(dev);
1662
1663         if (unlikely(!idev))
1664                 return ERR_PTR(-ENODEV);
1665
1666         rt = ip6_dst_alloc(net, dev, 0);
1667         if (unlikely(!rt)) {
1668                 in6_dev_put(idev);
1669                 dst = ERR_PTR(-ENOMEM);
1670                 goto out;
1671         }
1672
1673         rt->dst.flags |= DST_HOST;
1674         rt->dst.output  = ip6_output;
1675         rt->rt6i_gateway  = fl6->daddr;
1676         rt->rt6i_dst.addr = fl6->daddr;
1677         rt->rt6i_dst.plen = 128;
1678         rt->rt6i_idev     = idev;
1679         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
1680
1681         /* Add this dst into uncached_list so that rt6_ifdown() can
1682          * do proper release of the net_device
1683          */
1684         rt6_uncached_list_add(rt);
1685
1686         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
1687
1688 out:
1689         return dst;
1690 }
1691
1692 static int ip6_dst_gc(struct dst_ops *ops)
1693 {
1694         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
1695         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
1696         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
1697         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
1698         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
1699         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
1700         int entries;
1701
1702         entries = dst_entries_get_fast(ops);
1703         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
1704             entries <= rt_max_size)
1705                 goto out;
1706
1707         net->ipv6.ip6_rt_gc_expire++;
1708         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
1709         entries = dst_entries_get_slow(ops);
1710         if (entries < ops->gc_thresh)
1711                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
1712 out:
1713         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
1714         return entries > rt_max_size;
1715 }
1716
1717 static int ip6_convert_metrics(struct mx6_config *mxc,
1718                                const struct fib6_config *cfg)
1719 {
1720         bool ecn_ca = false;
1721         struct nlattr *nla;
1722         int remaining;
1723         u32 *mp;
1724
1725         if (!cfg->fc_mx)
1726                 return 0;
1727
1728         mp = kzalloc(sizeof(u32) * RTAX_MAX, GFP_KERNEL);
1729         if (unlikely(!mp))
1730                 return -ENOMEM;
1731
1732         nla_for_each_attr(nla, cfg->fc_mx, cfg->fc_mx_len, remaining) {
1733                 int type = nla_type(nla);
1734                 u32 val;
1735
1736                 if (!type)
1737                         continue;
1738                 if (unlikely(type > RTAX_MAX))
1739                         goto err;
1740
1741                 if (type == RTAX_CC_ALGO) {
1742                         char tmp[TCP_CA_NAME_MAX];
1743
1744                         nla_strlcpy(tmp, nla, sizeof(tmp));
1745                         val = tcp_ca_get_key_by_name(tmp, &ecn_ca);
1746                         if (val == TCP_CA_UNSPEC)
1747                                 goto err;
1748                 } else {
1749                         val = nla_get_u32(nla);
1750                 }
1751                 if (type == RTAX_HOPLIMIT && val > 255)
1752                         val = 255;
1753                 if (type == RTAX_FEATURES && (val & ~RTAX_FEATURE_MASK))
1754                         goto err;
1755
1756                 mp[type - 1] = val;
1757                 __set_bit(type - 1, mxc->mx_valid);
1758         }
1759
1760         if (ecn_ca) {
1761                 __set_bit(RTAX_FEATURES - 1, mxc->mx_valid);
1762                 mp[RTAX_FEATURES - 1] |= DST_FEATURE_ECN_CA;
1763         }
1764
1765         mxc->mx = mp;
1766         return 0;
1767  err:
1768         kfree(mp);
1769         return -EINVAL;
1770 }
1771
1772 static struct rt6_info *ip6_nh_lookup_table(struct net *net,
1773                                             struct fib6_config *cfg,
1774                                             const struct in6_addr *gw_addr)
1775 {
1776         struct flowi6 fl6 = {
1777                 .flowi6_oif = cfg->fc_ifindex,
1778                 .daddr = *gw_addr,
1779                 .saddr = cfg->fc_prefsrc,
1780         };
1781         struct fib6_table *table;
1782         struct rt6_info *rt;
1783         int flags = RT6_LOOKUP_F_IFACE | RT6_LOOKUP_F_IGNORE_LINKSTATE;
1784
1785         table = fib6_get_table(net, cfg->fc_table);
1786         if (!table)
1787                 return NULL;
1788
1789         if (!ipv6_addr_any(&cfg->fc_prefsrc))
1790                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1791
1792         rt = ip6_pol_route(net, table, cfg->fc_ifindex, &fl6, flags);
1793
1794         /* if table lookup failed, fall back to full lookup */
1795         if (rt == net->ipv6.ip6_null_entry) {
1796                 ip6_rt_put(rt);
1797                 rt = NULL;
1798         }
1799
1800         return rt;
1801 }
1802
1803 static struct rt6_info *ip6_route_info_create(struct fib6_config *cfg,
1804                                               struct netlink_ext_ack *extack)
1805 {
1806         struct net *net = cfg->fc_nlinfo.nl_net;
1807         struct rt6_info *rt = NULL;
1808         struct net_device *dev = NULL;
1809         struct inet6_dev *idev = NULL;
1810         struct fib6_table *table;
1811         int addr_type;
1812         int err = -EINVAL;
1813
1814         /* RTF_PCPU is an internal flag; can not be set by userspace */
1815         if (cfg->fc_flags & RTF_PCPU) {
1816                 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
1817                 goto out;
1818         }
1819
1820         if (cfg->fc_dst_len > 128) {
1821                 NL_SET_ERR_MSG(extack, "Invalid prefix length");
1822                 goto out;
1823         }
1824         if (cfg->fc_src_len > 128) {
1825                 NL_SET_ERR_MSG(extack, "Invalid source address length");
1826                 goto out;
1827         }
1828 #ifndef CONFIG_IPV6_SUBTREES
1829         if (cfg->fc_src_len) {
1830                 NL_SET_ERR_MSG(extack,
1831                                "Specifying source address requires IPV6_SUBTREES to be enabled");
1832                 goto out;
1833         }
1834 #endif
1835         if (cfg->fc_ifindex) {
1836                 err = -ENODEV;
1837                 dev = dev_get_by_index(net, cfg->fc_ifindex);
1838                 if (!dev)
1839                         goto out;
1840                 idev = in6_dev_get(dev);
1841                 if (!idev)
1842                         goto out;
1843         }
1844
1845         if (cfg->fc_metric == 0)
1846                 cfg->fc_metric = IP6_RT_PRIO_USER;
1847
1848         err = -ENOBUFS;
1849         if (cfg->fc_nlinfo.nlh &&
1850             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
1851                 table = fib6_get_table(net, cfg->fc_table);
1852                 if (!table) {
1853                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1854                         table = fib6_new_table(net, cfg->fc_table);
1855                 }
1856         } else {
1857                 table = fib6_new_table(net, cfg->fc_table);
1858         }
1859
1860         if (!table)
1861                 goto out;
1862
1863         rt = ip6_dst_alloc(net, NULL,
1864                            (cfg->fc_flags & RTF_ADDRCONF) ? 0 : DST_NOCOUNT);
1865
1866         if (!rt) {
1867                 err = -ENOMEM;
1868                 goto out;
1869         }
1870
1871         if (cfg->fc_flags & RTF_EXPIRES)
1872                 rt6_set_expires(rt, jiffies +
1873                                 clock_t_to_jiffies(cfg->fc_expires));
1874         else
1875                 rt6_clean_expires(rt);
1876
1877         if (cfg->fc_protocol == RTPROT_UNSPEC)
1878                 cfg->fc_protocol = RTPROT_BOOT;
1879         rt->rt6i_protocol = cfg->fc_protocol;
1880
1881         addr_type = ipv6_addr_type(&cfg->fc_dst);
1882
1883         if (addr_type & IPV6_ADDR_MULTICAST)
1884                 rt->dst.input = ip6_mc_input;
1885         else if (cfg->fc_flags & RTF_LOCAL)
1886                 rt->dst.input = ip6_input;
1887         else
1888                 rt->dst.input = ip6_forward;
1889
1890         rt->dst.output = ip6_output;
1891
1892         if (cfg->fc_encap) {
1893                 struct lwtunnel_state *lwtstate;
1894
1895                 err = lwtunnel_build_state(cfg->fc_encap_type,
1896                                            cfg->fc_encap, AF_INET6, cfg,
1897                                            &lwtstate, extack);
1898                 if (err)
1899                         goto out;
1900                 rt->dst.lwtstate = lwtstate_get(lwtstate);
1901                 if (lwtunnel_output_redirect(rt->dst.lwtstate)) {
1902                         rt->dst.lwtstate->orig_output = rt->dst.output;
1903                         rt->dst.output = lwtunnel_output;
1904                 }
1905                 if (lwtunnel_input_redirect(rt->dst.lwtstate)) {
1906                         rt->dst.lwtstate->orig_input = rt->dst.input;
1907                         rt->dst.input = lwtunnel_input;
1908                 }
1909         }
1910
1911         ipv6_addr_prefix(&rt->rt6i_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
1912         rt->rt6i_dst.plen = cfg->fc_dst_len;
1913         if (rt->rt6i_dst.plen == 128)
1914                 rt->dst.flags |= DST_HOST;
1915
1916 #ifdef CONFIG_IPV6_SUBTREES
1917         ipv6_addr_prefix(&rt->rt6i_src.addr, &cfg->fc_src, cfg->fc_src_len);
1918         rt->rt6i_src.plen = cfg->fc_src_len;
1919 #endif
1920
1921         rt->rt6i_metric = cfg->fc_metric;
1922
1923         /* We cannot add true routes via loopback here,
1924            they would result in kernel looping; promote them to reject routes
1925          */
1926         if ((cfg->fc_flags & RTF_REJECT) ||
1927             (dev && (dev->flags & IFF_LOOPBACK) &&
1928              !(addr_type & IPV6_ADDR_LOOPBACK) &&
1929              !(cfg->fc_flags & RTF_LOCAL))) {
1930                 /* hold loopback dev/idev if we haven't done so. */
1931                 if (dev != net->loopback_dev) {
1932                         if (dev) {
1933                                 dev_put(dev);
1934                                 in6_dev_put(idev);
1935                         }
1936                         dev = net->loopback_dev;
1937                         dev_hold(dev);
1938                         idev = in6_dev_get(dev);
1939                         if (!idev) {
1940                                 err = -ENODEV;
1941                                 goto out;
1942                         }
1943                 }
1944                 rt->rt6i_flags = RTF_REJECT|RTF_NONEXTHOP;
1945                 switch (cfg->fc_type) {
1946                 case RTN_BLACKHOLE:
1947                         rt->dst.error = -EINVAL;
1948                         rt->dst.output = dst_discard_out;
1949                         rt->dst.input = dst_discard;
1950                         break;
1951                 case RTN_PROHIBIT:
1952                         rt->dst.error = -EACCES;
1953                         rt->dst.output = ip6_pkt_prohibit_out;
1954                         rt->dst.input = ip6_pkt_prohibit;
1955                         break;
1956                 case RTN_THROW:
1957                 case RTN_UNREACHABLE:
1958                 default:
1959                         rt->dst.error = (cfg->fc_type == RTN_THROW) ? -EAGAIN
1960                                         : (cfg->fc_type == RTN_UNREACHABLE)
1961                                         ? -EHOSTUNREACH : -ENETUNREACH;
1962                         rt->dst.output = ip6_pkt_discard_out;
1963                         rt->dst.input = ip6_pkt_discard;
1964                         break;
1965                 }
1966                 goto install_route;
1967         }
1968
1969         if (cfg->fc_flags & RTF_GATEWAY) {
1970                 const struct in6_addr *gw_addr;
1971                 int gwa_type;
1972
1973                 gw_addr = &cfg->fc_gateway;
1974                 gwa_type = ipv6_addr_type(gw_addr);
1975
1976                 /* if gw_addr is local we will fail to detect this in case
1977                  * address is still TENTATIVE (DAD in progress). rt6_lookup()
1978                  * will return already-added prefix route via interface that
1979                  * prefix route was assigned to, which might be non-loopback.
1980                  */
1981                 err = -EINVAL;
1982                 if (ipv6_chk_addr_and_flags(net, gw_addr,
1983                                             gwa_type & IPV6_ADDR_LINKLOCAL ?
1984                                             dev : NULL, 0, 0)) {
1985                         NL_SET_ERR_MSG(extack, "Invalid gateway address");
1986                         goto out;
1987                 }
1988                 rt->rt6i_gateway = *gw_addr;
1989
1990                 if (gwa_type != (IPV6_ADDR_LINKLOCAL|IPV6_ADDR_UNICAST)) {
1991                         struct rt6_info *grt = NULL;
1992
1993                         /* IPv6 strictly inhibits using not link-local
1994                            addresses as nexthop address.
1995                            Otherwise, router will not able to send redirects.
1996                            It is very good, but in some (rare!) circumstances
1997                            (SIT, PtP, NBMA NOARP links) it is handy to allow
1998                            some exceptions. --ANK
1999                            We allow IPv4-mapped nexthops to support RFC4798-type
2000                            addressing
2001                          */
2002                         if (!(gwa_type & (IPV6_ADDR_UNICAST |
2003                                           IPV6_ADDR_MAPPED))) {
2004                                 NL_SET_ERR_MSG(extack,
2005                                                "Invalid gateway address");
2006                                 goto out;
2007                         }
2008
2009                         if (cfg->fc_table) {
2010                                 grt = ip6_nh_lookup_table(net, cfg, gw_addr);
2011
2012                                 if (grt) {
2013                                         if (grt->rt6i_flags & RTF_GATEWAY ||
2014                                             (dev && dev != grt->dst.dev)) {
2015                                                 ip6_rt_put(grt);
2016                                                 grt = NULL;
2017                                         }
2018                                 }
2019                         }
2020
2021                         if (!grt)
2022                                 grt = rt6_lookup(net, gw_addr, NULL,
2023                                                  cfg->fc_ifindex, 1);
2024
2025                         err = -EHOSTUNREACH;
2026                         if (!grt)
2027                                 goto out;
2028                         if (dev) {
2029                                 if (dev != grt->dst.dev) {
2030                                         ip6_rt_put(grt);
2031                                         goto out;
2032                                 }
2033                         } else {
2034                                 dev = grt->dst.dev;
2035                                 idev = grt->rt6i_idev;
2036                                 dev_hold(dev);
2037                                 in6_dev_hold(grt->rt6i_idev);
2038                         }
2039                         if (!(grt->rt6i_flags & RTF_GATEWAY))
2040                                 err = 0;
2041                         ip6_rt_put(grt);
2042
2043                         if (err)
2044                                 goto out;
2045                 }
2046                 err = -EINVAL;
2047                 if (!dev) {
2048                         NL_SET_ERR_MSG(extack, "Egress device not specified");
2049                         goto out;
2050                 } else if (dev->flags & IFF_LOOPBACK) {
2051                         NL_SET_ERR_MSG(extack,
2052                                        "Egress device can not be loopback device for this route");
2053                         goto out;
2054                 }
2055         }
2056
2057         err = -ENODEV;
2058         if (!dev)
2059                 goto out;
2060
2061         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
2062                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
2063                         NL_SET_ERR_MSG(extack, "Invalid source address");
2064                         err = -EINVAL;
2065                         goto out;
2066                 }
2067                 rt->rt6i_prefsrc.addr = cfg->fc_prefsrc;
2068                 rt->rt6i_prefsrc.plen = 128;
2069         } else
2070                 rt->rt6i_prefsrc.plen = 0;
2071
2072         rt->rt6i_flags = cfg->fc_flags;
2073
2074 install_route:
2075         rt->dst.dev = dev;
2076         rt->rt6i_idev = idev;
2077         rt->rt6i_table = table;
2078
2079         cfg->fc_nlinfo.nl_net = dev_net(dev);
2080
2081         return rt;
2082 out:
2083         if (dev)
2084                 dev_put(dev);
2085         if (idev)
2086                 in6_dev_put(idev);
2087         if (rt)
2088                 dst_release_immediate(&rt->dst);
2089
2090         return ERR_PTR(err);
2091 }
2092
2093 int ip6_route_add(struct fib6_config *cfg,
2094                   struct netlink_ext_ack *extack)
2095 {
2096         struct mx6_config mxc = { .mx = NULL, };
2097         struct rt6_info *rt;
2098         int err;
2099
2100         rt = ip6_route_info_create(cfg, extack);
2101         if (IS_ERR(rt)) {
2102                 err = PTR_ERR(rt);
2103                 rt = NULL;
2104                 goto out;
2105         }
2106
2107         err = ip6_convert_metrics(&mxc, cfg);
2108         if (err)
2109                 goto out;
2110
2111         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, &mxc, extack);
2112
2113         kfree(mxc.mx);
2114
2115         return err;
2116 out:
2117         if (rt)
2118                 dst_release_immediate(&rt->dst);
2119
2120         return err;
2121 }
2122
2123 static int __ip6_del_rt(struct rt6_info *rt, struct nl_info *info)
2124 {
2125         int err;
2126         struct fib6_table *table;
2127         struct net *net = dev_net(rt->dst.dev);
2128
2129         if (rt == net->ipv6.ip6_null_entry) {
2130                 err = -ENOENT;
2131                 goto out;
2132         }
2133
2134         table = rt->rt6i_table;
2135         write_lock_bh(&table->tb6_lock);
2136         err = fib6_del(rt, info);
2137         write_unlock_bh(&table->tb6_lock);
2138
2139 out:
2140         ip6_rt_put(rt);
2141         return err;
2142 }
2143
2144 int ip6_del_rt(struct rt6_info *rt)
2145 {
2146         struct nl_info info = {
2147                 .nl_net = dev_net(rt->dst.dev),
2148         };
2149         return __ip6_del_rt(rt, &info);
2150 }
2151
2152 static int __ip6_del_rt_siblings(struct rt6_info *rt, struct fib6_config *cfg)
2153 {
2154         struct nl_info *info = &cfg->fc_nlinfo;
2155         struct net *net = info->nl_net;
2156         struct sk_buff *skb = NULL;
2157         struct fib6_table *table;
2158         int err = -ENOENT;
2159
2160         if (rt == net->ipv6.ip6_null_entry)
2161                 goto out_put;
2162         table = rt->rt6i_table;
2163         write_lock_bh(&table->tb6_lock);
2164
2165         if (rt->rt6i_nsiblings && cfg->fc_delete_all_nh) {
2166                 struct rt6_info *sibling, *next_sibling;
2167
2168                 /* prefer to send a single notification with all hops */
2169                 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
2170                 if (skb) {
2171                         u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
2172
2173                         if (rt6_fill_node(net, skb, rt,
2174                                           NULL, NULL, 0, RTM_DELROUTE,
2175                                           info->portid, seq, 0) < 0) {
2176                                 kfree_skb(skb);
2177                                 skb = NULL;
2178                         } else
2179                                 info->skip_notify = 1;
2180                 }
2181
2182                 list_for_each_entry_safe(sibling, next_sibling,
2183                                          &rt->rt6i_siblings,
2184                                          rt6i_siblings) {
2185                         err = fib6_del(sibling, info);
2186                         if (err)
2187                                 goto out_unlock;
2188                 }
2189         }
2190
2191         err = fib6_del(rt, info);
2192 out_unlock:
2193         write_unlock_bh(&table->tb6_lock);
2194 out_put:
2195         ip6_rt_put(rt);
2196
2197         if (skb) {
2198                 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
2199                             info->nlh, gfp_any());
2200         }
2201         return err;
2202 }
2203
2204 static int ip6_route_del(struct fib6_config *cfg,
2205                          struct netlink_ext_ack *extack)
2206 {
2207         struct fib6_table *table;
2208         struct fib6_node *fn;
2209         struct rt6_info *rt;
2210         int err = -ESRCH;
2211
2212         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
2213         if (!table) {
2214                 NL_SET_ERR_MSG(extack, "FIB table does not exist");
2215                 return err;
2216         }
2217
2218         read_lock_bh(&table->tb6_lock);
2219
2220         fn = fib6_locate(&table->tb6_root,
2221                          &cfg->fc_dst, cfg->fc_dst_len,
2222                          &cfg->fc_src, cfg->fc_src_len);
2223
2224         if (fn) {
2225                 for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2226                         if ((rt->rt6i_flags & RTF_CACHE) &&
2227                             !(cfg->fc_flags & RTF_CACHE))
2228                                 continue;
2229                         if (cfg->fc_ifindex &&
2230                             (!rt->dst.dev ||
2231                              rt->dst.dev->ifindex != cfg->fc_ifindex))
2232                                 continue;
2233                         if (cfg->fc_flags & RTF_GATEWAY &&
2234                             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
2235                                 continue;
2236                         if (cfg->fc_metric && cfg->fc_metric != rt->rt6i_metric)
2237                                 continue;
2238                         if (cfg->fc_protocol && cfg->fc_protocol != rt->rt6i_protocol)
2239                                 continue;
2240                         dst_hold(&rt->dst);
2241                         read_unlock_bh(&table->tb6_lock);
2242
2243                         /* if gateway was specified only delete the one hop */
2244                         if (cfg->fc_flags & RTF_GATEWAY)
2245                                 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
2246
2247                         return __ip6_del_rt_siblings(rt, cfg);
2248                 }
2249         }
2250         read_unlock_bh(&table->tb6_lock);
2251
2252         return err;
2253 }
2254
2255 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
2256 {
2257         struct netevent_redirect netevent;
2258         struct rt6_info *rt, *nrt = NULL;
2259         struct ndisc_options ndopts;
2260         struct inet6_dev *in6_dev;
2261         struct neighbour *neigh;
2262         struct rd_msg *msg;
2263         int optlen, on_link;
2264         u8 *lladdr;
2265
2266         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
2267         optlen -= sizeof(*msg);
2268
2269         if (optlen < 0) {
2270                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2271                 return;
2272         }
2273
2274         msg = (struct rd_msg *)icmp6_hdr(skb);
2275
2276         if (ipv6_addr_is_multicast(&msg->dest)) {
2277                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2278                 return;
2279         }
2280
2281         on_link = 0;
2282         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
2283                 on_link = 1;
2284         } else if (ipv6_addr_type(&msg->target) !=
2285                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
2286                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2287                 return;
2288         }
2289
2290         in6_dev = __in6_dev_get(skb->dev);
2291         if (!in6_dev)
2292                 return;
2293         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
2294                 return;
2295
2296         /* RFC2461 8.1:
2297          *      The IP source address of the Redirect MUST be the same as the current
2298          *      first-hop router for the specified ICMP Destination Address.
2299          */
2300
2301         if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
2302                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2303                 return;
2304         }
2305
2306         lladdr = NULL;
2307         if (ndopts.nd_opts_tgt_lladdr) {
2308                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
2309                                              skb->dev);
2310                 if (!lladdr) {
2311                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2312                         return;
2313                 }
2314         }
2315
2316         rt = (struct rt6_info *) dst;
2317         if (rt->rt6i_flags & RTF_REJECT) {
2318                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2319                 return;
2320         }
2321
2322         /* Redirect received -> path was valid.
2323          * Look, redirects are sent only in response to data packets,
2324          * so that this nexthop apparently is reachable. --ANK
2325          */
2326         dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
2327
2328         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
2329         if (!neigh)
2330                 return;
2331
2332         /*
2333          *      We have finally decided to accept it.
2334          */
2335
2336         ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
2337                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
2338                      NEIGH_UPDATE_F_OVERRIDE|
2339                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
2340                                      NEIGH_UPDATE_F_ISROUTER)),
2341                      NDISC_REDIRECT, &ndopts);
2342
2343         nrt = ip6_rt_cache_alloc(rt, &msg->dest, NULL);
2344         if (!nrt)
2345                 goto out;
2346
2347         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
2348         if (on_link)
2349                 nrt->rt6i_flags &= ~RTF_GATEWAY;
2350
2351         nrt->rt6i_protocol = RTPROT_REDIRECT;
2352         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
2353
2354         if (ip6_ins_rt(nrt))
2355                 goto out_release;
2356
2357         netevent.old = &rt->dst;
2358         netevent.new = &nrt->dst;
2359         netevent.daddr = &msg->dest;
2360         netevent.neigh = neigh;
2361         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
2362
2363         if (rt->rt6i_flags & RTF_CACHE) {
2364                 rt = (struct rt6_info *) dst_clone(&rt->dst);
2365                 ip6_del_rt(rt);
2366         }
2367
2368 out_release:
2369         /* Release the reference taken in
2370          * ip6_rt_cache_alloc()
2371          */
2372         dst_release(&nrt->dst);
2373
2374 out:
2375         neigh_release(neigh);
2376 }
2377
2378 /*
2379  *      Misc support functions
2380  */
2381
2382 static void rt6_set_from(struct rt6_info *rt, struct rt6_info *from)
2383 {
2384         BUG_ON(from->dst.from);
2385
2386         rt->rt6i_flags &= ~RTF_EXPIRES;
2387         dst_hold(&from->dst);
2388         rt->dst.from = &from->dst;
2389         dst_init_metrics(&rt->dst, dst_metrics_ptr(&from->dst), true);
2390 }
2391
2392 static void ip6_rt_copy_init(struct rt6_info *rt, struct rt6_info *ort)
2393 {
2394         rt->dst.input = ort->dst.input;
2395         rt->dst.output = ort->dst.output;
2396         rt->rt6i_dst = ort->rt6i_dst;
2397         rt->dst.error = ort->dst.error;
2398         rt->rt6i_idev = ort->rt6i_idev;
2399         if (rt->rt6i_idev)
2400                 in6_dev_hold(rt->rt6i_idev);
2401         rt->dst.lastuse = jiffies;
2402         rt->rt6i_gateway = ort->rt6i_gateway;
2403         rt->rt6i_flags = ort->rt6i_flags;
2404         rt6_set_from(rt, ort);
2405         rt->rt6i_metric = ort->rt6i_metric;
2406 #ifdef CONFIG_IPV6_SUBTREES
2407         rt->rt6i_src = ort->rt6i_src;
2408 #endif
2409         rt->rt6i_prefsrc = ort->rt6i_prefsrc;
2410         rt->rt6i_table = ort->rt6i_table;
2411         rt->dst.lwtstate = lwtstate_get(ort->dst.lwtstate);
2412 }
2413
2414 #ifdef CONFIG_IPV6_ROUTE_INFO
2415 static struct rt6_info *rt6_get_route_info(struct net *net,
2416                                            const struct in6_addr *prefix, int prefixlen,
2417                                            const struct in6_addr *gwaddr,
2418                                            struct net_device *dev)
2419 {
2420         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
2421         int ifindex = dev->ifindex;
2422         struct fib6_node *fn;
2423         struct rt6_info *rt = NULL;
2424         struct fib6_table *table;
2425
2426         table = fib6_get_table(net, tb_id);
2427         if (!table)
2428                 return NULL;
2429
2430         read_lock_bh(&table->tb6_lock);
2431         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0);
2432         if (!fn)
2433                 goto out;
2434
2435         for (rt = fn->leaf; rt; rt = rt->dst.rt6_next) {
2436                 if (rt->dst.dev->ifindex != ifindex)
2437                         continue;
2438                 if ((rt->rt6i_flags & (RTF_ROUTEINFO|RTF_GATEWAY)) != (RTF_ROUTEINFO|RTF_GATEWAY))
2439                         continue;
2440                 if (!ipv6_addr_equal(&rt->rt6i_gateway, gwaddr))
2441                         continue;
2442                 dst_hold(&rt->dst);
2443                 break;
2444         }
2445 out:
2446         read_unlock_bh(&table->tb6_lock);
2447         return rt;
2448 }
2449
2450 static struct rt6_info *rt6_add_route_info(struct net *net,
2451                                            const struct in6_addr *prefix, int prefixlen,
2452                                            const struct in6_addr *gwaddr,
2453                                            struct net_device *dev,
2454                                            unsigned int pref)
2455 {
2456         struct fib6_config cfg = {
2457                 .fc_metric      = IP6_RT_PRIO_USER,
2458                 .fc_ifindex     = dev->ifindex,
2459                 .fc_dst_len     = prefixlen,
2460                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
2461                                   RTF_UP | RTF_PREF(pref),
2462                 .fc_protocol = RTPROT_RA,
2463                 .fc_nlinfo.portid = 0,
2464                 .fc_nlinfo.nlh = NULL,
2465                 .fc_nlinfo.nl_net = net,
2466         };
2467
2468         cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
2469         cfg.fc_dst = *prefix;
2470         cfg.fc_gateway = *gwaddr;
2471
2472         /* We should treat it as a default route if prefix length is 0. */
2473         if (!prefixlen)
2474                 cfg.fc_flags |= RTF_DEFAULT;
2475
2476         ip6_route_add(&cfg, NULL);
2477
2478         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
2479 }
2480 #endif
2481
2482 struct rt6_info *rt6_get_dflt_router(const struct in6_addr *addr, struct net_device *dev)
2483 {
2484         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
2485         struct rt6_info *rt;
2486         struct fib6_table *table;
2487
2488         table = fib6_get_table(dev_net(dev), tb_id);
2489         if (!table)
2490                 return NULL;
2491
2492         read_lock_bh(&table->tb6_lock);
2493         for (rt = table->tb6_root.leaf; rt; rt = rt->dst.rt6_next) {
2494                 if (dev == rt->dst.dev &&
2495                     ((rt->rt6i_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
2496                     ipv6_addr_equal(&rt->rt6i_gateway, addr))
2497                         break;
2498         }
2499         if (rt)
2500                 dst_hold(&rt->dst);
2501         read_unlock_bh(&table->tb6_lock);
2502         return rt;
2503 }
2504
2505 struct rt6_info *rt6_add_dflt_router(const struct in6_addr *gwaddr,
2506                                      struct net_device *dev,
2507                                      unsigned int pref)
2508 {
2509         struct fib6_config cfg = {
2510                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
2511                 .fc_metric      = IP6_RT_PRIO_USER,
2512                 .fc_ifindex     = dev->ifindex,
2513                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
2514                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
2515                 .fc_protocol = RTPROT_RA,
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
3428         if (rt->rt6i_flags & RTF_CACHE)
3429                 rtm->rtm_flags |= RTM_F_CLONED;
3430
3431         if (dst) {
3432                 if (nla_put_in6_addr(skb, RTA_DST, dst))
3433                         goto nla_put_failure;
3434                 rtm->rtm_dst_len = 128;
3435         } else if (rtm->rtm_dst_len)
3436                 if (nla_put_in6_addr(skb, RTA_DST, &rt->rt6i_dst.addr))
3437                         goto nla_put_failure;
3438 #ifdef CONFIG_IPV6_SUBTREES
3439         if (src) {
3440                 if (nla_put_in6_addr(skb, RTA_SRC, src))
3441                         goto nla_put_failure;
3442                 rtm->rtm_src_len = 128;
3443         } else if (rtm->rtm_src_len &&
3444                    nla_put_in6_addr(skb, RTA_SRC, &rt->rt6i_src.addr))
3445                 goto nla_put_failure;
3446 #endif
3447         if (iif) {
3448 #ifdef CONFIG_IPV6_MROUTE
3449                 if (ipv6_addr_is_multicast(&rt->rt6i_dst.addr)) {
3450                         int err = ip6mr_get_route(net, skb, rtm, portid);
3451
3452                         if (err == 0)
3453                                 return 0;
3454                         if (err < 0)
3455                                 goto nla_put_failure;
3456                 } else
3457 #endif
3458                         if (nla_put_u32(skb, RTA_IIF, iif))
3459                                 goto nla_put_failure;
3460         } else if (dst) {
3461                 struct in6_addr saddr_buf;
3462                 if (ip6_route_get_saddr(net, rt, dst, 0, &saddr_buf) == 0 &&
3463                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3464                         goto nla_put_failure;
3465         }
3466
3467         if (rt->rt6i_prefsrc.plen) {
3468                 struct in6_addr saddr_buf;
3469                 saddr_buf = rt->rt6i_prefsrc.addr;
3470                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
3471                         goto nla_put_failure;
3472         }
3473
3474         memcpy(metrics, dst_metrics_ptr(&rt->dst), sizeof(metrics));
3475         if (rt->rt6i_pmtu)
3476                 metrics[RTAX_MTU - 1] = rt->rt6i_pmtu;
3477         if (rtnetlink_put_metrics(skb, metrics) < 0)
3478                 goto nla_put_failure;
3479
3480         if (nla_put_u32(skb, RTA_PRIORITY, rt->rt6i_metric))
3481                 goto nla_put_failure;
3482
3483         /* For multipath routes, walk the siblings list and add
3484          * each as a nexthop within RTA_MULTIPATH.
3485          */
3486         if (rt->rt6i_nsiblings) {
3487                 struct rt6_info *sibling, *next_sibling;
3488                 struct nlattr *mp;
3489
3490                 mp = nla_nest_start(skb, RTA_MULTIPATH);
3491                 if (!mp)
3492                         goto nla_put_failure;
3493
3494                 if (rt6_add_nexthop(skb, rt) < 0)
3495                         goto nla_put_failure;
3496
3497                 list_for_each_entry_safe(sibling, next_sibling,
3498                                          &rt->rt6i_siblings, rt6i_siblings) {
3499                         if (rt6_add_nexthop(skb, sibling) < 0)
3500                                 goto nla_put_failure;
3501                 }
3502
3503                 nla_nest_end(skb, mp);
3504         } else {
3505                 if (rt6_nexthop_info(skb, rt, &rtm->rtm_flags, false) < 0)
3506                         goto nla_put_failure;
3507         }
3508
3509         expires = (rt->rt6i_flags & RTF_EXPIRES) ? rt->dst.expires - jiffies : 0;
3510
3511         if (rtnl_put_cacheinfo(skb, &rt->dst, 0, expires, rt->dst.error) < 0)
3512                 goto nla_put_failure;
3513
3514         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt->rt6i_flags)))
3515                 goto nla_put_failure;
3516
3517
3518         nlmsg_end(skb, nlh);
3519         return 0;
3520
3521 nla_put_failure:
3522         nlmsg_cancel(skb, nlh);
3523         return -EMSGSIZE;
3524 }
3525
3526 int rt6_dump_route(struct rt6_info *rt, void *p_arg)
3527 {
3528         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
3529         struct net *net = arg->net;
3530
3531         if (rt == net->ipv6.ip6_null_entry)
3532                 return 0;
3533
3534         if (nlmsg_len(arg->cb->nlh) >= sizeof(struct rtmsg)) {
3535                 struct rtmsg *rtm = nlmsg_data(arg->cb->nlh);
3536
3537                 /* user wants prefix routes only */
3538                 if (rtm->rtm_flags & RTM_F_PREFIX &&
3539                     !(rt->rt6i_flags & RTF_PREFIX_RT)) {
3540                         /* success since this is not a prefix route */
3541                         return 1;
3542                 }
3543         }
3544
3545         return rt6_fill_node(net,
3546                      arg->skb, rt, NULL, NULL, 0, RTM_NEWROUTE,
3547                      NETLINK_CB(arg->cb->skb).portid, arg->cb->nlh->nlmsg_seq,
3548                      NLM_F_MULTI);
3549 }
3550
3551 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
3552                               struct netlink_ext_ack *extack)
3553 {
3554         struct net *net = sock_net(in_skb->sk);
3555         struct nlattr *tb[RTA_MAX+1];
3556         int err, iif = 0, oif = 0;
3557         struct dst_entry *dst;
3558         struct rt6_info *rt;
3559         struct sk_buff *skb;
3560         struct rtmsg *rtm;
3561         struct flowi6 fl6;
3562         bool fibmatch;
3563
3564         err = nlmsg_parse(nlh, sizeof(*rtm), tb, RTA_MAX, rtm_ipv6_policy,
3565                           extack);
3566         if (err < 0)
3567                 goto errout;
3568
3569         err = -EINVAL;
3570         memset(&fl6, 0, sizeof(fl6));
3571         rtm = nlmsg_data(nlh);
3572         fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
3573         fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
3574
3575         if (tb[RTA_SRC]) {
3576                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
3577                         goto errout;
3578
3579                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
3580         }
3581
3582         if (tb[RTA_DST]) {
3583                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
3584                         goto errout;
3585
3586                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
3587         }
3588
3589         if (tb[RTA_IIF])
3590                 iif = nla_get_u32(tb[RTA_IIF]);
3591
3592         if (tb[RTA_OIF])
3593                 oif = nla_get_u32(tb[RTA_OIF]);
3594
3595         if (tb[RTA_MARK])
3596                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
3597
3598         if (tb[RTA_UID])
3599                 fl6.flowi6_uid = make_kuid(current_user_ns(),
3600                                            nla_get_u32(tb[RTA_UID]));
3601         else
3602                 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
3603
3604         if (iif) {
3605                 struct net_device *dev;
3606                 int flags = 0;
3607
3608                 dev = __dev_get_by_index(net, iif);
3609                 if (!dev) {
3610                         err = -ENODEV;
3611                         goto errout;
3612                 }
3613
3614                 fl6.flowi6_iif = iif;
3615
3616                 if (!ipv6_addr_any(&fl6.saddr))
3617                         flags |= RT6_LOOKUP_F_HAS_SADDR;
3618
3619                 if (!fibmatch)
3620                         dst = ip6_route_input_lookup(net, dev, &fl6, flags);
3621         } else {
3622                 fl6.flowi6_oif = oif;
3623
3624                 if (!fibmatch)
3625                         dst = ip6_route_output(net, NULL, &fl6);
3626         }
3627
3628         if (fibmatch)
3629                 dst = ip6_route_lookup(net, &fl6, 0);
3630
3631         rt = container_of(dst, struct rt6_info, dst);
3632         if (rt->dst.error) {
3633                 err = rt->dst.error;
3634                 ip6_rt_put(rt);
3635                 goto errout;
3636         }
3637
3638         if (rt == net->ipv6.ip6_null_entry) {
3639                 err = rt->dst.error;
3640                 ip6_rt_put(rt);
3641                 goto errout;
3642         }
3643
3644         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
3645         if (!skb) {
3646                 ip6_rt_put(rt);
3647                 err = -ENOBUFS;
3648                 goto errout;
3649         }
3650
3651         skb_dst_set(skb, &rt->dst);
3652         if (fibmatch)
3653                 err = rt6_fill_node(net, skb, rt, NULL, NULL, iif,
3654                                     RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3655                                     nlh->nlmsg_seq, 0);
3656         else
3657                 err = rt6_fill_node(net, skb, rt, &fl6.daddr, &fl6.saddr, iif,
3658                                     RTM_NEWROUTE, NETLINK_CB(in_skb).portid,
3659                                     nlh->nlmsg_seq, 0);
3660         if (err < 0) {
3661                 kfree_skb(skb);
3662                 goto errout;
3663         }
3664
3665         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
3666 errout:
3667         return err;
3668 }
3669
3670 void inet6_rt_notify(int event, struct rt6_info *rt, struct nl_info *info,
3671                      unsigned int nlm_flags)
3672 {
3673         struct sk_buff *skb;
3674         struct net *net = info->nl_net;
3675         u32 seq;
3676         int err;
3677
3678         err = -ENOBUFS;
3679         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3680
3681         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3682         if (!skb)
3683                 goto errout;
3684
3685         err = rt6_fill_node(net, skb, rt, NULL, NULL, 0,
3686                                 event, info->portid, seq, nlm_flags);
3687         if (err < 0) {
3688                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3689                 WARN_ON(err == -EMSGSIZE);
3690                 kfree_skb(skb);
3691                 goto errout;
3692         }
3693         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3694                     info->nlh, gfp_any());
3695         return;
3696 errout:
3697         if (err < 0)
3698                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
3699 }
3700
3701 static int ip6_route_dev_notify(struct notifier_block *this,
3702                                 unsigned long event, void *ptr)
3703 {
3704         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
3705         struct net *net = dev_net(dev);
3706
3707         if (!(dev->flags & IFF_LOOPBACK))
3708                 return NOTIFY_OK;
3709
3710         if (event == NETDEV_REGISTER) {
3711                 net->ipv6.ip6_null_entry->dst.dev = dev;
3712                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
3713 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3714                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
3715                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
3716                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
3717                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
3718 #endif
3719          } else if (event == NETDEV_UNREGISTER &&
3720                     dev->reg_state != NETREG_UNREGISTERED) {
3721                 /* NETDEV_UNREGISTER could be fired for multiple times by
3722                  * netdev_wait_allrefs(). Make sure we only call this once.
3723                  */
3724                 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
3725 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3726                 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
3727                 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
3728 #endif
3729         }
3730
3731         return NOTIFY_OK;
3732 }
3733
3734 /*
3735  *      /proc
3736  */
3737
3738 #ifdef CONFIG_PROC_FS
3739
3740 static const struct file_operations ipv6_route_proc_fops = {
3741         .owner          = THIS_MODULE,
3742         .open           = ipv6_route_open,
3743         .read           = seq_read,
3744         .llseek         = seq_lseek,
3745         .release        = seq_release_net,
3746 };
3747
3748 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
3749 {
3750         struct net *net = (struct net *)seq->private;
3751         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
3752                    net->ipv6.rt6_stats->fib_nodes,
3753                    net->ipv6.rt6_stats->fib_route_nodes,
3754                    net->ipv6.rt6_stats->fib_rt_alloc,
3755                    net->ipv6.rt6_stats->fib_rt_entries,
3756                    net->ipv6.rt6_stats->fib_rt_cache,
3757                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
3758                    net->ipv6.rt6_stats->fib_discarded_routes);
3759
3760         return 0;
3761 }
3762
3763 static int rt6_stats_seq_open(struct inode *inode, struct file *file)
3764 {
3765         return single_open_net(inode, file, rt6_stats_seq_show);
3766 }
3767
3768 static const struct file_operations rt6_stats_seq_fops = {
3769         .owner   = THIS_MODULE,
3770         .open    = rt6_stats_seq_open,
3771         .read    = seq_read,
3772         .llseek  = seq_lseek,
3773         .release = single_release_net,
3774 };
3775 #endif  /* CONFIG_PROC_FS */
3776
3777 #ifdef CONFIG_SYSCTL
3778
3779 static
3780 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
3781                               void __user *buffer, size_t *lenp, loff_t *ppos)
3782 {
3783         struct net *net;
3784         int delay;
3785         if (!write)
3786                 return -EINVAL;
3787
3788         net = (struct net *)ctl->extra1;
3789         delay = net->ipv6.sysctl.flush_delay;
3790         proc_dointvec(ctl, write, buffer, lenp, ppos);
3791         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
3792         return 0;
3793 }
3794
3795 struct ctl_table ipv6_route_table_template[] = {
3796         {
3797                 .procname       =       "flush",
3798                 .data           =       &init_net.ipv6.sysctl.flush_delay,
3799                 .maxlen         =       sizeof(int),
3800                 .mode           =       0200,
3801                 .proc_handler   =       ipv6_sysctl_rtcache_flush
3802         },
3803         {
3804                 .procname       =       "gc_thresh",
3805                 .data           =       &ip6_dst_ops_template.gc_thresh,
3806                 .maxlen         =       sizeof(int),
3807                 .mode           =       0644,
3808                 .proc_handler   =       proc_dointvec,
3809         },
3810         {
3811                 .procname       =       "max_size",
3812                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
3813                 .maxlen         =       sizeof(int),
3814                 .mode           =       0644,
3815                 .proc_handler   =       proc_dointvec,
3816         },
3817         {
3818                 .procname       =       "gc_min_interval",
3819                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3820                 .maxlen         =       sizeof(int),
3821                 .mode           =       0644,
3822                 .proc_handler   =       proc_dointvec_jiffies,
3823         },
3824         {
3825                 .procname       =       "gc_timeout",
3826                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
3827                 .maxlen         =       sizeof(int),
3828                 .mode           =       0644,
3829                 .proc_handler   =       proc_dointvec_jiffies,
3830         },
3831         {
3832                 .procname       =       "gc_interval",
3833                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
3834                 .maxlen         =       sizeof(int),
3835                 .mode           =       0644,
3836                 .proc_handler   =       proc_dointvec_jiffies,
3837         },
3838         {
3839                 .procname       =       "gc_elasticity",
3840                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
3841                 .maxlen         =       sizeof(int),
3842                 .mode           =       0644,
3843                 .proc_handler   =       proc_dointvec,
3844         },
3845         {
3846                 .procname       =       "mtu_expires",
3847                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
3848                 .maxlen         =       sizeof(int),
3849                 .mode           =       0644,
3850                 .proc_handler   =       proc_dointvec_jiffies,
3851         },
3852         {
3853                 .procname       =       "min_adv_mss",
3854                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
3855                 .maxlen         =       sizeof(int),
3856                 .mode           =       0644,
3857                 .proc_handler   =       proc_dointvec,
3858         },
3859         {
3860                 .procname       =       "gc_min_interval_ms",
3861                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
3862                 .maxlen         =       sizeof(int),
3863                 .mode           =       0644,
3864                 .proc_handler   =       proc_dointvec_ms_jiffies,
3865         },
3866         { }
3867 };
3868
3869 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
3870 {
3871         struct ctl_table *table;
3872
3873         table = kmemdup(ipv6_route_table_template,
3874                         sizeof(ipv6_route_table_template),
3875                         GFP_KERNEL);
3876
3877         if (table) {
3878                 table[0].data = &net->ipv6.sysctl.flush_delay;
3879                 table[0].extra1 = net;
3880                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
3881                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
3882                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3883                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
3884                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
3885                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
3886                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
3887                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
3888                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
3889
3890                 /* Don't export sysctls to unprivileged users */
3891                 if (net->user_ns != &init_user_ns)
3892                         table[0].procname = NULL;
3893         }
3894
3895         return table;
3896 }
3897 #endif
3898
3899 static int __net_init ip6_route_net_init(struct net *net)
3900 {
3901         int ret = -ENOMEM;
3902
3903         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
3904                sizeof(net->ipv6.ip6_dst_ops));
3905
3906         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
3907                 goto out_ip6_dst_ops;
3908
3909         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
3910                                            sizeof(*net->ipv6.ip6_null_entry),
3911                                            GFP_KERNEL);
3912         if (!net->ipv6.ip6_null_entry)
3913                 goto out_ip6_dst_entries;
3914         net->ipv6.ip6_null_entry->dst.path =
3915                 (struct dst_entry *)net->ipv6.ip6_null_entry;
3916         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3917         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
3918                          ip6_template_metrics, true);
3919
3920 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3921         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
3922                                                sizeof(*net->ipv6.ip6_prohibit_entry),
3923                                                GFP_KERNEL);
3924         if (!net->ipv6.ip6_prohibit_entry)
3925                 goto out_ip6_null_entry;
3926         net->ipv6.ip6_prohibit_entry->dst.path =
3927                 (struct dst_entry *)net->ipv6.ip6_prohibit_entry;
3928         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3929         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
3930                          ip6_template_metrics, true);
3931
3932         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
3933                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
3934                                                GFP_KERNEL);
3935         if (!net->ipv6.ip6_blk_hole_entry)
3936                 goto out_ip6_prohibit_entry;
3937         net->ipv6.ip6_blk_hole_entry->dst.path =
3938                 (struct dst_entry *)net->ipv6.ip6_blk_hole_entry;
3939         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
3940         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
3941                          ip6_template_metrics, true);
3942 #endif
3943
3944         net->ipv6.sysctl.flush_delay = 0;
3945         net->ipv6.sysctl.ip6_rt_max_size = 4096;
3946         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
3947         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
3948         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
3949         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
3950         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
3951         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
3952
3953         net->ipv6.ip6_rt_gc_expire = 30*HZ;
3954
3955         ret = 0;
3956 out:
3957         return ret;
3958
3959 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3960 out_ip6_prohibit_entry:
3961         kfree(net->ipv6.ip6_prohibit_entry);
3962 out_ip6_null_entry:
3963         kfree(net->ipv6.ip6_null_entry);
3964 #endif
3965 out_ip6_dst_entries:
3966         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3967 out_ip6_dst_ops:
3968         goto out;
3969 }
3970
3971 static void __net_exit ip6_route_net_exit(struct net *net)
3972 {
3973         kfree(net->ipv6.ip6_null_entry);
3974 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3975         kfree(net->ipv6.ip6_prohibit_entry);
3976         kfree(net->ipv6.ip6_blk_hole_entry);
3977 #endif
3978         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
3979 }
3980
3981 static int __net_init ip6_route_net_init_late(struct net *net)
3982 {
3983 #ifdef CONFIG_PROC_FS
3984         proc_create("ipv6_route", 0, net->proc_net, &ipv6_route_proc_fops);
3985         proc_create("rt6_stats", S_IRUGO, net->proc_net, &rt6_stats_seq_fops);
3986 #endif
3987         return 0;
3988 }
3989
3990 static void __net_exit ip6_route_net_exit_late(struct net *net)
3991 {
3992 #ifdef CONFIG_PROC_FS
3993         remove_proc_entry("ipv6_route", net->proc_net);
3994         remove_proc_entry("rt6_stats", net->proc_net);
3995 #endif
3996 }
3997
3998 static struct pernet_operations ip6_route_net_ops = {
3999         .init = ip6_route_net_init,
4000         .exit = ip6_route_net_exit,
4001 };
4002
4003 static int __net_init ipv6_inetpeer_init(struct net *net)
4004 {
4005         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
4006
4007         if (!bp)
4008                 return -ENOMEM;
4009         inet_peer_base_init(bp);
4010         net->ipv6.peers = bp;
4011         return 0;
4012 }
4013
4014 static void __net_exit ipv6_inetpeer_exit(struct net *net)
4015 {
4016         struct inet_peer_base *bp = net->ipv6.peers;
4017
4018         net->ipv6.peers = NULL;
4019         inetpeer_invalidate_tree(bp);
4020         kfree(bp);
4021 }
4022
4023 static struct pernet_operations ipv6_inetpeer_ops = {
4024         .init   =       ipv6_inetpeer_init,
4025         .exit   =       ipv6_inetpeer_exit,
4026 };
4027
4028 static struct pernet_operations ip6_route_net_late_ops = {
4029         .init = ip6_route_net_init_late,
4030         .exit = ip6_route_net_exit_late,
4031 };
4032
4033 static struct notifier_block ip6_route_dev_notifier = {
4034         .notifier_call = ip6_route_dev_notify,
4035         .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
4036 };
4037
4038 void __init ip6_route_init_special_entries(void)
4039 {
4040         /* Registering of the loopback is done before this portion of code,
4041          * the loopback reference in rt6_info will not be taken, do it
4042          * manually for init_net */
4043         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
4044         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4045   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4046         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
4047         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4048         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
4049         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
4050   #endif
4051 }
4052
4053 int __init ip6_route_init(void)
4054 {
4055         int ret;
4056         int cpu;
4057
4058         ret = -ENOMEM;
4059         ip6_dst_ops_template.kmem_cachep =
4060                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
4061                                   SLAB_HWCACHE_ALIGN, NULL);
4062         if (!ip6_dst_ops_template.kmem_cachep)
4063                 goto out;
4064
4065         ret = dst_entries_init(&ip6_dst_blackhole_ops);
4066         if (ret)
4067                 goto out_kmem_cache;
4068
4069         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
4070         if (ret)
4071                 goto out_dst_entries;
4072
4073         ret = register_pernet_subsys(&ip6_route_net_ops);
4074         if (ret)
4075                 goto out_register_inetpeer;
4076
4077         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
4078
4079         ret = fib6_init();
4080         if (ret)
4081                 goto out_register_subsys;
4082
4083         ret = xfrm6_init();
4084         if (ret)
4085                 goto out_fib6_init;
4086
4087         ret = fib6_rules_init();
4088         if (ret)
4089                 goto xfrm6_init;
4090
4091         ret = register_pernet_subsys(&ip6_route_net_late_ops);
4092         if (ret)
4093                 goto fib6_rules_init;
4094
4095         ret = -ENOBUFS;
4096         if (__rtnl_register(PF_INET6, RTM_NEWROUTE, inet6_rtm_newroute, NULL, NULL) ||
4097             __rtnl_register(PF_INET6, RTM_DELROUTE, inet6_rtm_delroute, NULL, NULL) ||
4098             __rtnl_register(PF_INET6, RTM_GETROUTE, inet6_rtm_getroute, NULL, NULL))
4099                 goto out_register_late_subsys;
4100
4101         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
4102         if (ret)
4103                 goto out_register_late_subsys;
4104
4105         for_each_possible_cpu(cpu) {
4106                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
4107
4108                 INIT_LIST_HEAD(&ul->head);
4109                 spin_lock_init(&ul->lock);
4110         }
4111
4112 out:
4113         return ret;
4114
4115 out_register_late_subsys:
4116         unregister_pernet_subsys(&ip6_route_net_late_ops);
4117 fib6_rules_init:
4118         fib6_rules_cleanup();
4119 xfrm6_init:
4120         xfrm6_fini();
4121 out_fib6_init:
4122         fib6_gc_cleanup();
4123 out_register_subsys:
4124         unregister_pernet_subsys(&ip6_route_net_ops);
4125 out_register_inetpeer:
4126         unregister_pernet_subsys(&ipv6_inetpeer_ops);
4127 out_dst_entries:
4128         dst_entries_destroy(&ip6_dst_blackhole_ops);
4129 out_kmem_cache:
4130         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4131         goto out;
4132 }
4133
4134 void ip6_route_cleanup(void)
4135 {
4136         unregister_netdevice_notifier(&ip6_route_dev_notifier);
4137         unregister_pernet_subsys(&ip6_route_net_late_ops);
4138         fib6_rules_cleanup();
4139         xfrm6_fini();
4140         fib6_gc_cleanup();
4141         unregister_pernet_subsys(&ipv6_inetpeer_ops);
4142         unregister_pernet_subsys(&ip6_route_net_ops);
4143         dst_entries_destroy(&ip6_dst_blackhole_ops);
4144         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
4145 }