~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/net/ipv6/route.c

Version: ~ [ linux-5.8-rc5 ] ~ [ linux-5.7.8 ] ~ [ linux-5.6.19 ] ~ [ linux-5.5.19 ] ~ [ linux-5.4.51 ] ~ [ linux-5.3.18 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.132 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.188 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.230 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.230 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.85 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  1 // SPDX-License-Identifier: GPL-2.0-or-later
  2 /*
  3  *      Linux INET6 implementation
  4  *      FIB front-end.
  5  *
  6  *      Authors:
  7  *      Pedro Roque             <roque@di.fc.ul.pt>
  8  */
  9 
 10 /*      Changes:
 11  *
 12  *      YOSHIFUJI Hideaki @USAGI
 13  *              reworked default router selection.
 14  *              - respect outgoing interface
 15  *              - select from (probably) reachable routers (i.e.
 16  *              routers in REACHABLE, STALE, DELAY or PROBE states).
 17  *              - always select the same router if it is (probably)
 18  *              reachable.  otherwise, round-robin the list.
 19  *      Ville Nuorvala
 20  *              Fixed routing subtrees.
 21  */
 22 
 23 #define pr_fmt(fmt) "IPv6: " fmt
 24 
 25 #include <linux/capability.h>
 26 #include <linux/errno.h>
 27 #include <linux/export.h>
 28 #include <linux/types.h>
 29 #include <linux/times.h>
 30 #include <linux/socket.h>
 31 #include <linux/sockios.h>
 32 #include <linux/net.h>
 33 #include <linux/route.h>
 34 #include <linux/netdevice.h>
 35 #include <linux/in6.h>
 36 #include <linux/mroute6.h>
 37 #include <linux/init.h>
 38 #include <linux/if_arp.h>
 39 #include <linux/proc_fs.h>
 40 #include <linux/seq_file.h>
 41 #include <linux/nsproxy.h>
 42 #include <linux/slab.h>
 43 #include <linux/jhash.h>
 44 #include <net/net_namespace.h>
 45 #include <net/snmp.h>
 46 #include <net/ipv6.h>
 47 #include <net/ip6_fib.h>
 48 #include <net/ip6_route.h>
 49 #include <net/ndisc.h>
 50 #include <net/addrconf.h>
 51 #include <net/tcp.h>
 52 #include <linux/rtnetlink.h>
 53 #include <net/dst.h>
 54 #include <net/dst_metadata.h>
 55 #include <net/xfrm.h>
 56 #include <net/netevent.h>
 57 #include <net/netlink.h>
 58 #include <net/rtnh.h>
 59 #include <net/lwtunnel.h>
 60 #include <net/ip_tunnels.h>
 61 #include <net/l3mdev.h>
 62 #include <net/ip.h>
 63 #include <linux/uaccess.h>
 64 
 65 #ifdef CONFIG_SYSCTL
 66 #include <linux/sysctl.h>
 67 #endif
 68 
 69 static int ip6_rt_type_to_error(u8 fib6_type);
 70 
 71 #define CREATE_TRACE_POINTS
 72 #include <trace/events/fib6.h>
 73 EXPORT_TRACEPOINT_SYMBOL_GPL(fib6_table_lookup);
 74 #undef CREATE_TRACE_POINTS
 75 
 76 enum rt6_nud_state {
 77         RT6_NUD_FAIL_HARD = -3,
 78         RT6_NUD_FAIL_PROBE = -2,
 79         RT6_NUD_FAIL_DO_RR = -1,
 80         RT6_NUD_SUCCEED = 1
 81 };
 82 
 83 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie);
 84 static unsigned int      ip6_default_advmss(const struct dst_entry *dst);
 85 static unsigned int      ip6_mtu(const struct dst_entry *dst);
 86 static struct dst_entry *ip6_negative_advice(struct dst_entry *);
 87 static void             ip6_dst_destroy(struct dst_entry *);
 88 static void             ip6_dst_ifdown(struct dst_entry *,
 89                                        struct net_device *dev, int how);
 90 static int               ip6_dst_gc(struct dst_ops *ops);
 91 
 92 static int              ip6_pkt_discard(struct sk_buff *skb);
 93 static int              ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb);
 94 static int              ip6_pkt_prohibit(struct sk_buff *skb);
 95 static int              ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb);
 96 static void             ip6_link_failure(struct sk_buff *skb);
 97 static void             ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
 98                                            struct sk_buff *skb, u32 mtu);
 99 static void             rt6_do_redirect(struct dst_entry *dst, struct sock *sk,
100                                         struct sk_buff *skb);
101 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
102                            int strict);
103 static size_t rt6_nlmsg_size(struct fib6_info *f6i);
104 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
105                          struct fib6_info *rt, struct dst_entry *dst,
106                          struct in6_addr *dest, struct in6_addr *src,
107                          int iif, int type, u32 portid, u32 seq,
108                          unsigned int flags);
109 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
110                                            const struct in6_addr *daddr,
111                                            const struct in6_addr *saddr);
112 
113 #ifdef CONFIG_IPV6_ROUTE_INFO
114 static struct fib6_info *rt6_add_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                                            unsigned int pref);
119 static struct fib6_info *rt6_get_route_info(struct net *net,
120                                            const struct in6_addr *prefix, int prefixlen,
121                                            const struct in6_addr *gwaddr,
122                                            struct net_device *dev);
123 #endif
124 
125 struct uncached_list {
126         spinlock_t              lock;
127         struct list_head        head;
128 };
129 
130 static DEFINE_PER_CPU_ALIGNED(struct uncached_list, rt6_uncached_list);
131 
132 void rt6_uncached_list_add(struct rt6_info *rt)
133 {
134         struct uncached_list *ul = raw_cpu_ptr(&rt6_uncached_list);
135 
136         rt->rt6i_uncached_list = ul;
137 
138         spin_lock_bh(&ul->lock);
139         list_add_tail(&rt->rt6i_uncached, &ul->head);
140         spin_unlock_bh(&ul->lock);
141 }
142 
143 void rt6_uncached_list_del(struct rt6_info *rt)
144 {
145         if (!list_empty(&rt->rt6i_uncached)) {
146                 struct uncached_list *ul = rt->rt6i_uncached_list;
147                 struct net *net = dev_net(rt->dst.dev);
148 
149                 spin_lock_bh(&ul->lock);
150                 list_del(&rt->rt6i_uncached);
151                 atomic_dec(&net->ipv6.rt6_stats->fib_rt_uncache);
152                 spin_unlock_bh(&ul->lock);
153         }
154 }
155 
156 static void rt6_uncached_list_flush_dev(struct net *net, struct net_device *dev)
157 {
158         struct net_device *loopback_dev = net->loopback_dev;
159         int cpu;
160 
161         if (dev == loopback_dev)
162                 return;
163 
164         for_each_possible_cpu(cpu) {
165                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
166                 struct rt6_info *rt;
167 
168                 spin_lock_bh(&ul->lock);
169                 list_for_each_entry(rt, &ul->head, rt6i_uncached) {
170                         struct inet6_dev *rt_idev = rt->rt6i_idev;
171                         struct net_device *rt_dev = rt->dst.dev;
172 
173                         if (rt_idev->dev == dev) {
174                                 rt->rt6i_idev = in6_dev_get(loopback_dev);
175                                 in6_dev_put(rt_idev);
176                         }
177 
178                         if (rt_dev == dev) {
179                                 rt->dst.dev = blackhole_netdev;
180                                 dev_hold(rt->dst.dev);
181                                 dev_put(rt_dev);
182                         }
183                 }
184                 spin_unlock_bh(&ul->lock);
185         }
186 }
187 
188 static inline const void *choose_neigh_daddr(const struct in6_addr *p,
189                                              struct sk_buff *skb,
190                                              const void *daddr)
191 {
192         if (!ipv6_addr_any(p))
193                 return (const void *) p;
194         else if (skb)
195                 return &ipv6_hdr(skb)->daddr;
196         return daddr;
197 }
198 
199 struct neighbour *ip6_neigh_lookup(const struct in6_addr *gw,
200                                    struct net_device *dev,
201                                    struct sk_buff *skb,
202                                    const void *daddr)
203 {
204         struct neighbour *n;
205 
206         daddr = choose_neigh_daddr(gw, skb, daddr);
207         n = __ipv6_neigh_lookup(dev, daddr);
208         if (n)
209                 return n;
210 
211         n = neigh_create(&nd_tbl, daddr, dev);
212         return IS_ERR(n) ? NULL : n;
213 }
214 
215 static struct neighbour *ip6_dst_neigh_lookup(const struct dst_entry *dst,
216                                               struct sk_buff *skb,
217                                               const void *daddr)
218 {
219         const struct rt6_info *rt = container_of(dst, struct rt6_info, dst);
220 
221         return ip6_neigh_lookup(rt6_nexthop(rt, &in6addr_any),
222                                 dst->dev, skb, daddr);
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->rt6i_gateway, 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            =       dst_cow_metrics_generic,
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_dst_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_dst_neigh_lookup,
286 };
287 
288 static const u32 ip6_template_metrics[RTAX_MAX] = {
289         [RTAX_HOPLIMIT - 1] = 0,
290 };
291 
292 static const struct fib6_info fib6_null_entry_template = {
293         .fib6_flags     = (RTF_REJECT | RTF_NONEXTHOP),
294         .fib6_protocol  = RTPROT_KERNEL,
295         .fib6_metric    = ~(u32)0,
296         .fib6_ref       = REFCOUNT_INIT(1),
297         .fib6_type      = RTN_UNREACHABLE,
298         .fib6_metrics   = (struct dst_metrics *)&dst_default_metrics,
299 };
300 
301 static const struct rt6_info ip6_null_entry_template = {
302         .dst = {
303                 .__refcnt       = ATOMIC_INIT(1),
304                 .__use          = 1,
305                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
306                 .error          = -ENETUNREACH,
307                 .input          = ip6_pkt_discard,
308                 .output         = ip6_pkt_discard_out,
309         },
310         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
311 };
312 
313 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
314 
315 static const struct rt6_info ip6_prohibit_entry_template = {
316         .dst = {
317                 .__refcnt       = ATOMIC_INIT(1),
318                 .__use          = 1,
319                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
320                 .error          = -EACCES,
321                 .input          = ip6_pkt_prohibit,
322                 .output         = ip6_pkt_prohibit_out,
323         },
324         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
325 };
326 
327 static const struct rt6_info ip6_blk_hole_entry_template = {
328         .dst = {
329                 .__refcnt       = ATOMIC_INIT(1),
330                 .__use          = 1,
331                 .obsolete       = DST_OBSOLETE_FORCE_CHK,
332                 .error          = -EINVAL,
333                 .input          = dst_discard,
334                 .output         = dst_discard_out,
335         },
336         .rt6i_flags     = (RTF_REJECT | RTF_NONEXTHOP),
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_uncached);
347 }
348 
349 /* allocate dst with ip6_dst_ops */
350 struct rt6_info *ip6_dst_alloc(struct net *net, struct net_device *dev,
351                                int flags)
352 {
353         struct rt6_info *rt = dst_alloc(&net->ipv6.ip6_dst_ops, dev,
354                                         1, DST_OBSOLETE_FORCE_CHK, flags);
355 
356         if (rt) {
357                 rt6_info_init(rt);
358                 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
359         }
360 
361         return rt;
362 }
363 EXPORT_SYMBOL(ip6_dst_alloc);
364 
365 static void ip6_dst_destroy(struct dst_entry *dst)
366 {
367         struct rt6_info *rt = (struct rt6_info *)dst;
368         struct fib6_info *from;
369         struct inet6_dev *idev;
370 
371         ip_dst_metrics_put(dst);
372         rt6_uncached_list_del(rt);
373 
374         idev = rt->rt6i_idev;
375         if (idev) {
376                 rt->rt6i_idev = NULL;
377                 in6_dev_put(idev);
378         }
379 
380         from = xchg((__force struct fib6_info **)&rt->from, NULL);
381         fib6_info_release(from);
382 }
383 
384 static void ip6_dst_ifdown(struct dst_entry *dst, struct net_device *dev,
385                            int how)
386 {
387         struct rt6_info *rt = (struct rt6_info *)dst;
388         struct inet6_dev *idev = rt->rt6i_idev;
389         struct net_device *loopback_dev =
390                 dev_net(dev)->loopback_dev;
391 
392         if (idev && idev->dev != loopback_dev) {
393                 struct inet6_dev *loopback_idev = in6_dev_get(loopback_dev);
394                 if (loopback_idev) {
395                         rt->rt6i_idev = loopback_idev;
396                         in6_dev_put(idev);
397                 }
398         }
399 }
400 
401 static bool __rt6_check_expired(const struct rt6_info *rt)
402 {
403         if (rt->rt6i_flags & RTF_EXPIRES)
404                 return time_after(jiffies, rt->dst.expires);
405         else
406                 return false;
407 }
408 
409 static bool rt6_check_expired(const struct rt6_info *rt)
410 {
411         struct fib6_info *from;
412 
413         from = rcu_dereference(rt->from);
414 
415         if (rt->rt6i_flags & RTF_EXPIRES) {
416                 if (time_after(jiffies, rt->dst.expires))
417                         return true;
418         } else if (from) {
419                 return rt->dst.obsolete != DST_OBSOLETE_FORCE_CHK ||
420                         fib6_check_expired(from);
421         }
422         return false;
423 }
424 
425 void fib6_select_path(const struct net *net, struct fib6_result *res,
426                       struct flowi6 *fl6, int oif, bool have_oif_match,
427                       const struct sk_buff *skb, int strict)
428 {
429         struct fib6_info *sibling, *next_sibling;
430         struct fib6_info *match = res->f6i;
431 
432         if ((!match->fib6_nsiblings && !match->nh) || have_oif_match)
433                 goto out;
434 
435         /* We might have already computed the hash for ICMPv6 errors. In such
436          * case it will always be non-zero. Otherwise now is the time to do it.
437          */
438         if (!fl6->mp_hash &&
439             (!match->nh || nexthop_is_multipath(match->nh)))
440                 fl6->mp_hash = rt6_multipath_hash(net, fl6, skb, NULL);
441 
442         if (unlikely(match->nh)) {
443                 nexthop_path_fib6_result(res, fl6->mp_hash);
444                 return;
445         }
446 
447         if (fl6->mp_hash <= atomic_read(&match->fib6_nh->fib_nh_upper_bound))
448                 goto out;
449 
450         list_for_each_entry_safe(sibling, next_sibling, &match->fib6_siblings,
451                                  fib6_siblings) {
452                 const struct fib6_nh *nh = sibling->fib6_nh;
453                 int nh_upper_bound;
454 
455                 nh_upper_bound = atomic_read(&nh->fib_nh_upper_bound);
456                 if (fl6->mp_hash > nh_upper_bound)
457                         continue;
458                 if (rt6_score_route(nh, sibling->fib6_flags, oif, strict) < 0)
459                         break;
460                 match = sibling;
461                 break;
462         }
463 
464 out:
465         res->f6i = match;
466         res->nh = match->fib6_nh;
467 }
468 
469 /*
470  *      Route lookup. rcu_read_lock() should be held.
471  */
472 
473 static bool __rt6_device_match(struct net *net, const struct fib6_nh *nh,
474                                const struct in6_addr *saddr, int oif, int flags)
475 {
476         const struct net_device *dev;
477 
478         if (nh->fib_nh_flags & RTNH_F_DEAD)
479                 return false;
480 
481         dev = nh->fib_nh_dev;
482         if (oif) {
483                 if (dev->ifindex == oif)
484                         return true;
485         } else {
486                 if (ipv6_chk_addr(net, saddr, dev,
487                                   flags & RT6_LOOKUP_F_IFACE))
488                         return true;
489         }
490 
491         return false;
492 }
493 
494 struct fib6_nh_dm_arg {
495         struct net              *net;
496         const struct in6_addr   *saddr;
497         int                     oif;
498         int                     flags;
499         struct fib6_nh          *nh;
500 };
501 
502 static int __rt6_nh_dev_match(struct fib6_nh *nh, void *_arg)
503 {
504         struct fib6_nh_dm_arg *arg = _arg;
505 
506         arg->nh = nh;
507         return __rt6_device_match(arg->net, nh, arg->saddr, arg->oif,
508                                   arg->flags);
509 }
510 
511 /* returns fib6_nh from nexthop or NULL */
512 static struct fib6_nh *rt6_nh_dev_match(struct net *net, struct nexthop *nh,
513                                         struct fib6_result *res,
514                                         const struct in6_addr *saddr,
515                                         int oif, int flags)
516 {
517         struct fib6_nh_dm_arg arg = {
518                 .net   = net,
519                 .saddr = saddr,
520                 .oif   = oif,
521                 .flags = flags,
522         };
523 
524         if (nexthop_is_blackhole(nh))
525                 return NULL;
526 
527         if (nexthop_for_each_fib6_nh(nh, __rt6_nh_dev_match, &arg))
528                 return arg.nh;
529 
530         return NULL;
531 }
532 
533 static void rt6_device_match(struct net *net, struct fib6_result *res,
534                              const struct in6_addr *saddr, int oif, int flags)
535 {
536         struct fib6_info *f6i = res->f6i;
537         struct fib6_info *spf6i;
538         struct fib6_nh *nh;
539 
540         if (!oif && ipv6_addr_any(saddr)) {
541                 if (unlikely(f6i->nh)) {
542                         nh = nexthop_fib6_nh(f6i->nh);
543                         if (nexthop_is_blackhole(f6i->nh))
544                                 goto out_blackhole;
545                 } else {
546                         nh = f6i->fib6_nh;
547                 }
548                 if (!(nh->fib_nh_flags & RTNH_F_DEAD))
549                         goto out;
550         }
551 
552         for (spf6i = f6i; spf6i; spf6i = rcu_dereference(spf6i->fib6_next)) {
553                 bool matched = false;
554 
555                 if (unlikely(spf6i->nh)) {
556                         nh = rt6_nh_dev_match(net, spf6i->nh, res, saddr,
557                                               oif, flags);
558                         if (nh)
559                                 matched = true;
560                 } else {
561                         nh = spf6i->fib6_nh;
562                         if (__rt6_device_match(net, nh, saddr, oif, flags))
563                                 matched = true;
564                 }
565                 if (matched) {
566                         res->f6i = spf6i;
567                         goto out;
568                 }
569         }
570 
571         if (oif && flags & RT6_LOOKUP_F_IFACE) {
572                 res->f6i = net->ipv6.fib6_null_entry;
573                 nh = res->f6i->fib6_nh;
574                 goto out;
575         }
576 
577         if (unlikely(f6i->nh)) {
578                 nh = nexthop_fib6_nh(f6i->nh);
579                 if (nexthop_is_blackhole(f6i->nh))
580                         goto out_blackhole;
581         } else {
582                 nh = f6i->fib6_nh;
583         }
584 
585         if (nh->fib_nh_flags & RTNH_F_DEAD) {
586                 res->f6i = net->ipv6.fib6_null_entry;
587                 nh = res->f6i->fib6_nh;
588         }
589 out:
590         res->nh = nh;
591         res->fib6_type = res->f6i->fib6_type;
592         res->fib6_flags = res->f6i->fib6_flags;
593         return;
594 
595 out_blackhole:
596         res->fib6_flags |= RTF_REJECT;
597         res->fib6_type = RTN_BLACKHOLE;
598         res->nh = nh;
599 }
600 
601 #ifdef CONFIG_IPV6_ROUTER_PREF
602 struct __rt6_probe_work {
603         struct work_struct work;
604         struct in6_addr target;
605         struct net_device *dev;
606 };
607 
608 static void rt6_probe_deferred(struct work_struct *w)
609 {
610         struct in6_addr mcaddr;
611         struct __rt6_probe_work *work =
612                 container_of(w, struct __rt6_probe_work, work);
613 
614         addrconf_addr_solict_mult(&work->target, &mcaddr);
615         ndisc_send_ns(work->dev, &work->target, &mcaddr, NULL, 0);
616         dev_put(work->dev);
617         kfree(work);
618 }
619 
620 static void rt6_probe(struct fib6_nh *fib6_nh)
621 {
622         struct __rt6_probe_work *work = NULL;
623         const struct in6_addr *nh_gw;
624         unsigned long last_probe;
625         struct neighbour *neigh;
626         struct net_device *dev;
627         struct inet6_dev *idev;
628 
629         /*
630          * Okay, this does not seem to be appropriate
631          * for now, however, we need to check if it
632          * is really so; aka Router Reachability Probing.
633          *
634          * Router Reachability Probe MUST be rate-limited
635          * to no more than one per minute.
636          */
637         if (!fib6_nh->fib_nh_gw_family)
638                 return;
639 
640         nh_gw = &fib6_nh->fib_nh_gw6;
641         dev = fib6_nh->fib_nh_dev;
642         rcu_read_lock_bh();
643         last_probe = READ_ONCE(fib6_nh->last_probe);
644         idev = __in6_dev_get(dev);
645         neigh = __ipv6_neigh_lookup_noref(dev, nh_gw);
646         if (neigh) {
647                 if (neigh->nud_state & NUD_VALID)
648                         goto out;
649 
650                 write_lock(&neigh->lock);
651                 if (!(neigh->nud_state & NUD_VALID) &&
652                     time_after(jiffies,
653                                neigh->updated + idev->cnf.rtr_probe_interval)) {
654                         work = kmalloc(sizeof(*work), GFP_ATOMIC);
655                         if (work)
656                                 __neigh_set_probe_once(neigh);
657                 }
658                 write_unlock(&neigh->lock);
659         } else if (time_after(jiffies, last_probe +
660                                        idev->cnf.rtr_probe_interval)) {
661                 work = kmalloc(sizeof(*work), GFP_ATOMIC);
662         }
663 
664         if (!work || cmpxchg(&fib6_nh->last_probe,
665                              last_probe, jiffies) != last_probe) {
666                 kfree(work);
667         } else {
668                 INIT_WORK(&work->work, rt6_probe_deferred);
669                 work->target = *nh_gw;
670                 dev_hold(dev);
671                 work->dev = dev;
672                 schedule_work(&work->work);
673         }
674 
675 out:
676         rcu_read_unlock_bh();
677 }
678 #else
679 static inline void rt6_probe(struct fib6_nh *fib6_nh)
680 {
681 }
682 #endif
683 
684 /*
685  * Default Router Selection (RFC 2461 6.3.6)
686  */
687 static enum rt6_nud_state rt6_check_neigh(const struct fib6_nh *fib6_nh)
688 {
689         enum rt6_nud_state ret = RT6_NUD_FAIL_HARD;
690         struct neighbour *neigh;
691 
692         rcu_read_lock_bh();
693         neigh = __ipv6_neigh_lookup_noref(fib6_nh->fib_nh_dev,
694                                           &fib6_nh->fib_nh_gw6);
695         if (neigh) {
696                 read_lock(&neigh->lock);
697                 if (neigh->nud_state & NUD_VALID)
698                         ret = RT6_NUD_SUCCEED;
699 #ifdef CONFIG_IPV6_ROUTER_PREF
700                 else if (!(neigh->nud_state & NUD_FAILED))
701                         ret = RT6_NUD_SUCCEED;
702                 else
703                         ret = RT6_NUD_FAIL_PROBE;
704 #endif
705                 read_unlock(&neigh->lock);
706         } else {
707                 ret = IS_ENABLED(CONFIG_IPV6_ROUTER_PREF) ?
708                       RT6_NUD_SUCCEED : RT6_NUD_FAIL_DO_RR;
709         }
710         rcu_read_unlock_bh();
711 
712         return ret;
713 }
714 
715 static int rt6_score_route(const struct fib6_nh *nh, u32 fib6_flags, int oif,
716                            int strict)
717 {
718         int m = 0;
719 
720         if (!oif || nh->fib_nh_dev->ifindex == oif)
721                 m = 2;
722 
723         if (!m && (strict & RT6_LOOKUP_F_IFACE))
724                 return RT6_NUD_FAIL_HARD;
725 #ifdef CONFIG_IPV6_ROUTER_PREF
726         m |= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(fib6_flags)) << 2;
727 #endif
728         if ((strict & RT6_LOOKUP_F_REACHABLE) &&
729             !(fib6_flags & RTF_NONEXTHOP) && nh->fib_nh_gw_family) {
730                 int n = rt6_check_neigh(nh);
731                 if (n < 0)
732                         return n;
733         }
734         return m;
735 }
736 
737 static bool find_match(struct fib6_nh *nh, u32 fib6_flags,
738                        int oif, int strict, int *mpri, bool *do_rr)
739 {
740         bool match_do_rr = false;
741         bool rc = false;
742         int m;
743 
744         if (nh->fib_nh_flags & RTNH_F_DEAD)
745                 goto out;
746 
747         if (ip6_ignore_linkdown(nh->fib_nh_dev) &&
748             nh->fib_nh_flags & RTNH_F_LINKDOWN &&
749             !(strict & RT6_LOOKUP_F_IGNORE_LINKSTATE))
750                 goto out;
751 
752         m = rt6_score_route(nh, fib6_flags, oif, strict);
753         if (m == RT6_NUD_FAIL_DO_RR) {
754                 match_do_rr = true;
755                 m = 0; /* lowest valid score */
756         } else if (m == RT6_NUD_FAIL_HARD) {
757                 goto out;
758         }
759 
760         if (strict & RT6_LOOKUP_F_REACHABLE)
761                 rt6_probe(nh);
762 
763         /* note that m can be RT6_NUD_FAIL_PROBE at this point */
764         if (m > *mpri) {
765                 *do_rr = match_do_rr;
766                 *mpri = m;
767                 rc = true;
768         }
769 out:
770         return rc;
771 }
772 
773 struct fib6_nh_frl_arg {
774         u32             flags;
775         int             oif;
776         int             strict;
777         int             *mpri;
778         bool            *do_rr;
779         struct fib6_nh  *nh;
780 };
781 
782 static int rt6_nh_find_match(struct fib6_nh *nh, void *_arg)
783 {
784         struct fib6_nh_frl_arg *arg = _arg;
785 
786         arg->nh = nh;
787         return find_match(nh, arg->flags, arg->oif, arg->strict,
788                           arg->mpri, arg->do_rr);
789 }
790 
791 static void __find_rr_leaf(struct fib6_info *f6i_start,
792                            struct fib6_info *nomatch, u32 metric,
793                            struct fib6_result *res, struct fib6_info **cont,
794                            int oif, int strict, bool *do_rr, int *mpri)
795 {
796         struct fib6_info *f6i;
797 
798         for (f6i = f6i_start;
799              f6i && f6i != nomatch;
800              f6i = rcu_dereference(f6i->fib6_next)) {
801                 bool matched = false;
802                 struct fib6_nh *nh;
803 
804                 if (cont && f6i->fib6_metric != metric) {
805                         *cont = f6i;
806                         return;
807                 }
808 
809                 if (fib6_check_expired(f6i))
810                         continue;
811 
812                 if (unlikely(f6i->nh)) {
813                         struct fib6_nh_frl_arg arg = {
814                                 .flags  = f6i->fib6_flags,
815                                 .oif    = oif,
816                                 .strict = strict,
817                                 .mpri   = mpri,
818                                 .do_rr  = do_rr
819                         };
820 
821                         if (nexthop_is_blackhole(f6i->nh)) {
822                                 res->fib6_flags = RTF_REJECT;
823                                 res->fib6_type = RTN_BLACKHOLE;
824                                 res->f6i = f6i;
825                                 res->nh = nexthop_fib6_nh(f6i->nh);
826                                 return;
827                         }
828                         if (nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_find_match,
829                                                      &arg)) {
830                                 matched = true;
831                                 nh = arg.nh;
832                         }
833                 } else {
834                         nh = f6i->fib6_nh;
835                         if (find_match(nh, f6i->fib6_flags, oif, strict,
836                                        mpri, do_rr))
837                                 matched = true;
838                 }
839                 if (matched) {
840                         res->f6i = f6i;
841                         res->nh = nh;
842                         res->fib6_flags = f6i->fib6_flags;
843                         res->fib6_type = f6i->fib6_type;
844                 }
845         }
846 }
847 
848 static void find_rr_leaf(struct fib6_node *fn, struct fib6_info *leaf,
849                          struct fib6_info *rr_head, int oif, int strict,
850                          bool *do_rr, struct fib6_result *res)
851 {
852         u32 metric = rr_head->fib6_metric;
853         struct fib6_info *cont = NULL;
854         int mpri = -1;
855 
856         __find_rr_leaf(rr_head, NULL, metric, res, &cont,
857                        oif, strict, do_rr, &mpri);
858 
859         __find_rr_leaf(leaf, rr_head, metric, res, &cont,
860                        oif, strict, do_rr, &mpri);
861 
862         if (res->f6i || !cont)
863                 return;
864 
865         __find_rr_leaf(cont, NULL, metric, res, NULL,
866                        oif, strict, do_rr, &mpri);
867 }
868 
869 static void rt6_select(struct net *net, struct fib6_node *fn, int oif,
870                        struct fib6_result *res, int strict)
871 {
872         struct fib6_info *leaf = rcu_dereference(fn->leaf);
873         struct fib6_info *rt0;
874         bool do_rr = false;
875         int key_plen;
876 
877         /* make sure this function or its helpers sets f6i */
878         res->f6i = NULL;
879 
880         if (!leaf || leaf == net->ipv6.fib6_null_entry)
881                 goto out;
882 
883         rt0 = rcu_dereference(fn->rr_ptr);
884         if (!rt0)
885                 rt0 = leaf;
886 
887         /* Double check to make sure fn is not an intermediate node
888          * and fn->leaf does not points to its child's leaf
889          * (This might happen if all routes under fn are deleted from
890          * the tree and fib6_repair_tree() is called on the node.)
891          */
892         key_plen = rt0->fib6_dst.plen;
893 #ifdef CONFIG_IPV6_SUBTREES
894         if (rt0->fib6_src.plen)
895                 key_plen = rt0->fib6_src.plen;
896 #endif
897         if (fn->fn_bit != key_plen)
898                 goto out;
899 
900         find_rr_leaf(fn, leaf, rt0, oif, strict, &do_rr, res);
901         if (do_rr) {
902                 struct fib6_info *next = rcu_dereference(rt0->fib6_next);
903 
904                 /* no entries matched; do round-robin */
905                 if (!next || next->fib6_metric != rt0->fib6_metric)
906                         next = leaf;
907 
908                 if (next != rt0) {
909                         spin_lock_bh(&leaf->fib6_table->tb6_lock);
910                         /* make sure next is not being deleted from the tree */
911                         if (next->fib6_node)
912                                 rcu_assign_pointer(fn->rr_ptr, next);
913                         spin_unlock_bh(&leaf->fib6_table->tb6_lock);
914                 }
915         }
916 
917 out:
918         if (!res->f6i) {
919                 res->f6i = net->ipv6.fib6_null_entry;
920                 res->nh = res->f6i->fib6_nh;
921                 res->fib6_flags = res->f6i->fib6_flags;
922                 res->fib6_type = res->f6i->fib6_type;
923         }
924 }
925 
926 static bool rt6_is_gw_or_nonexthop(const struct fib6_result *res)
927 {
928         return (res->f6i->fib6_flags & RTF_NONEXTHOP) ||
929                res->nh->fib_nh_gw_family;
930 }
931 
932 #ifdef CONFIG_IPV6_ROUTE_INFO
933 int rt6_route_rcv(struct net_device *dev, u8 *opt, int len,
934                   const struct in6_addr *gwaddr)
935 {
936         struct net *net = dev_net(dev);
937         struct route_info *rinfo = (struct route_info *) opt;
938         struct in6_addr prefix_buf, *prefix;
939         unsigned int pref;
940         unsigned long lifetime;
941         struct fib6_info *rt;
942 
943         if (len < sizeof(struct route_info)) {
944                 return -EINVAL;
945         }
946 
947         /* Sanity check for prefix_len and length */
948         if (rinfo->length > 3) {
949                 return -EINVAL;
950         } else if (rinfo->prefix_len > 128) {
951                 return -EINVAL;
952         } else if (rinfo->prefix_len > 64) {
953                 if (rinfo->length < 2) {
954                         return -EINVAL;
955                 }
956         } else if (rinfo->prefix_len > 0) {
957                 if (rinfo->length < 1) {
958                         return -EINVAL;
959                 }
960         }
961 
962         pref = rinfo->route_pref;
963         if (pref == ICMPV6_ROUTER_PREF_INVALID)
964                 return -EINVAL;
965 
966         lifetime = addrconf_timeout_fixup(ntohl(rinfo->lifetime), HZ);
967 
968         if (rinfo->length == 3)
969                 prefix = (struct in6_addr *)rinfo->prefix;
970         else {
971                 /* this function is safe */
972                 ipv6_addr_prefix(&prefix_buf,
973                                  (struct in6_addr *)rinfo->prefix,
974                                  rinfo->prefix_len);
975                 prefix = &prefix_buf;
976         }
977 
978         if (rinfo->prefix_len == 0)
979                 rt = rt6_get_dflt_router(net, gwaddr, dev);
980         else
981                 rt = rt6_get_route_info(net, prefix, rinfo->prefix_len,
982                                         gwaddr, dev);
983 
984         if (rt && !lifetime) {
985                 ip6_del_rt(net, rt);
986                 rt = NULL;
987         }
988 
989         if (!rt && lifetime)
990                 rt = rt6_add_route_info(net, prefix, rinfo->prefix_len, gwaddr,
991                                         dev, pref);
992         else if (rt)
993                 rt->fib6_flags = RTF_ROUTEINFO |
994                                  (rt->fib6_flags & ~RTF_PREF_MASK) | RTF_PREF(pref);
995 
996         if (rt) {
997                 if (!addrconf_finite_timeout(lifetime))
998                         fib6_clean_expires(rt);
999                 else
1000                         fib6_set_expires(rt, jiffies + HZ * lifetime);
1001 
1002                 fib6_info_release(rt);
1003         }
1004         return 0;
1005 }
1006 #endif
1007 
1008 /*
1009  *      Misc support functions
1010  */
1011 
1012 /* called with rcu_lock held */
1013 static struct net_device *ip6_rt_get_dev_rcu(const struct fib6_result *res)
1014 {
1015         struct net_device *dev = res->nh->fib_nh_dev;
1016 
1017         if (res->fib6_flags & (RTF_LOCAL | RTF_ANYCAST)) {
1018                 /* for copies of local routes, dst->dev needs to be the
1019                  * device if it is a master device, the master device if
1020                  * device is enslaved, and the loopback as the default
1021                  */
1022                 if (netif_is_l3_slave(dev) &&
1023                     !rt6_need_strict(&res->f6i->fib6_dst.addr))
1024                         dev = l3mdev_master_dev_rcu(dev);
1025                 else if (!netif_is_l3_master(dev))
1026                         dev = dev_net(dev)->loopback_dev;
1027                 /* last case is netif_is_l3_master(dev) is true in which
1028                  * case we want dev returned to be dev
1029                  */
1030         }
1031 
1032         return dev;
1033 }
1034 
1035 static const int fib6_prop[RTN_MAX + 1] = {
1036         [RTN_UNSPEC]    = 0,
1037         [RTN_UNICAST]   = 0,
1038         [RTN_LOCAL]     = 0,
1039         [RTN_BROADCAST] = 0,
1040         [RTN_ANYCAST]   = 0,
1041         [RTN_MULTICAST] = 0,
1042         [RTN_BLACKHOLE] = -EINVAL,
1043         [RTN_UNREACHABLE] = -EHOSTUNREACH,
1044         [RTN_PROHIBIT]  = -EACCES,
1045         [RTN_THROW]     = -EAGAIN,
1046         [RTN_NAT]       = -EINVAL,
1047         [RTN_XRESOLVE]  = -EINVAL,
1048 };
1049 
1050 static int ip6_rt_type_to_error(u8 fib6_type)
1051 {
1052         return fib6_prop[fib6_type];
1053 }
1054 
1055 static unsigned short fib6_info_dst_flags(struct fib6_info *rt)
1056 {
1057         unsigned short flags = 0;
1058 
1059         if (rt->dst_nocount)
1060                 flags |= DST_NOCOUNT;
1061         if (rt->dst_nopolicy)
1062                 flags |= DST_NOPOLICY;
1063         if (rt->dst_host)
1064                 flags |= DST_HOST;
1065 
1066         return flags;
1067 }
1068 
1069 static void ip6_rt_init_dst_reject(struct rt6_info *rt, u8 fib6_type)
1070 {
1071         rt->dst.error = ip6_rt_type_to_error(fib6_type);
1072 
1073         switch (fib6_type) {
1074         case RTN_BLACKHOLE:
1075                 rt->dst.output = dst_discard_out;
1076                 rt->dst.input = dst_discard;
1077                 break;
1078         case RTN_PROHIBIT:
1079                 rt->dst.output = ip6_pkt_prohibit_out;
1080                 rt->dst.input = ip6_pkt_prohibit;
1081                 break;
1082         case RTN_THROW:
1083         case RTN_UNREACHABLE:
1084         default:
1085                 rt->dst.output = ip6_pkt_discard_out;
1086                 rt->dst.input = ip6_pkt_discard;
1087                 break;
1088         }
1089 }
1090 
1091 static void ip6_rt_init_dst(struct rt6_info *rt, const struct fib6_result *res)
1092 {
1093         struct fib6_info *f6i = res->f6i;
1094 
1095         if (res->fib6_flags & RTF_REJECT) {
1096                 ip6_rt_init_dst_reject(rt, res->fib6_type);
1097                 return;
1098         }
1099 
1100         rt->dst.error = 0;
1101         rt->dst.output = ip6_output;
1102 
1103         if (res->fib6_type == RTN_LOCAL || res->fib6_type == RTN_ANYCAST) {
1104                 rt->dst.input = ip6_input;
1105         } else if (ipv6_addr_type(&f6i->fib6_dst.addr) & IPV6_ADDR_MULTICAST) {
1106                 rt->dst.input = ip6_mc_input;
1107         } else {
1108                 rt->dst.input = ip6_forward;
1109         }
1110 
1111         if (res->nh->fib_nh_lws) {
1112                 rt->dst.lwtstate = lwtstate_get(res->nh->fib_nh_lws);
1113                 lwtunnel_set_redirect(&rt->dst);
1114         }
1115 
1116         rt->dst.lastuse = jiffies;
1117 }
1118 
1119 /* Caller must already hold reference to @from */
1120 static void rt6_set_from(struct rt6_info *rt, struct fib6_info *from)
1121 {
1122         rt->rt6i_flags &= ~RTF_EXPIRES;
1123         rcu_assign_pointer(rt->from, from);
1124         ip_dst_init_metrics(&rt->dst, from->fib6_metrics);
1125 }
1126 
1127 /* Caller must already hold reference to f6i in result */
1128 static void ip6_rt_copy_init(struct rt6_info *rt, const struct fib6_result *res)
1129 {
1130         const struct fib6_nh *nh = res->nh;
1131         const struct net_device *dev = nh->fib_nh_dev;
1132         struct fib6_info *f6i = res->f6i;
1133 
1134         ip6_rt_init_dst(rt, res);
1135 
1136         rt->rt6i_dst = f6i->fib6_dst;
1137         rt->rt6i_idev = dev ? in6_dev_get(dev) : NULL;
1138         rt->rt6i_flags = res->fib6_flags;
1139         if (nh->fib_nh_gw_family) {
1140                 rt->rt6i_gateway = nh->fib_nh_gw6;
1141                 rt->rt6i_flags |= RTF_GATEWAY;
1142         }
1143         rt6_set_from(rt, f6i);
1144 #ifdef CONFIG_IPV6_SUBTREES
1145         rt->rt6i_src = f6i->fib6_src;
1146 #endif
1147 }
1148 
1149 static struct fib6_node* fib6_backtrack(struct fib6_node *fn,
1150                                         struct in6_addr *saddr)
1151 {
1152         struct fib6_node *pn, *sn;
1153         while (1) {
1154                 if (fn->fn_flags & RTN_TL_ROOT)
1155                         return NULL;
1156                 pn = rcu_dereference(fn->parent);
1157                 sn = FIB6_SUBTREE(pn);
1158                 if (sn && sn != fn)
1159                         fn = fib6_node_lookup(sn, NULL, saddr);
1160                 else
1161                         fn = pn;
1162                 if (fn->fn_flags & RTN_RTINFO)
1163                         return fn;
1164         }
1165 }
1166 
1167 static bool ip6_hold_safe(struct net *net, struct rt6_info **prt)
1168 {
1169         struct rt6_info *rt = *prt;
1170 
1171         if (dst_hold_safe(&rt->dst))
1172                 return true;
1173         if (net) {
1174                 rt = net->ipv6.ip6_null_entry;
1175                 dst_hold(&rt->dst);
1176         } else {
1177                 rt = NULL;
1178         }
1179         *prt = rt;
1180         return false;
1181 }
1182 
1183 /* called with rcu_lock held */
1184 static struct rt6_info *ip6_create_rt_rcu(const struct fib6_result *res)
1185 {
1186         struct net_device *dev = res->nh->fib_nh_dev;
1187         struct fib6_info *f6i = res->f6i;
1188         unsigned short flags;
1189         struct rt6_info *nrt;
1190 
1191         if (!fib6_info_hold_safe(f6i))
1192                 goto fallback;
1193 
1194         flags = fib6_info_dst_flags(f6i);
1195         nrt = ip6_dst_alloc(dev_net(dev), dev, flags);
1196         if (!nrt) {
1197                 fib6_info_release(f6i);
1198                 goto fallback;
1199         }
1200 
1201         ip6_rt_copy_init(nrt, res);
1202         return nrt;
1203 
1204 fallback:
1205         nrt = dev_net(dev)->ipv6.ip6_null_entry;
1206         dst_hold(&nrt->dst);
1207         return nrt;
1208 }
1209 
1210 static struct rt6_info *ip6_pol_route_lookup(struct net *net,
1211                                              struct fib6_table *table,
1212                                              struct flowi6 *fl6,
1213                                              const struct sk_buff *skb,
1214                                              int flags)
1215 {
1216         struct fib6_result res = {};
1217         struct fib6_node *fn;
1218         struct rt6_info *rt;
1219 
1220         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
1221                 flags &= ~RT6_LOOKUP_F_IFACE;
1222 
1223         rcu_read_lock();
1224         fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
1225 restart:
1226         res.f6i = rcu_dereference(fn->leaf);
1227         if (!res.f6i)
1228                 res.f6i = net->ipv6.fib6_null_entry;
1229         else
1230                 rt6_device_match(net, &res, &fl6->saddr, fl6->flowi6_oif,
1231                                  flags);
1232 
1233         if (res.f6i == net->ipv6.fib6_null_entry) {
1234                 fn = fib6_backtrack(fn, &fl6->saddr);
1235                 if (fn)
1236                         goto restart;
1237 
1238                 rt = net->ipv6.ip6_null_entry;
1239                 dst_hold(&rt->dst);
1240                 goto out;
1241         } else if (res.fib6_flags & RTF_REJECT) {
1242                 goto do_create;
1243         }
1244 
1245         fib6_select_path(net, &res, fl6, fl6->flowi6_oif,
1246                          fl6->flowi6_oif != 0, skb, flags);
1247 
1248         /* Search through exception table */
1249         rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
1250         if (rt) {
1251                 if (ip6_hold_safe(net, &rt))
1252                         dst_use_noref(&rt->dst, jiffies);
1253         } else {
1254 do_create:
1255                 rt = ip6_create_rt_rcu(&res);
1256         }
1257 
1258 out:
1259         trace_fib6_table_lookup(net, &res, table, fl6);
1260 
1261         rcu_read_unlock();
1262 
1263         return rt;
1264 }
1265 
1266 struct dst_entry *ip6_route_lookup(struct net *net, struct flowi6 *fl6,
1267                                    const struct sk_buff *skb, int flags)
1268 {
1269         return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_lookup);
1270 }
1271 EXPORT_SYMBOL_GPL(ip6_route_lookup);
1272 
1273 struct rt6_info *rt6_lookup(struct net *net, const struct in6_addr *daddr,
1274                             const struct in6_addr *saddr, int oif,
1275                             const struct sk_buff *skb, int strict)
1276 {
1277         struct flowi6 fl6 = {
1278                 .flowi6_oif = oif,
1279                 .daddr = *daddr,
1280         };
1281         struct dst_entry *dst;
1282         int flags = strict ? RT6_LOOKUP_F_IFACE : 0;
1283 
1284         if (saddr) {
1285                 memcpy(&fl6.saddr, saddr, sizeof(*saddr));
1286                 flags |= RT6_LOOKUP_F_HAS_SADDR;
1287         }
1288 
1289         dst = fib6_rule_lookup(net, &fl6, skb, flags, ip6_pol_route_lookup);
1290         if (dst->error == 0)
1291                 return (struct rt6_info *) dst;
1292 
1293         dst_release(dst);
1294 
1295         return NULL;
1296 }
1297 EXPORT_SYMBOL(rt6_lookup);
1298 
1299 /* ip6_ins_rt is called with FREE table->tb6_lock.
1300  * It takes new route entry, the addition fails by any reason the
1301  * route is released.
1302  * Caller must hold dst before calling it.
1303  */
1304 
1305 static int __ip6_ins_rt(struct fib6_info *rt, struct nl_info *info,
1306                         struct netlink_ext_ack *extack)
1307 {
1308         int err;
1309         struct fib6_table *table;
1310 
1311         table = rt->fib6_table;
1312         spin_lock_bh(&table->tb6_lock);
1313         err = fib6_add(&table->tb6_root, rt, info, extack);
1314         spin_unlock_bh(&table->tb6_lock);
1315 
1316         return err;
1317 }
1318 
1319 int ip6_ins_rt(struct net *net, struct fib6_info *rt)
1320 {
1321         struct nl_info info = { .nl_net = net, };
1322 
1323         return __ip6_ins_rt(rt, &info, NULL);
1324 }
1325 
1326 static struct rt6_info *ip6_rt_cache_alloc(const struct fib6_result *res,
1327                                            const struct in6_addr *daddr,
1328                                            const struct in6_addr *saddr)
1329 {
1330         struct fib6_info *f6i = res->f6i;
1331         struct net_device *dev;
1332         struct rt6_info *rt;
1333 
1334         /*
1335          *      Clone the route.
1336          */
1337 
1338         if (!fib6_info_hold_safe(f6i))
1339                 return NULL;
1340 
1341         dev = ip6_rt_get_dev_rcu(res);
1342         rt = ip6_dst_alloc(dev_net(dev), dev, 0);
1343         if (!rt) {
1344                 fib6_info_release(f6i);
1345                 return NULL;
1346         }
1347 
1348         ip6_rt_copy_init(rt, res);
1349         rt->rt6i_flags |= RTF_CACHE;
1350         rt->dst.flags |= DST_HOST;
1351         rt->rt6i_dst.addr = *daddr;
1352         rt->rt6i_dst.plen = 128;
1353 
1354         if (!rt6_is_gw_or_nonexthop(res)) {
1355                 if (f6i->fib6_dst.plen != 128 &&
1356                     ipv6_addr_equal(&f6i->fib6_dst.addr, daddr))
1357                         rt->rt6i_flags |= RTF_ANYCAST;
1358 #ifdef CONFIG_IPV6_SUBTREES
1359                 if (rt->rt6i_src.plen && saddr) {
1360                         rt->rt6i_src.addr = *saddr;
1361                         rt->rt6i_src.plen = 128;
1362                 }
1363 #endif
1364         }
1365 
1366         return rt;
1367 }
1368 
1369 static struct rt6_info *ip6_rt_pcpu_alloc(const struct fib6_result *res)
1370 {
1371         struct fib6_info *f6i = res->f6i;
1372         unsigned short flags = fib6_info_dst_flags(f6i);
1373         struct net_device *dev;
1374         struct rt6_info *pcpu_rt;
1375 
1376         if (!fib6_info_hold_safe(f6i))
1377                 return NULL;
1378 
1379         rcu_read_lock();
1380         dev = ip6_rt_get_dev_rcu(res);
1381         pcpu_rt = ip6_dst_alloc(dev_net(dev), dev, flags);
1382         rcu_read_unlock();
1383         if (!pcpu_rt) {
1384                 fib6_info_release(f6i);
1385                 return NULL;
1386         }
1387         ip6_rt_copy_init(pcpu_rt, res);
1388         pcpu_rt->rt6i_flags |= RTF_PCPU;
1389         return pcpu_rt;
1390 }
1391 
1392 /* It should be called with rcu_read_lock() acquired */
1393 static struct rt6_info *rt6_get_pcpu_route(const struct fib6_result *res)
1394 {
1395         struct rt6_info *pcpu_rt;
1396 
1397         pcpu_rt = this_cpu_read(*res->nh->rt6i_pcpu);
1398 
1399         return pcpu_rt;
1400 }
1401 
1402 static struct rt6_info *rt6_make_pcpu_route(struct net *net,
1403                                             const struct fib6_result *res)
1404 {
1405         struct rt6_info *pcpu_rt, *prev, **p;
1406 
1407         pcpu_rt = ip6_rt_pcpu_alloc(res);
1408         if (!pcpu_rt)
1409                 return NULL;
1410 
1411         p = this_cpu_ptr(res->nh->rt6i_pcpu);
1412         prev = cmpxchg(p, NULL, pcpu_rt);
1413         BUG_ON(prev);
1414 
1415         if (res->f6i->fib6_destroying) {
1416                 struct fib6_info *from;
1417 
1418                 from = xchg((__force struct fib6_info **)&pcpu_rt->from, NULL);
1419                 fib6_info_release(from);
1420         }
1421 
1422         return pcpu_rt;
1423 }
1424 
1425 /* exception hash table implementation
1426  */
1427 static DEFINE_SPINLOCK(rt6_exception_lock);
1428 
1429 /* Remove rt6_ex from hash table and free the memory
1430  * Caller must hold rt6_exception_lock
1431  */
1432 static void rt6_remove_exception(struct rt6_exception_bucket *bucket,
1433                                  struct rt6_exception *rt6_ex)
1434 {
1435         struct fib6_info *from;
1436         struct net *net;
1437 
1438         if (!bucket || !rt6_ex)
1439                 return;
1440 
1441         net = dev_net(rt6_ex->rt6i->dst.dev);
1442         net->ipv6.rt6_stats->fib_rt_cache--;
1443 
1444         /* purge completely the exception to allow releasing the held resources:
1445          * some [sk] cache may keep the dst around for unlimited time
1446          */
1447         from = xchg((__force struct fib6_info **)&rt6_ex->rt6i->from, NULL);
1448         fib6_info_release(from);
1449         dst_dev_put(&rt6_ex->rt6i->dst);
1450 
1451         hlist_del_rcu(&rt6_ex->hlist);
1452         dst_release(&rt6_ex->rt6i->dst);
1453         kfree_rcu(rt6_ex, rcu);
1454         WARN_ON_ONCE(!bucket->depth);
1455         bucket->depth--;
1456 }
1457 
1458 /* Remove oldest rt6_ex in bucket and free the memory
1459  * Caller must hold rt6_exception_lock
1460  */
1461 static void rt6_exception_remove_oldest(struct rt6_exception_bucket *bucket)
1462 {
1463         struct rt6_exception *rt6_ex, *oldest = NULL;
1464 
1465         if (!bucket)
1466                 return;
1467 
1468         hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
1469                 if (!oldest || time_before(rt6_ex->stamp, oldest->stamp))
1470                         oldest = rt6_ex;
1471         }
1472         rt6_remove_exception(bucket, oldest);
1473 }
1474 
1475 static u32 rt6_exception_hash(const struct in6_addr *dst,
1476                               const struct in6_addr *src)
1477 {
1478         static u32 seed __read_mostly;
1479         u32 val;
1480 
1481         net_get_random_once(&seed, sizeof(seed));
1482         val = jhash(dst, sizeof(*dst), seed);
1483 
1484 #ifdef CONFIG_IPV6_SUBTREES
1485         if (src)
1486                 val = jhash(src, sizeof(*src), val);
1487 #endif
1488         return hash_32(val, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT);
1489 }
1490 
1491 /* Helper function to find the cached rt in the hash table
1492  * and update bucket pointer to point to the bucket for this
1493  * (daddr, saddr) pair
1494  * Caller must hold rt6_exception_lock
1495  */
1496 static struct rt6_exception *
1497 __rt6_find_exception_spinlock(struct rt6_exception_bucket **bucket,
1498                               const struct in6_addr *daddr,
1499                               const struct in6_addr *saddr)
1500 {
1501         struct rt6_exception *rt6_ex;
1502         u32 hval;
1503 
1504         if (!(*bucket) || !daddr)
1505                 return NULL;
1506 
1507         hval = rt6_exception_hash(daddr, saddr);
1508         *bucket += hval;
1509 
1510         hlist_for_each_entry(rt6_ex, &(*bucket)->chain, hlist) {
1511                 struct rt6_info *rt6 = rt6_ex->rt6i;
1512                 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1513 
1514 #ifdef CONFIG_IPV6_SUBTREES
1515                 if (matched && saddr)
1516                         matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1517 #endif
1518                 if (matched)
1519                         return rt6_ex;
1520         }
1521         return NULL;
1522 }
1523 
1524 /* Helper function to find the cached rt in the hash table
1525  * and update bucket pointer to point to the bucket for this
1526  * (daddr, saddr) pair
1527  * Caller must hold rcu_read_lock()
1528  */
1529 static struct rt6_exception *
1530 __rt6_find_exception_rcu(struct rt6_exception_bucket **bucket,
1531                          const struct in6_addr *daddr,
1532                          const struct in6_addr *saddr)
1533 {
1534         struct rt6_exception *rt6_ex;
1535         u32 hval;
1536 
1537         WARN_ON_ONCE(!rcu_read_lock_held());
1538 
1539         if (!(*bucket) || !daddr)
1540                 return NULL;
1541 
1542         hval = rt6_exception_hash(daddr, saddr);
1543         *bucket += hval;
1544 
1545         hlist_for_each_entry_rcu(rt6_ex, &(*bucket)->chain, hlist) {
1546                 struct rt6_info *rt6 = rt6_ex->rt6i;
1547                 bool matched = ipv6_addr_equal(daddr, &rt6->rt6i_dst.addr);
1548 
1549 #ifdef CONFIG_IPV6_SUBTREES
1550                 if (matched && saddr)
1551                         matched = ipv6_addr_equal(saddr, &rt6->rt6i_src.addr);
1552 #endif
1553                 if (matched)
1554                         return rt6_ex;
1555         }
1556         return NULL;
1557 }
1558 
1559 static unsigned int fib6_mtu(const struct fib6_result *res)
1560 {
1561         const struct fib6_nh *nh = res->nh;
1562         unsigned int mtu;
1563 
1564         if (res->f6i->fib6_pmtu) {
1565                 mtu = res->f6i->fib6_pmtu;
1566         } else {
1567                 struct net_device *dev = nh->fib_nh_dev;
1568                 struct inet6_dev *idev;
1569 
1570                 rcu_read_lock();
1571                 idev = __in6_dev_get(dev);
1572                 mtu = idev->cnf.mtu6;
1573                 rcu_read_unlock();
1574         }
1575 
1576         mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
1577 
1578         return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
1579 }
1580 
1581 #define FIB6_EXCEPTION_BUCKET_FLUSHED  0x1UL
1582 
1583 /* used when the flushed bit is not relevant, only access to the bucket
1584  * (ie., all bucket users except rt6_insert_exception);
1585  *
1586  * called under rcu lock; sometimes called with rt6_exception_lock held
1587  */
1588 static
1589 struct rt6_exception_bucket *fib6_nh_get_excptn_bucket(const struct fib6_nh *nh,
1590                                                        spinlock_t *lock)
1591 {
1592         struct rt6_exception_bucket *bucket;
1593 
1594         if (lock)
1595                 bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1596                                                    lockdep_is_held(lock));
1597         else
1598                 bucket = rcu_dereference(nh->rt6i_exception_bucket);
1599 
1600         /* remove bucket flushed bit if set */
1601         if (bucket) {
1602                 unsigned long p = (unsigned long)bucket;
1603 
1604                 p &= ~FIB6_EXCEPTION_BUCKET_FLUSHED;
1605                 bucket = (struct rt6_exception_bucket *)p;
1606         }
1607 
1608         return bucket;
1609 }
1610 
1611 static bool fib6_nh_excptn_bucket_flushed(struct rt6_exception_bucket *bucket)
1612 {
1613         unsigned long p = (unsigned long)bucket;
1614 
1615         return !!(p & FIB6_EXCEPTION_BUCKET_FLUSHED);
1616 }
1617 
1618 /* called with rt6_exception_lock held */
1619 static void fib6_nh_excptn_bucket_set_flushed(struct fib6_nh *nh,
1620                                               spinlock_t *lock)
1621 {
1622         struct rt6_exception_bucket *bucket;
1623         unsigned long p;
1624 
1625         bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1626                                            lockdep_is_held(lock));
1627 
1628         p = (unsigned long)bucket;
1629         p |= FIB6_EXCEPTION_BUCKET_FLUSHED;
1630         bucket = (struct rt6_exception_bucket *)p;
1631         rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1632 }
1633 
1634 static int rt6_insert_exception(struct rt6_info *nrt,
1635                                 const struct fib6_result *res)
1636 {
1637         struct net *net = dev_net(nrt->dst.dev);
1638         struct rt6_exception_bucket *bucket;
1639         struct fib6_info *f6i = res->f6i;
1640         struct in6_addr *src_key = NULL;
1641         struct rt6_exception *rt6_ex;
1642         struct fib6_nh *nh = res->nh;
1643         int err = 0;
1644 
1645         spin_lock_bh(&rt6_exception_lock);
1646 
1647         bucket = rcu_dereference_protected(nh->rt6i_exception_bucket,
1648                                           lockdep_is_held(&rt6_exception_lock));
1649         if (!bucket) {
1650                 bucket = kcalloc(FIB6_EXCEPTION_BUCKET_SIZE, sizeof(*bucket),
1651                                  GFP_ATOMIC);
1652                 if (!bucket) {
1653                         err = -ENOMEM;
1654                         goto out;
1655                 }
1656                 rcu_assign_pointer(nh->rt6i_exception_bucket, bucket);
1657         } else if (fib6_nh_excptn_bucket_flushed(bucket)) {
1658                 err = -EINVAL;
1659                 goto out;
1660         }
1661 
1662 #ifdef CONFIG_IPV6_SUBTREES
1663         /* fib6_src.plen != 0 indicates f6i is in subtree
1664          * and exception table is indexed by a hash of
1665          * both fib6_dst and fib6_src.
1666          * Otherwise, the exception table is indexed by
1667          * a hash of only fib6_dst.
1668          */
1669         if (f6i->fib6_src.plen)
1670                 src_key = &nrt->rt6i_src.addr;
1671 #endif
1672         /* rt6_mtu_change() might lower mtu on f6i.
1673          * Only insert this exception route if its mtu
1674          * is less than f6i's mtu value.
1675          */
1676         if (dst_metric_raw(&nrt->dst, RTAX_MTU) >= fib6_mtu(res)) {
1677                 err = -EINVAL;
1678                 goto out;
1679         }
1680 
1681         rt6_ex = __rt6_find_exception_spinlock(&bucket, &nrt->rt6i_dst.addr,
1682                                                src_key);
1683         if (rt6_ex)
1684                 rt6_remove_exception(bucket, rt6_ex);
1685 
1686         rt6_ex = kzalloc(sizeof(*rt6_ex), GFP_ATOMIC);
1687         if (!rt6_ex) {
1688                 err = -ENOMEM;
1689                 goto out;
1690         }
1691         rt6_ex->rt6i = nrt;
1692         rt6_ex->stamp = jiffies;
1693         hlist_add_head_rcu(&rt6_ex->hlist, &bucket->chain);
1694         bucket->depth++;
1695         net->ipv6.rt6_stats->fib_rt_cache++;
1696 
1697         if (bucket->depth > FIB6_MAX_DEPTH)
1698                 rt6_exception_remove_oldest(bucket);
1699 
1700 out:
1701         spin_unlock_bh(&rt6_exception_lock);
1702 
1703         /* Update fn->fn_sernum to invalidate all cached dst */
1704         if (!err) {
1705                 spin_lock_bh(&f6i->fib6_table->tb6_lock);
1706                 fib6_update_sernum(net, f6i);
1707                 spin_unlock_bh(&f6i->fib6_table->tb6_lock);
1708                 fib6_force_start_gc(net);
1709         }
1710 
1711         return err;
1712 }
1713 
1714 static void fib6_nh_flush_exceptions(struct fib6_nh *nh, struct fib6_info *from)
1715 {
1716         struct rt6_exception_bucket *bucket;
1717         struct rt6_exception *rt6_ex;
1718         struct hlist_node *tmp;
1719         int i;
1720 
1721         spin_lock_bh(&rt6_exception_lock);
1722 
1723         bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1724         if (!bucket)
1725                 goto out;
1726 
1727         /* Prevent rt6_insert_exception() to recreate the bucket list */
1728         if (!from)
1729                 fib6_nh_excptn_bucket_set_flushed(nh, &rt6_exception_lock);
1730 
1731         for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
1732                 hlist_for_each_entry_safe(rt6_ex, tmp, &bucket->chain, hlist) {
1733                         if (!from ||
1734                             rcu_access_pointer(rt6_ex->rt6i->from) == from)
1735                                 rt6_remove_exception(bucket, rt6_ex);
1736                 }
1737                 WARN_ON_ONCE(!from && bucket->depth);
1738                 bucket++;
1739         }
1740 out:
1741         spin_unlock_bh(&rt6_exception_lock);
1742 }
1743 
1744 static int rt6_nh_flush_exceptions(struct fib6_nh *nh, void *arg)
1745 {
1746         struct fib6_info *f6i = arg;
1747 
1748         fib6_nh_flush_exceptions(nh, f6i);
1749 
1750         return 0;
1751 }
1752 
1753 void rt6_flush_exceptions(struct fib6_info *f6i)
1754 {
1755         if (f6i->nh)
1756                 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_flush_exceptions,
1757                                          f6i);
1758         else
1759                 fib6_nh_flush_exceptions(f6i->fib6_nh, f6i);
1760 }
1761 
1762 /* Find cached rt in the hash table inside passed in rt
1763  * Caller has to hold rcu_read_lock()
1764  */
1765 static struct rt6_info *rt6_find_cached_rt(const struct fib6_result *res,
1766                                            const struct in6_addr *daddr,
1767                                            const struct in6_addr *saddr)
1768 {
1769         const struct in6_addr *src_key = NULL;
1770         struct rt6_exception_bucket *bucket;
1771         struct rt6_exception *rt6_ex;
1772         struct rt6_info *ret = NULL;
1773 
1774 #ifdef CONFIG_IPV6_SUBTREES
1775         /* fib6i_src.plen != 0 indicates f6i is in subtree
1776          * and exception table is indexed by a hash of
1777          * both fib6_dst and fib6_src.
1778          * However, the src addr used to create the hash
1779          * might not be exactly the passed in saddr which
1780          * is a /128 addr from the flow.
1781          * So we need to use f6i->fib6_src to redo lookup
1782          * if the passed in saddr does not find anything.
1783          * (See the logic in ip6_rt_cache_alloc() on how
1784          * rt->rt6i_src is updated.)
1785          */
1786         if (res->f6i->fib6_src.plen)
1787                 src_key = saddr;
1788 find_ex:
1789 #endif
1790         bucket = fib6_nh_get_excptn_bucket(res->nh, NULL);
1791         rt6_ex = __rt6_find_exception_rcu(&bucket, daddr, src_key);
1792 
1793         if (rt6_ex && !rt6_check_expired(rt6_ex->rt6i))
1794                 ret = rt6_ex->rt6i;
1795 
1796 #ifdef CONFIG_IPV6_SUBTREES
1797         /* Use fib6_src as src_key and redo lookup */
1798         if (!ret && src_key && src_key != &res->f6i->fib6_src.addr) {
1799                 src_key = &res->f6i->fib6_src.addr;
1800                 goto find_ex;
1801         }
1802 #endif
1803 
1804         return ret;
1805 }
1806 
1807 /* Remove the passed in cached rt from the hash table that contains it */
1808 static int fib6_nh_remove_exception(const struct fib6_nh *nh, int plen,
1809                                     const struct rt6_info *rt)
1810 {
1811         const struct in6_addr *src_key = NULL;
1812         struct rt6_exception_bucket *bucket;
1813         struct rt6_exception *rt6_ex;
1814         int err;
1815 
1816         if (!rcu_access_pointer(nh->rt6i_exception_bucket))
1817                 return -ENOENT;
1818 
1819         spin_lock_bh(&rt6_exception_lock);
1820         bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1821 
1822 #ifdef CONFIG_IPV6_SUBTREES
1823         /* rt6i_src.plen != 0 indicates 'from' is in subtree
1824          * and exception table is indexed by a hash of
1825          * both rt6i_dst and rt6i_src.
1826          * Otherwise, the exception table is indexed by
1827          * a hash of only rt6i_dst.
1828          */
1829         if (plen)
1830                 src_key = &rt->rt6i_src.addr;
1831 #endif
1832         rt6_ex = __rt6_find_exception_spinlock(&bucket,
1833                                                &rt->rt6i_dst.addr,
1834                                                src_key);
1835         if (rt6_ex) {
1836                 rt6_remove_exception(bucket, rt6_ex);
1837                 err = 0;
1838         } else {
1839                 err = -ENOENT;
1840         }
1841 
1842         spin_unlock_bh(&rt6_exception_lock);
1843         return err;
1844 }
1845 
1846 struct fib6_nh_excptn_arg {
1847         struct rt6_info *rt;
1848         int             plen;
1849 };
1850 
1851 static int rt6_nh_remove_exception_rt(struct fib6_nh *nh, void *_arg)
1852 {
1853         struct fib6_nh_excptn_arg *arg = _arg;
1854         int err;
1855 
1856         err = fib6_nh_remove_exception(nh, arg->plen, arg->rt);
1857         if (err == 0)
1858                 return 1;
1859 
1860         return 0;
1861 }
1862 
1863 static int rt6_remove_exception_rt(struct rt6_info *rt)
1864 {
1865         struct fib6_info *from;
1866 
1867         from = rcu_dereference(rt->from);
1868         if (!from || !(rt->rt6i_flags & RTF_CACHE))
1869                 return -EINVAL;
1870 
1871         if (from->nh) {
1872                 struct fib6_nh_excptn_arg arg = {
1873                         .rt = rt,
1874                         .plen = from->fib6_src.plen
1875                 };
1876                 int rc;
1877 
1878                 /* rc = 1 means an entry was found */
1879                 rc = nexthop_for_each_fib6_nh(from->nh,
1880                                               rt6_nh_remove_exception_rt,
1881                                               &arg);
1882                 return rc ? 0 : -ENOENT;
1883         }
1884 
1885         return fib6_nh_remove_exception(from->fib6_nh,
1886                                         from->fib6_src.plen, rt);
1887 }
1888 
1889 /* Find rt6_ex which contains the passed in rt cache and
1890  * refresh its stamp
1891  */
1892 static void fib6_nh_update_exception(const struct fib6_nh *nh, int plen,
1893                                      const struct rt6_info *rt)
1894 {
1895         const struct in6_addr *src_key = NULL;
1896         struct rt6_exception_bucket *bucket;
1897         struct rt6_exception *rt6_ex;
1898 
1899         bucket = fib6_nh_get_excptn_bucket(nh, NULL);
1900 #ifdef CONFIG_IPV6_SUBTREES
1901         /* rt6i_src.plen != 0 indicates 'from' is in subtree
1902          * and exception table is indexed by a hash of
1903          * both rt6i_dst and rt6i_src.
1904          * Otherwise, the exception table is indexed by
1905          * a hash of only rt6i_dst.
1906          */
1907         if (plen)
1908                 src_key = &rt->rt6i_src.addr;
1909 #endif
1910         rt6_ex = __rt6_find_exception_rcu(&bucket, &rt->rt6i_dst.addr, src_key);
1911         if (rt6_ex)
1912                 rt6_ex->stamp = jiffies;
1913 }
1914 
1915 struct fib6_nh_match_arg {
1916         const struct net_device *dev;
1917         const struct in6_addr   *gw;
1918         struct fib6_nh          *match;
1919 };
1920 
1921 /* determine if fib6_nh has given device and gateway */
1922 static int fib6_nh_find_match(struct fib6_nh *nh, void *_arg)
1923 {
1924         struct fib6_nh_match_arg *arg = _arg;
1925 
1926         if (arg->dev != nh->fib_nh_dev ||
1927             (arg->gw && !nh->fib_nh_gw_family) ||
1928             (!arg->gw && nh->fib_nh_gw_family) ||
1929             (arg->gw && !ipv6_addr_equal(arg->gw, &nh->fib_nh_gw6)))
1930                 return 0;
1931 
1932         arg->match = nh;
1933 
1934         /* found a match, break the loop */
1935         return 1;
1936 }
1937 
1938 static void rt6_update_exception_stamp_rt(struct rt6_info *rt)
1939 {
1940         struct fib6_info *from;
1941         struct fib6_nh *fib6_nh;
1942 
1943         rcu_read_lock();
1944 
1945         from = rcu_dereference(rt->from);
1946         if (!from || !(rt->rt6i_flags & RTF_CACHE))
1947                 goto unlock;
1948 
1949         if (from->nh) {
1950                 struct fib6_nh_match_arg arg = {
1951                         .dev = rt->dst.dev,
1952                         .gw = &rt->rt6i_gateway,
1953                 };
1954 
1955                 nexthop_for_each_fib6_nh(from->nh, fib6_nh_find_match, &arg);
1956 
1957                 if (!arg.match)
1958                         goto unlock;
1959                 fib6_nh = arg.match;
1960         } else {
1961                 fib6_nh = from->fib6_nh;
1962         }
1963         fib6_nh_update_exception(fib6_nh, from->fib6_src.plen, rt);
1964 unlock:
1965         rcu_read_unlock();
1966 }
1967 
1968 static bool rt6_mtu_change_route_allowed(struct inet6_dev *idev,
1969                                          struct rt6_info *rt, int mtu)
1970 {
1971         /* If the new MTU is lower than the route PMTU, this new MTU will be the
1972          * lowest MTU in the path: always allow updating the route PMTU to
1973          * reflect PMTU decreases.
1974          *
1975          * If the new MTU is higher, and the route PMTU is equal to the local
1976          * MTU, this means the old MTU is the lowest in the path, so allow
1977          * updating it: if other nodes now have lower MTUs, PMTU discovery will
1978          * handle this.
1979          */
1980 
1981         if (dst_mtu(&rt->dst) >= mtu)
1982                 return true;
1983 
1984         if (dst_mtu(&rt->dst) == idev->cnf.mtu6)
1985                 return true;
1986 
1987         return false;
1988 }
1989 
1990 static void rt6_exceptions_update_pmtu(struct inet6_dev *idev,
1991                                        const struct fib6_nh *nh, int mtu)
1992 {
1993         struct rt6_exception_bucket *bucket;
1994         struct rt6_exception *rt6_ex;
1995         int i;
1996 
1997         bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
1998         if (!bucket)
1999                 return;
2000 
2001         for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2002                 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
2003                         struct rt6_info *entry = rt6_ex->rt6i;
2004 
2005                         /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
2006                          * route), the metrics of its rt->from have already
2007                          * been updated.
2008                          */
2009                         if (dst_metric_raw(&entry->dst, RTAX_MTU) &&
2010                             rt6_mtu_change_route_allowed(idev, entry, mtu))
2011                                 dst_metric_set(&entry->dst, RTAX_MTU, mtu);
2012                 }
2013                 bucket++;
2014         }
2015 }
2016 
2017 #define RTF_CACHE_GATEWAY       (RTF_GATEWAY | RTF_CACHE)
2018 
2019 static void fib6_nh_exceptions_clean_tohost(const struct fib6_nh *nh,
2020                                             const struct in6_addr *gateway)
2021 {
2022         struct rt6_exception_bucket *bucket;
2023         struct rt6_exception *rt6_ex;
2024         struct hlist_node *tmp;
2025         int i;
2026 
2027         if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2028                 return;
2029 
2030         spin_lock_bh(&rt6_exception_lock);
2031         bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2032         if (bucket) {
2033                 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2034                         hlist_for_each_entry_safe(rt6_ex, tmp,
2035                                                   &bucket->chain, hlist) {
2036                                 struct rt6_info *entry = rt6_ex->rt6i;
2037 
2038                                 if ((entry->rt6i_flags & RTF_CACHE_GATEWAY) ==
2039                                     RTF_CACHE_GATEWAY &&
2040                                     ipv6_addr_equal(gateway,
2041                                                     &entry->rt6i_gateway)) {
2042                                         rt6_remove_exception(bucket, rt6_ex);
2043                                 }
2044                         }
2045                         bucket++;
2046                 }
2047         }
2048 
2049         spin_unlock_bh(&rt6_exception_lock);
2050 }
2051 
2052 static void rt6_age_examine_exception(struct rt6_exception_bucket *bucket,
2053                                       struct rt6_exception *rt6_ex,
2054                                       struct fib6_gc_args *gc_args,
2055                                       unsigned long now)
2056 {
2057         struct rt6_info *rt = rt6_ex->rt6i;
2058 
2059         /* we are pruning and obsoleting aged-out and non gateway exceptions
2060          * even if others have still references to them, so that on next
2061          * dst_check() such references can be dropped.
2062          * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
2063          * expired, independently from their aging, as per RFC 8201 section 4
2064          */
2065         if (!(rt->rt6i_flags & RTF_EXPIRES)) {
2066                 if (time_after_eq(now, rt->dst.lastuse + gc_args->timeout)) {
2067                         RT6_TRACE("aging clone %p\n", rt);
2068                         rt6_remove_exception(bucket, rt6_ex);
2069                         return;
2070                 }
2071         } else if (time_after(jiffies, rt->dst.expires)) {
2072                 RT6_TRACE("purging expired route %p\n", rt);
2073                 rt6_remove_exception(bucket, rt6_ex);
2074                 return;
2075         }
2076 
2077         if (rt->rt6i_flags & RTF_GATEWAY) {
2078                 struct neighbour *neigh;
2079                 __u8 neigh_flags = 0;
2080 
2081                 neigh = __ipv6_neigh_lookup_noref(rt->dst.dev, &rt->rt6i_gateway);
2082                 if (neigh)
2083                         neigh_flags = neigh->flags;
2084 
2085                 if (!(neigh_flags & NTF_ROUTER)) {
2086                         RT6_TRACE("purging route %p via non-router but gateway\n",
2087                                   rt);
2088                         rt6_remove_exception(bucket, rt6_ex);
2089                         return;
2090                 }
2091         }
2092 
2093         gc_args->more++;
2094 }
2095 
2096 static void fib6_nh_age_exceptions(const struct fib6_nh *nh,
2097                                    struct fib6_gc_args *gc_args,
2098                                    unsigned long now)
2099 {
2100         struct rt6_exception_bucket *bucket;
2101         struct rt6_exception *rt6_ex;
2102         struct hlist_node *tmp;
2103         int i;
2104 
2105         if (!rcu_access_pointer(nh->rt6i_exception_bucket))
2106                 return;
2107 
2108         rcu_read_lock_bh();
2109         spin_lock(&rt6_exception_lock);
2110         bucket = fib6_nh_get_excptn_bucket(nh, &rt6_exception_lock);
2111         if (bucket) {
2112                 for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
2113                         hlist_for_each_entry_safe(rt6_ex, tmp,
2114                                                   &bucket->chain, hlist) {
2115                                 rt6_age_examine_exception(bucket, rt6_ex,
2116                                                           gc_args, now);
2117                         }
2118                         bucket++;
2119                 }
2120         }
2121         spin_unlock(&rt6_exception_lock);
2122         rcu_read_unlock_bh();
2123 }
2124 
2125 struct fib6_nh_age_excptn_arg {
2126         struct fib6_gc_args     *gc_args;
2127         unsigned long           now;
2128 };
2129 
2130 static int rt6_nh_age_exceptions(struct fib6_nh *nh, void *_arg)
2131 {
2132         struct fib6_nh_age_excptn_arg *arg = _arg;
2133 
2134         fib6_nh_age_exceptions(nh, arg->gc_args, arg->now);
2135         return 0;
2136 }
2137 
2138 void rt6_age_exceptions(struct fib6_info *f6i,
2139                         struct fib6_gc_args *gc_args,
2140                         unsigned long now)
2141 {
2142         if (f6i->nh) {
2143                 struct fib6_nh_age_excptn_arg arg = {
2144                         .gc_args = gc_args,
2145                         .now = now
2146                 };
2147 
2148                 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_age_exceptions,
2149                                          &arg);
2150         } else {
2151                 fib6_nh_age_exceptions(f6i->fib6_nh, gc_args, now);
2152         }
2153 }
2154 
2155 /* must be called with rcu lock held */
2156 int fib6_table_lookup(struct net *net, struct fib6_table *table, int oif,
2157                       struct flowi6 *fl6, struct fib6_result *res, int strict)
2158 {
2159         struct fib6_node *fn, *saved_fn;
2160 
2161         fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2162         saved_fn = fn;
2163 
2164         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2165                 oif = 0;
2166 
2167 redo_rt6_select:
2168         rt6_select(net, fn, oif, res, strict);
2169         if (res->f6i == net->ipv6.fib6_null_entry) {
2170                 fn = fib6_backtrack(fn, &fl6->saddr);
2171                 if (fn)
2172                         goto redo_rt6_select;
2173                 else if (strict & RT6_LOOKUP_F_REACHABLE) {
2174                         /* also consider unreachable route */
2175                         strict &= ~RT6_LOOKUP_F_REACHABLE;
2176                         fn = saved_fn;
2177                         goto redo_rt6_select;
2178                 }
2179         }
2180 
2181         trace_fib6_table_lookup(net, res, table, fl6);
2182 
2183         return 0;
2184 }
2185 
2186 struct rt6_info *ip6_pol_route(struct net *net, struct fib6_table *table,
2187                                int oif, struct flowi6 *fl6,
2188                                const struct sk_buff *skb, int flags)
2189 {
2190         struct fib6_result res = {};
2191         struct rt6_info *rt = NULL;
2192         int strict = 0;
2193 
2194         WARN_ON_ONCE((flags & RT6_LOOKUP_F_DST_NOREF) &&
2195                      !rcu_read_lock_held());
2196 
2197         strict |= flags & RT6_LOOKUP_F_IFACE;
2198         strict |= flags & RT6_LOOKUP_F_IGNORE_LINKSTATE;
2199         if (net->ipv6.devconf_all->forwarding == 0)
2200                 strict |= RT6_LOOKUP_F_REACHABLE;
2201 
2202         rcu_read_lock();
2203 
2204         fib6_table_lookup(net, table, oif, fl6, &res, strict);
2205         if (res.f6i == net->ipv6.fib6_null_entry)
2206                 goto out;
2207 
2208         fib6_select_path(net, &res, fl6, oif, false, skb, strict);
2209 
2210         /*Search through exception table */
2211         rt = rt6_find_cached_rt(&res, &fl6->daddr, &fl6->saddr);
2212         if (rt) {
2213                 goto out;
2214         } else if (unlikely((fl6->flowi6_flags & FLOWI_FLAG_KNOWN_NH) &&
2215                             !res.nh->fib_nh_gw_family)) {
2216                 /* Create a RTF_CACHE clone which will not be
2217                  * owned by the fib6 tree.  It is for the special case where
2218                  * the daddr in the skb during the neighbor look-up is different
2219                  * from the fl6->daddr used to look-up route here.
2220                  */
2221                 rt = ip6_rt_cache_alloc(&res, &fl6->daddr, NULL);
2222 
2223                 if (rt) {
2224                         /* 1 refcnt is taken during ip6_rt_cache_alloc().
2225                          * As rt6_uncached_list_add() does not consume refcnt,
2226                          * this refcnt is always returned to the caller even
2227                          * if caller sets RT6_LOOKUP_F_DST_NOREF flag.
2228                          */
2229                         rt6_uncached_list_add(rt);
2230                         atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
2231                         rcu_read_unlock();
2232 
2233                         return rt;
2234                 }
2235         } else {
2236                 /* Get a percpu copy */
2237                 local_bh_disable();
2238                 rt = rt6_get_pcpu_route(&res);
2239 
2240                 if (!rt)
2241                         rt = rt6_make_pcpu_route(net, &res);
2242 
2243                 local_bh_enable();
2244         }
2245 out:
2246         if (!rt)
2247                 rt = net->ipv6.ip6_null_entry;
2248         if (!(flags & RT6_LOOKUP_F_DST_NOREF))
2249                 ip6_hold_safe(net, &rt);
2250         rcu_read_unlock();
2251 
2252         return rt;
2253 }
2254 EXPORT_SYMBOL_GPL(ip6_pol_route);
2255 
2256 static struct rt6_info *ip6_pol_route_input(struct net *net,
2257                                             struct fib6_table *table,
2258                                             struct flowi6 *fl6,
2259                                             const struct sk_buff *skb,
2260                                             int flags)
2261 {
2262         return ip6_pol_route(net, table, fl6->flowi6_iif, fl6, skb, flags);
2263 }
2264 
2265 struct dst_entry *ip6_route_input_lookup(struct net *net,
2266                                          struct net_device *dev,
2267                                          struct flowi6 *fl6,
2268                                          const struct sk_buff *skb,
2269                                          int flags)
2270 {
2271         if (rt6_need_strict(&fl6->daddr) && dev->type != ARPHRD_PIMREG)
2272                 flags |= RT6_LOOKUP_F_IFACE;
2273 
2274         return fib6_rule_lookup(net, fl6, skb, flags, ip6_pol_route_input);
2275 }
2276 EXPORT_SYMBOL_GPL(ip6_route_input_lookup);
2277 
2278 static void ip6_multipath_l3_keys(const struct sk_buff *skb,
2279                                   struct flow_keys *keys,
2280                                   struct flow_keys *flkeys)
2281 {
2282         const struct ipv6hdr *outer_iph = ipv6_hdr(skb);
2283         const struct ipv6hdr *key_iph = outer_iph;
2284         struct flow_keys *_flkeys = flkeys;
2285         const struct ipv6hdr *inner_iph;
2286         const struct icmp6hdr *icmph;
2287         struct ipv6hdr _inner_iph;
2288         struct icmp6hdr _icmph;
2289 
2290         if (likely(outer_iph->nexthdr != IPPROTO_ICMPV6))
2291                 goto out;
2292 
2293         icmph = skb_header_pointer(skb, skb_transport_offset(skb),
2294                                    sizeof(_icmph), &_icmph);
2295         if (!icmph)
2296                 goto out;
2297 
2298         if (icmph->icmp6_type != ICMPV6_DEST_UNREACH &&
2299             icmph->icmp6_type != ICMPV6_PKT_TOOBIG &&
2300             icmph->icmp6_type != ICMPV6_TIME_EXCEED &&
2301             icmph->icmp6_type != ICMPV6_PARAMPROB)
2302                 goto out;
2303 
2304         inner_iph = skb_header_pointer(skb,
2305                                        skb_transport_offset(skb) + sizeof(*icmph),
2306                                        sizeof(_inner_iph), &_inner_iph);
2307         if (!inner_iph)
2308                 goto out;
2309 
2310         key_iph = inner_iph;
2311         _flkeys = NULL;
2312 out:
2313         if (_flkeys) {
2314                 keys->addrs.v6addrs.src = _flkeys->addrs.v6addrs.src;
2315                 keys->addrs.v6addrs.dst = _flkeys->addrs.v6addrs.dst;
2316                 keys->tags.flow_label = _flkeys->tags.flow_label;
2317                 keys->basic.ip_proto = _flkeys->basic.ip_proto;
2318         } else {
2319                 keys->addrs.v6addrs.src = key_iph->saddr;
2320                 keys->addrs.v6addrs.dst = key_iph->daddr;
2321                 keys->tags.flow_label = ip6_flowlabel(key_iph);
2322                 keys->basic.ip_proto = key_iph->nexthdr;
2323         }
2324 }
2325 
2326 /* if skb is set it will be used and fl6 can be NULL */
2327 u32 rt6_multipath_hash(const struct net *net, const struct flowi6 *fl6,
2328                        const struct sk_buff *skb, struct flow_keys *flkeys)
2329 {
2330         struct flow_keys hash_keys;
2331         u32 mhash;
2332 
2333         switch (ip6_multipath_hash_policy(net)) {
2334         case 0:
2335                 memset(&hash_keys, 0, sizeof(hash_keys));
2336                 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2337                 if (skb) {
2338                         ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2339                 } else {
2340                         hash_keys.addrs.v6addrs.src = fl6->saddr;
2341                         hash_keys.addrs.v6addrs.dst = fl6->daddr;
2342                         hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2343                         hash_keys.basic.ip_proto = fl6->flowi6_proto;
2344                 }
2345                 break;
2346         case 1:
2347                 if (skb) {
2348                         unsigned int flag = FLOW_DISSECTOR_F_STOP_AT_ENCAP;
2349                         struct flow_keys keys;
2350 
2351                         /* short-circuit if we already have L4 hash present */
2352                         if (skb->l4_hash)
2353                                 return skb_get_hash_raw(skb) >> 1;
2354 
2355                         memset(&hash_keys, 0, sizeof(hash_keys));
2356 
2357                         if (!flkeys) {
2358                                 skb_flow_dissect_flow_keys(skb, &keys, flag);
2359                                 flkeys = &keys;
2360                         }
2361                         hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2362                         hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2363                         hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2364                         hash_keys.ports.src = flkeys->ports.src;
2365                         hash_keys.ports.dst = flkeys->ports.dst;
2366                         hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2367                 } else {
2368                         memset(&hash_keys, 0, sizeof(hash_keys));
2369                         hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2370                         hash_keys.addrs.v6addrs.src = fl6->saddr;
2371                         hash_keys.addrs.v6addrs.dst = fl6->daddr;
2372                         hash_keys.ports.src = fl6->fl6_sport;
2373                         hash_keys.ports.dst = fl6->fl6_dport;
2374                         hash_keys.basic.ip_proto = fl6->flowi6_proto;
2375                 }
2376                 break;
2377         case 2:
2378                 memset(&hash_keys, 0, sizeof(hash_keys));
2379                 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2380                 if (skb) {
2381                         struct flow_keys keys;
2382 
2383                         if (!flkeys) {
2384                                 skb_flow_dissect_flow_keys(skb, &keys, 0);
2385                                 flkeys = &keys;
2386                         }
2387 
2388                         /* Inner can be v4 or v6 */
2389                         if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV4_ADDRS) {
2390                                 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV4_ADDRS;
2391                                 hash_keys.addrs.v4addrs.src = flkeys->addrs.v4addrs.src;
2392                                 hash_keys.addrs.v4addrs.dst = flkeys->addrs.v4addrs.dst;
2393                         } else if (flkeys->control.addr_type == FLOW_DISSECTOR_KEY_IPV6_ADDRS) {
2394                                 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2395                                 hash_keys.addrs.v6addrs.src = flkeys->addrs.v6addrs.src;
2396                                 hash_keys.addrs.v6addrs.dst = flkeys->addrs.v6addrs.dst;
2397                                 hash_keys.tags.flow_label = flkeys->tags.flow_label;
2398                                 hash_keys.basic.ip_proto = flkeys->basic.ip_proto;
2399                         } else {
2400                                 /* Same as case 0 */
2401                                 hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2402                                 ip6_multipath_l3_keys(skb, &hash_keys, flkeys);
2403                         }
2404                 } else {
2405                         /* Same as case 0 */
2406                         hash_keys.control.addr_type = FLOW_DISSECTOR_KEY_IPV6_ADDRS;
2407                         hash_keys.addrs.v6addrs.src = fl6->saddr;
2408                         hash_keys.addrs.v6addrs.dst = fl6->daddr;
2409                         hash_keys.tags.flow_label = (__force u32)flowi6_get_flowlabel(fl6);
2410                         hash_keys.basic.ip_proto = fl6->flowi6_proto;
2411                 }
2412                 break;
2413         }
2414         mhash = flow_hash_from_keys(&hash_keys);
2415 
2416         return mhash >> 1;
2417 }
2418 
2419 /* Called with rcu held */
2420 void ip6_route_input(struct sk_buff *skb)
2421 {
2422         const struct ipv6hdr *iph = ipv6_hdr(skb);
2423         struct net *net = dev_net(skb->dev);
2424         int flags = RT6_LOOKUP_F_HAS_SADDR | RT6_LOOKUP_F_DST_NOREF;
2425         struct ip_tunnel_info *tun_info;
2426         struct flowi6 fl6 = {
2427                 .flowi6_iif = skb->dev->ifindex,
2428                 .daddr = iph->daddr,
2429                 .saddr = iph->saddr,
2430                 .flowlabel = ip6_flowinfo(iph),
2431                 .flowi6_mark = skb->mark,
2432                 .flowi6_proto = iph->nexthdr,
2433         };
2434         struct flow_keys *flkeys = NULL, _flkeys;
2435 
2436         tun_info = skb_tunnel_info(skb);
2437         if (tun_info && !(tun_info->mode & IP_TUNNEL_INFO_TX))
2438                 fl6.flowi6_tun_key.tun_id = tun_info->key.tun_id;
2439 
2440         if (fib6_rules_early_flow_dissect(net, skb, &fl6, &_flkeys))
2441                 flkeys = &_flkeys;
2442 
2443         if (unlikely(fl6.flowi6_proto == IPPROTO_ICMPV6))
2444                 fl6.mp_hash = rt6_multipath_hash(net, &fl6, skb, flkeys);
2445         skb_dst_drop(skb);
2446         skb_dst_set_noref(skb, ip6_route_input_lookup(net, skb->dev,
2447                                                       &fl6, skb, flags));
2448 }
2449 
2450 static struct rt6_info *ip6_pol_route_output(struct net *net,
2451                                              struct fib6_table *table,
2452                                              struct flowi6 *fl6,
2453                                              const struct sk_buff *skb,
2454                                              int flags)
2455 {
2456         return ip6_pol_route(net, table, fl6->flowi6_oif, fl6, skb, flags);
2457 }
2458 
2459 struct dst_entry *ip6_route_output_flags_noref(struct net *net,
2460                                                const struct sock *sk,
2461                                                struct flowi6 *fl6, int flags)
2462 {
2463         bool any_src;
2464 
2465         if (ipv6_addr_type(&fl6->daddr) &
2466             (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL)) {
2467                 struct dst_entry *dst;
2468 
2469                 /* This function does not take refcnt on the dst */
2470                 dst = l3mdev_link_scope_lookup(net, fl6);
2471                 if (dst)
2472                         return dst;
2473         }
2474 
2475         fl6->flowi6_iif = LOOPBACK_IFINDEX;
2476 
2477         flags |= RT6_LOOKUP_F_DST_NOREF;
2478         any_src = ipv6_addr_any(&fl6->saddr);
2479         if ((sk && sk->sk_bound_dev_if) || rt6_need_strict(&fl6->daddr) ||
2480             (fl6->flowi6_oif && any_src))
2481                 flags |= RT6_LOOKUP_F_IFACE;
2482 
2483         if (!any_src)
2484                 flags |= RT6_LOOKUP_F_HAS_SADDR;
2485         else if (sk)
2486                 flags |= rt6_srcprefs2flags(inet6_sk(sk)->srcprefs);
2487 
2488         return fib6_rule_lookup(net, fl6, NULL, flags, ip6_pol_route_output);
2489 }
2490 EXPORT_SYMBOL_GPL(ip6_route_output_flags_noref);
2491 
2492 struct dst_entry *ip6_route_output_flags(struct net *net,
2493                                          const struct sock *sk,
2494                                          struct flowi6 *fl6,
2495                                          int flags)
2496 {
2497         struct dst_entry *dst;
2498         struct rt6_info *rt6;
2499 
2500         rcu_read_lock();
2501         dst = ip6_route_output_flags_noref(net, sk, fl6, flags);
2502         rt6 = (struct rt6_info *)dst;
2503         /* For dst cached in uncached_list, refcnt is already taken. */
2504         if (list_empty(&rt6->rt6i_uncached) && !dst_hold_safe(dst)) {
2505                 dst = &net->ipv6.ip6_null_entry->dst;
2506                 dst_hold(dst);
2507         }
2508         rcu_read_unlock();
2509 
2510         return dst;
2511 }
2512 EXPORT_SYMBOL_GPL(ip6_route_output_flags);
2513 
2514 struct dst_entry *ip6_blackhole_route(struct net *net, struct dst_entry *dst_orig)
2515 {
2516         struct rt6_info *rt, *ort = (struct rt6_info *) dst_orig;
2517         struct net_device *loopback_dev = net->loopback_dev;
2518         struct dst_entry *new = NULL;
2519 
2520         rt = dst_alloc(&ip6_dst_blackhole_ops, loopback_dev, 1,
2521                        DST_OBSOLETE_DEAD, 0);
2522         if (rt) {
2523                 rt6_info_init(rt);
2524                 atomic_inc(&net->ipv6.rt6_stats->fib_rt_alloc);
2525 
2526                 new = &rt->dst;
2527                 new->__use = 1;
2528                 new->input = dst_discard;
2529                 new->output = dst_discard_out;
2530 
2531                 dst_copy_metrics(new, &ort->dst);
2532 
2533                 rt->rt6i_idev = in6_dev_get(loopback_dev);
2534                 rt->rt6i_gateway = ort->rt6i_gateway;
2535                 rt->rt6i_flags = ort->rt6i_flags & ~RTF_PCPU;
2536 
2537                 memcpy(&rt->rt6i_dst, &ort->rt6i_dst, sizeof(struct rt6key));
2538 #ifdef CONFIG_IPV6_SUBTREES
2539                 memcpy(&rt->rt6i_src, &ort->rt6i_src, sizeof(struct rt6key));
2540 #endif
2541         }
2542 
2543         dst_release(dst_orig);
2544         return new ? new : ERR_PTR(-ENOMEM);
2545 }
2546 
2547 /*
2548  *      Destination cache support functions
2549  */
2550 
2551 static bool fib6_check(struct fib6_info *f6i, u32 cookie)
2552 {
2553         u32 rt_cookie = 0;
2554 
2555         if (!fib6_get_cookie_safe(f6i, &rt_cookie) || rt_cookie != cookie)
2556                 return false;
2557 
2558         if (fib6_check_expired(f6i))
2559                 return false;
2560 
2561         return true;
2562 }
2563 
2564 static struct dst_entry *rt6_check(struct rt6_info *rt,
2565                                    struct fib6_info *from,
2566                                    u32 cookie)
2567 {
2568         u32 rt_cookie = 0;
2569 
2570         if (!from || !fib6_get_cookie_safe(from, &rt_cookie) ||
2571             rt_cookie != cookie)
2572                 return NULL;
2573 
2574         if (rt6_check_expired(rt))
2575                 return NULL;
2576 
2577         return &rt->dst;
2578 }
2579 
2580 static struct dst_entry *rt6_dst_from_check(struct rt6_info *rt,
2581                                             struct fib6_info *from,
2582                                             u32 cookie)
2583 {
2584         if (!__rt6_check_expired(rt) &&
2585             rt->dst.obsolete == DST_OBSOLETE_FORCE_CHK &&
2586             fib6_check(from, cookie))
2587                 return &rt->dst;
2588         else
2589                 return NULL;
2590 }
2591 
2592 static struct dst_entry *ip6_dst_check(struct dst_entry *dst, u32 cookie)
2593 {
2594         struct dst_entry *dst_ret;
2595         struct fib6_info *from;
2596         struct rt6_info *rt;
2597 
2598         rt = container_of(dst, struct rt6_info, dst);
2599 
2600         rcu_read_lock();
2601 
2602         /* All IPV6 dsts are created with ->obsolete set to the value
2603          * DST_OBSOLETE_FORCE_CHK which forces validation calls down
2604          * into this function always.
2605          */
2606 
2607         from = rcu_dereference(rt->from);
2608 
2609         if (from && (rt->rt6i_flags & RTF_PCPU ||
2610             unlikely(!list_empty(&rt->rt6i_uncached))))
2611                 dst_ret = rt6_dst_from_check(rt, from, cookie);
2612         else
2613                 dst_ret = rt6_check(rt, from, cookie);
2614 
2615         rcu_read_unlock();
2616 
2617         return dst_ret;
2618 }
2619 
2620 static struct dst_entry *ip6_negative_advice(struct dst_entry *dst)
2621 {
2622         struct rt6_info *rt = (struct rt6_info *) dst;
2623 
2624         if (rt) {
2625                 if (rt->rt6i_flags & RTF_CACHE) {
2626                         rcu_read_lock();
2627                         if (rt6_check_expired(rt)) {
2628                                 rt6_remove_exception_rt(rt);
2629                                 dst = NULL;
2630                         }
2631                         rcu_read_unlock();
2632                 } else {
2633                         dst_release(dst);
2634                         dst = NULL;
2635                 }
2636         }
2637         return dst;
2638 }
2639 
2640 static void ip6_link_failure(struct sk_buff *skb)
2641 {
2642         struct rt6_info *rt;
2643 
2644         icmpv6_send(skb, ICMPV6_DEST_UNREACH, ICMPV6_ADDR_UNREACH, 0);
2645 
2646         rt = (struct rt6_info *) skb_dst(skb);
2647         if (rt) {
2648                 rcu_read_lock();
2649                 if (rt->rt6i_flags & RTF_CACHE) {
2650                         rt6_remove_exception_rt(rt);
2651                 } else {
2652                         struct fib6_info *from;
2653                         struct fib6_node *fn;
2654 
2655                         from = rcu_dereference(rt->from);
2656                         if (from) {
2657                                 fn = rcu_dereference(from->fib6_node);
2658                                 if (fn && (rt->rt6i_flags & RTF_DEFAULT))
2659                                         fn->fn_sernum = -1;
2660                         }
2661                 }
2662                 rcu_read_unlock();
2663         }
2664 }
2665 
2666 static void rt6_update_expires(struct rt6_info *rt0, int timeout)
2667 {
2668         if (!(rt0->rt6i_flags & RTF_EXPIRES)) {
2669                 struct fib6_info *from;
2670 
2671                 rcu_read_lock();
2672                 from = rcu_dereference(rt0->from);
2673                 if (from)
2674                         rt0->dst.expires = from->expires;
2675                 rcu_read_unlock();
2676         }
2677 
2678         dst_set_expires(&rt0->dst, timeout);
2679         rt0->rt6i_flags |= RTF_EXPIRES;
2680 }
2681 
2682 static void rt6_do_update_pmtu(struct rt6_info *rt, u32 mtu)
2683 {
2684         struct net *net = dev_net(rt->dst.dev);
2685 
2686         dst_metric_set(&rt->dst, RTAX_MTU, mtu);
2687         rt->rt6i_flags |= RTF_MODIFIED;
2688         rt6_update_expires(rt, net->ipv6.sysctl.ip6_rt_mtu_expires);
2689 }
2690 
2691 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info *rt)
2692 {
2693         return !(rt->rt6i_flags & RTF_CACHE) &&
2694                 (rt->rt6i_flags & RTF_PCPU || rcu_access_pointer(rt->from));
2695 }
2696 
2697 static void __ip6_rt_update_pmtu(struct dst_entry *dst, const struct sock *sk,
2698                                  const struct ipv6hdr *iph, u32 mtu)
2699 {
2700         const struct in6_addr *daddr, *saddr;
2701         struct rt6_info *rt6 = (struct rt6_info *)dst;
2702 
2703         if (dst_metric_locked(dst, RTAX_MTU))
2704                 return;
2705 
2706         if (iph) {
2707                 daddr = &iph->daddr;
2708                 saddr = &iph->saddr;
2709         } else if (sk) {
2710                 daddr = &sk->sk_v6_daddr;
2711                 saddr = &inet6_sk(sk)->saddr;
2712         } else {
2713                 daddr = NULL;
2714                 saddr = NULL;
2715         }
2716         dst_confirm_neigh(dst, daddr);
2717         mtu = max_t(u32, mtu, IPV6_MIN_MTU);
2718         if (mtu >= dst_mtu(dst))
2719                 return;
2720 
2721         if (!rt6_cache_allowed_for_pmtu(rt6)) {
2722                 rt6_do_update_pmtu(rt6, mtu);
2723                 /* update rt6_ex->stamp for cache */
2724                 if (rt6->rt6i_flags & RTF_CACHE)
2725                         rt6_update_exception_stamp_rt(rt6);
2726         } else if (daddr) {
2727                 struct fib6_result res = {};
2728                 struct rt6_info *nrt6;
2729 
2730                 rcu_read_lock();
2731                 res.f6i = rcu_dereference(rt6->from);
2732                 if (!res.f6i) {
2733                         rcu_read_unlock();
2734                         return;
2735                 }
2736                 res.fib6_flags = res.f6i->fib6_flags;
2737                 res.fib6_type = res.f6i->fib6_type;
2738 
2739                 if (res.f6i->nh) {
2740                         struct fib6_nh_match_arg arg = {
2741                                 .dev = dst->dev,
2742                                 .gw = &rt6->rt6i_gateway,
2743                         };
2744 
2745                         nexthop_for_each_fib6_nh(res.f6i->nh,
2746                                                  fib6_nh_find_match, &arg);
2747 
2748                         /* fib6_info uses a nexthop that does not have fib6_nh
2749                          * using the dst->dev + gw. Should be impossible.
2750                          */
2751                         if (!arg.match) {
2752                                 rcu_read_unlock();
2753                                 return;
2754                         }
2755 
2756                         res.nh = arg.match;
2757                 } else {
2758                         res.nh = res.f6i->fib6_nh;
2759                 }
2760 
2761                 nrt6 = ip6_rt_cache_alloc(&res, daddr, saddr);
2762                 if (nrt6) {
2763                         rt6_do_update_pmtu(nrt6, mtu);
2764                         if (rt6_insert_exception(nrt6, &res))
2765                                 dst_release_immediate(&nrt6->dst);
2766                 }
2767                 rcu_read_unlock();
2768         }
2769 }
2770 
2771 static void ip6_rt_update_pmtu(struct dst_entry *dst, struct sock *sk,
2772                                struct sk_buff *skb, u32 mtu)
2773 {
2774         __ip6_rt_update_pmtu(dst, sk, skb ? ipv6_hdr(skb) : NULL, mtu);
2775 }
2776 
2777 void ip6_update_pmtu(struct sk_buff *skb, struct net *net, __be32 mtu,
2778                      int oif, u32 mark, kuid_t uid)
2779 {
2780         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2781         struct dst_entry *dst;
2782         struct flowi6 fl6 = {
2783                 .flowi6_oif = oif,
2784                 .flowi6_mark = mark ? mark : IP6_REPLY_MARK(net, skb->mark),
2785                 .daddr = iph->daddr,
2786                 .saddr = iph->saddr,
2787                 .flowlabel = ip6_flowinfo(iph),
2788                 .flowi6_uid = uid,
2789         };
2790 
2791         dst = ip6_route_output(net, NULL, &fl6);
2792         if (!dst->error)
2793                 __ip6_rt_update_pmtu(dst, NULL, iph, ntohl(mtu));
2794         dst_release(dst);
2795 }
2796 EXPORT_SYMBOL_GPL(ip6_update_pmtu);
2797 
2798 void ip6_sk_update_pmtu(struct sk_buff *skb, struct sock *sk, __be32 mtu)
2799 {
2800         int oif = sk->sk_bound_dev_if;
2801         struct dst_entry *dst;
2802 
2803         if (!oif && skb->dev)
2804                 oif = l3mdev_master_ifindex(skb->dev);
2805 
2806         ip6_update_pmtu(skb, sock_net(sk), mtu, oif, sk->sk_mark, sk->sk_uid);
2807 
2808         dst = __sk_dst_get(sk);
2809         if (!dst || !dst->obsolete ||
2810             dst->ops->check(dst, inet6_sk(sk)->dst_cookie))
2811                 return;
2812 
2813         bh_lock_sock(sk);
2814         if (!sock_owned_by_user(sk) && !ipv6_addr_v4mapped(&sk->sk_v6_daddr))
2815                 ip6_datagram_dst_update(sk, false);
2816         bh_unlock_sock(sk);
2817 }
2818 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu);
2819 
2820 void ip6_sk_dst_store_flow(struct sock *sk, struct dst_entry *dst,
2821                            const struct flowi6 *fl6)
2822 {
2823 #ifdef CONFIG_IPV6_SUBTREES
2824         struct ipv6_pinfo *np = inet6_sk(sk);
2825 #endif
2826 
2827         ip6_dst_store(sk, dst,
2828                       ipv6_addr_equal(&fl6->daddr, &sk->sk_v6_daddr) ?
2829                       &sk->sk_v6_daddr : NULL,
2830 #ifdef CONFIG_IPV6_SUBTREES
2831                       ipv6_addr_equal(&fl6->saddr, &np->saddr) ?
2832                       &np->saddr :
2833 #endif
2834                       NULL);
2835 }
2836 
2837 static bool ip6_redirect_nh_match(const struct fib6_result *res,
2838                                   struct flowi6 *fl6,
2839                                   const struct in6_addr *gw,
2840                                   struct rt6_info **ret)
2841 {
2842         const struct fib6_nh *nh = res->nh;
2843 
2844         if (nh->fib_nh_flags & RTNH_F_DEAD || !nh->fib_nh_gw_family ||
2845             fl6->flowi6_oif != nh->fib_nh_dev->ifindex)
2846                 return false;
2847 
2848         /* rt_cache's gateway might be different from its 'parent'
2849          * in the case of an ip redirect.
2850          * So we keep searching in the exception table if the gateway
2851          * is different.
2852          */
2853         if (!ipv6_addr_equal(gw, &nh->fib_nh_gw6)) {
2854                 struct rt6_info *rt_cache;
2855 
2856                 rt_cache = rt6_find_cached_rt(res, &fl6->daddr, &fl6->saddr);
2857                 if (rt_cache &&
2858                     ipv6_addr_equal(gw, &rt_cache->rt6i_gateway)) {
2859                         *ret = rt_cache;
2860                         return true;
2861                 }
2862                 return false;
2863         }
2864         return true;
2865 }
2866 
2867 struct fib6_nh_rd_arg {
2868         struct fib6_result      *res;
2869         struct flowi6           *fl6;
2870         const struct in6_addr   *gw;
2871         struct rt6_info         **ret;
2872 };
2873 
2874 static int fib6_nh_redirect_match(struct fib6_nh *nh, void *_arg)
2875 {
2876         struct fib6_nh_rd_arg *arg = _arg;
2877 
2878         arg->res->nh = nh;
2879         return ip6_redirect_nh_match(arg->res, arg->fl6, arg->gw, arg->ret);
2880 }
2881 
2882 /* Handle redirects */
2883 struct ip6rd_flowi {
2884         struct flowi6 fl6;
2885         struct in6_addr gateway;
2886 };
2887 
2888 static struct rt6_info *__ip6_route_redirect(struct net *net,
2889                                              struct fib6_table *table,
2890                                              struct flowi6 *fl6,
2891                                              const struct sk_buff *skb,
2892                                              int flags)
2893 {
2894         struct ip6rd_flowi *rdfl = (struct ip6rd_flowi *)fl6;
2895         struct rt6_info *ret = NULL;
2896         struct fib6_result res = {};
2897         struct fib6_nh_rd_arg arg = {
2898                 .res = &res,
2899                 .fl6 = fl6,
2900                 .gw  = &rdfl->gateway,
2901                 .ret = &ret
2902         };
2903         struct fib6_info *rt;
2904         struct fib6_node *fn;
2905 
2906         /* l3mdev_update_flow overrides oif if the device is enslaved; in
2907          * this case we must match on the real ingress device, so reset it
2908          */
2909         if (fl6->flowi6_flags & FLOWI_FLAG_SKIP_NH_OIF)
2910                 fl6->flowi6_oif = skb->dev->ifindex;
2911 
2912         /* Get the "current" route for this destination and
2913          * check if the redirect has come from appropriate router.
2914          *
2915          * RFC 4861 specifies that redirects should only be
2916          * accepted if they come from the nexthop to the target.
2917          * Due to the way the routes are chosen, this notion
2918          * is a bit fuzzy and one might need to check all possible
2919          * routes.
2920          */
2921 
2922         rcu_read_lock();
2923         fn = fib6_node_lookup(&table->tb6_root, &fl6->daddr, &fl6->saddr);
2924 restart:
2925         for_each_fib6_node_rt_rcu(fn) {
2926                 res.f6i = rt;
2927                 if (fib6_check_expired(rt))
2928                         continue;
2929                 if (rt->fib6_flags & RTF_REJECT)
2930                         break;
2931                 if (unlikely(rt->nh)) {
2932                         if (nexthop_is_blackhole(rt->nh))
2933                                 continue;
2934                         /* on match, res->nh is filled in and potentially ret */
2935                         if (nexthop_for_each_fib6_nh(rt->nh,
2936                                                      fib6_nh_redirect_match,
2937                                                      &arg))
2938                                 goto out;
2939                 } else {
2940                         res.nh = rt->fib6_nh;
2941                         if (ip6_redirect_nh_match(&res, fl6, &rdfl->gateway,
2942                                                   &ret))
2943                                 goto out;
2944                 }
2945         }
2946 
2947         if (!rt)
2948                 rt = net->ipv6.fib6_null_entry;
2949         else if (rt->fib6_flags & RTF_REJECT) {
2950                 ret = net->ipv6.ip6_null_entry;
2951                 goto out;
2952         }
2953 
2954         if (rt == net->ipv6.fib6_null_entry) {
2955                 fn = fib6_backtrack(fn, &fl6->saddr);
2956                 if (fn)
2957                         goto restart;
2958         }
2959 
2960         res.f6i = rt;
2961         res.nh = rt->fib6_nh;
2962 out:
2963         if (ret) {
2964                 ip6_hold_safe(net, &ret);
2965         } else {
2966                 res.fib6_flags = res.f6i->fib6_flags;
2967                 res.fib6_type = res.f6i->fib6_type;
2968                 ret = ip6_create_rt_rcu(&res);
2969         }
2970 
2971         rcu_read_unlock();
2972 
2973         trace_fib6_table_lookup(net, &res, table, fl6);
2974         return ret;
2975 };
2976 
2977 static struct dst_entry *ip6_route_redirect(struct net *net,
2978                                             const struct flowi6 *fl6,
2979                                             const struct sk_buff *skb,
2980                                             const struct in6_addr *gateway)
2981 {
2982         int flags = RT6_LOOKUP_F_HAS_SADDR;
2983         struct ip6rd_flowi rdfl;
2984 
2985         rdfl.fl6 = *fl6;
2986         rdfl.gateway = *gateway;
2987 
2988         return fib6_rule_lookup(net, &rdfl.fl6, skb,
2989                                 flags, __ip6_route_redirect);
2990 }
2991 
2992 void ip6_redirect(struct sk_buff *skb, struct net *net, int oif, u32 mark,
2993                   kuid_t uid)
2994 {
2995         const struct ipv6hdr *iph = (struct ipv6hdr *) skb->data;
2996         struct dst_entry *dst;
2997         struct flowi6 fl6 = {
2998                 .flowi6_iif = LOOPBACK_IFINDEX,
2999                 .flowi6_oif = oif,
3000                 .flowi6_mark = mark,
3001                 .daddr = iph->daddr,
3002                 .saddr = iph->saddr,
3003                 .flowlabel = ip6_flowinfo(iph),
3004                 .flowi6_uid = uid,
3005         };
3006 
3007         dst = ip6_route_redirect(net, &fl6, skb, &ipv6_hdr(skb)->saddr);
3008         rt6_do_redirect(dst, NULL, skb);
3009         dst_release(dst);
3010 }
3011 EXPORT_SYMBOL_GPL(ip6_redirect);
3012 
3013 void ip6_redirect_no_header(struct sk_buff *skb, struct net *net, int oif)
3014 {
3015         const struct ipv6hdr *iph = ipv6_hdr(skb);
3016         const struct rd_msg *msg = (struct rd_msg *)icmp6_hdr(skb);
3017         struct dst_entry *dst;
3018         struct flowi6 fl6 = {
3019                 .flowi6_iif = LOOPBACK_IFINDEX,
3020                 .flowi6_oif = oif,
3021                 .daddr = msg->dest,
3022                 .saddr = iph->daddr,
3023                 .flowi6_uid = sock_net_uid(net, NULL),
3024         };
3025 
3026         dst = ip6_route_redirect(net, &fl6, skb, &iph->saddr);
3027         rt6_do_redirect(dst, NULL, skb);
3028         dst_release(dst);
3029 }
3030 
3031 void ip6_sk_redirect(struct sk_buff *skb, struct sock *sk)
3032 {
3033         ip6_redirect(skb, sock_net(sk), sk->sk_bound_dev_if, sk->sk_mark,
3034                      sk->sk_uid);
3035 }
3036 EXPORT_SYMBOL_GPL(ip6_sk_redirect);
3037 
3038 static unsigned int ip6_default_advmss(const struct dst_entry *dst)
3039 {
3040         struct net_device *dev = dst->dev;
3041         unsigned int mtu = dst_mtu(dst);
3042         struct net *net = dev_net(dev);
3043 
3044         mtu -= sizeof(struct ipv6hdr) + sizeof(struct tcphdr);
3045 
3046         if (mtu < net->ipv6.sysctl.ip6_rt_min_advmss)
3047                 mtu = net->ipv6.sysctl.ip6_rt_min_advmss;
3048 
3049         /*
3050          * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
3051          * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
3052          * IPV6_MAXPLEN is also valid and means: "any MSS,
3053          * rely only on pmtu discovery"
3054          */
3055         if (mtu > IPV6_MAXPLEN - sizeof(struct tcphdr))
3056                 mtu = IPV6_MAXPLEN;
3057         return mtu;
3058 }
3059 
3060 static unsigned int ip6_mtu(const struct dst_entry *dst)
3061 {
3062         struct inet6_dev *idev;
3063         unsigned int mtu;
3064 
3065         mtu = dst_metric_raw(dst, RTAX_MTU);
3066         if (mtu)
3067                 goto out;
3068 
3069         mtu = IPV6_MIN_MTU;
3070 
3071         rcu_read_lock();
3072         idev = __in6_dev_get(dst->dev);
3073         if (idev)
3074                 mtu = idev->cnf.mtu6;
3075         rcu_read_unlock();
3076 
3077 out:
3078         mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3079 
3080         return mtu - lwtunnel_headroom(dst->lwtstate, mtu);
3081 }
3082 
3083 /* MTU selection:
3084  * 1. mtu on route is locked - use it
3085  * 2. mtu from nexthop exception
3086  * 3. mtu from egress device
3087  *
3088  * based on ip6_dst_mtu_forward and exception logic of
3089  * rt6_find_cached_rt; called with rcu_read_lock
3090  */
3091 u32 ip6_mtu_from_fib6(const struct fib6_result *res,
3092                       const struct in6_addr *daddr,
3093                       const struct in6_addr *saddr)
3094 {
3095         const struct fib6_nh *nh = res->nh;
3096         struct fib6_info *f6i = res->f6i;
3097         struct inet6_dev *idev;
3098         struct rt6_info *rt;
3099         u32 mtu = 0;
3100 
3101         if (unlikely(fib6_metric_locked(f6i, RTAX_MTU))) {
3102                 mtu = f6i->fib6_pmtu;
3103                 if (mtu)
3104                         goto out;
3105         }
3106 
3107         rt = rt6_find_cached_rt(res, daddr, saddr);
3108         if (unlikely(rt)) {
3109                 mtu = dst_metric_raw(&rt->dst, RTAX_MTU);
3110         } else {
3111                 struct net_device *dev = nh->fib_nh_dev;
3112 
3113                 mtu = IPV6_MIN_MTU;
3114                 idev = __in6_dev_get(dev);
3115                 if (idev && idev->cnf.mtu6 > mtu)
3116                         mtu = idev->cnf.mtu6;
3117         }
3118 
3119         mtu = min_t(unsigned int, mtu, IP6_MAX_MTU);
3120 out:
3121         return mtu - lwtunnel_headroom(nh->fib_nh_lws, mtu);
3122 }
3123 
3124 struct dst_entry *icmp6_dst_alloc(struct net_device *dev,
3125                                   struct flowi6 *fl6)
3126 {
3127         struct dst_entry *dst;
3128         struct rt6_info *rt;
3129         struct inet6_dev *idev = in6_dev_get(dev);
3130         struct net *net = dev_net(dev);
3131 
3132         if (unlikely(!idev))
3133                 return ERR_PTR(-ENODEV);
3134 
3135         rt = ip6_dst_alloc(net, dev, 0);
3136         if (unlikely(!rt)) {
3137                 in6_dev_put(idev);
3138                 dst = ERR_PTR(-ENOMEM);
3139                 goto out;
3140         }
3141 
3142         rt->dst.flags |= DST_HOST;
3143         rt->dst.input = ip6_input;
3144         rt->dst.output  = ip6_output;
3145         rt->rt6i_gateway  = fl6->daddr;
3146         rt->rt6i_dst.addr = fl6->daddr;
3147         rt->rt6i_dst.plen = 128;
3148         rt->rt6i_idev     = idev;
3149         dst_metric_set(&rt->dst, RTAX_HOPLIMIT, 0);
3150 
3151         /* Add this dst into uncached_list so that rt6_disable_ip() can
3152          * do proper release of the net_device
3153          */
3154         rt6_uncached_list_add(rt);
3155         atomic_inc(&net->ipv6.rt6_stats->fib_rt_uncache);
3156 
3157         dst = xfrm_lookup(net, &rt->dst, flowi6_to_flowi(fl6), NULL, 0);
3158 
3159 out:
3160         return dst;
3161 }
3162 
3163 static int ip6_dst_gc(struct dst_ops *ops)
3164 {
3165         struct net *net = container_of(ops, struct net, ipv6.ip6_dst_ops);
3166         int rt_min_interval = net->ipv6.sysctl.ip6_rt_gc_min_interval;
3167         int rt_max_size = net->ipv6.sysctl.ip6_rt_max_size;
3168         int rt_elasticity = net->ipv6.sysctl.ip6_rt_gc_elasticity;
3169         int rt_gc_timeout = net->ipv6.sysctl.ip6_rt_gc_timeout;
3170         unsigned long rt_last_gc = net->ipv6.ip6_rt_last_gc;
3171         int entries;
3172 
3173         entries = dst_entries_get_fast(ops);
3174         if (time_after(rt_last_gc + rt_min_interval, jiffies) &&
3175             entries <= rt_max_size)
3176                 goto out;
3177 
3178         net->ipv6.ip6_rt_gc_expire++;
3179         fib6_run_gc(net->ipv6.ip6_rt_gc_expire, net, true);
3180         entries = dst_entries_get_slow(ops);
3181         if (entries < ops->gc_thresh)
3182                 net->ipv6.ip6_rt_gc_expire = rt_gc_timeout>>1;
3183 out:
3184         net->ipv6.ip6_rt_gc_expire -= net->ipv6.ip6_rt_gc_expire>>rt_elasticity;
3185         return entries > rt_max_size;
3186 }
3187 
3188 static int ip6_nh_lookup_table(struct net *net, struct fib6_config *cfg,
3189                                const struct in6_addr *gw_addr, u32 tbid,
3190                                int flags, struct fib6_result *res)
3191 {
3192         struct flowi6 fl6 = {
3193                 .flowi6_oif = cfg->fc_ifindex,
3194                 .daddr = *gw_addr,
3195                 .saddr = cfg->fc_prefsrc,
3196         };
3197         struct fib6_table *table;
3198         int err;
3199 
3200         table = fib6_get_table(net, tbid);
3201         if (!table)
3202                 return -EINVAL;
3203 
3204         if (!ipv6_addr_any(&cfg->fc_prefsrc))
3205                 flags |= RT6_LOOKUP_F_HAS_SADDR;
3206 
3207         flags |= RT6_LOOKUP_F_IGNORE_LINKSTATE;
3208 
3209         err = fib6_table_lookup(net, table, cfg->fc_ifindex, &fl6, res, flags);
3210         if (!err && res->f6i != net->ipv6.fib6_null_entry)
3211                 fib6_select_path(net, res, &fl6, cfg->fc_ifindex,
3212                                  cfg->fc_ifindex != 0, NULL, flags);
3213 
3214         return err;
3215 }
3216 
3217 static int ip6_route_check_nh_onlink(struct net *net,
3218                                      struct fib6_config *cfg,
3219                                      const struct net_device *dev,
3220                                      struct netlink_ext_ack *extack)
3221 {
3222         u32 tbid = l3mdev_fib_table_rcu(dev) ? : RT_TABLE_MAIN;
3223         const struct in6_addr *gw_addr = &cfg->fc_gateway;
3224         struct fib6_result res = {};
3225         int err;
3226 
3227         err = ip6_nh_lookup_table(net, cfg, gw_addr, tbid, 0, &res);
3228         if (!err && !(res.fib6_flags & RTF_REJECT) &&
3229             /* ignore match if it is the default route */
3230             !ipv6_addr_any(&res.f6i->fib6_dst.addr) &&
3231             (res.fib6_type != RTN_UNICAST || dev != res.nh->fib_nh_dev)) {
3232                 NL_SET_ERR_MSG(extack,
3233                                "Nexthop has invalid gateway or device mismatch");
3234                 err = -EINVAL;
3235         }
3236 
3237         return err;
3238 }
3239 
3240 static int ip6_route_check_nh(struct net *net,
3241                               struct fib6_config *cfg,
3242                               struct net_device **_dev,
3243                               struct inet6_dev **idev)
3244 {
3245         const struct in6_addr *gw_addr = &cfg->fc_gateway;
3246         struct net_device *dev = _dev ? *_dev : NULL;
3247         int flags = RT6_LOOKUP_F_IFACE;
3248         struct fib6_result res = {};
3249         int err = -EHOSTUNREACH;
3250 
3251         if (cfg->fc_table) {
3252                 err = ip6_nh_lookup_table(net, cfg, gw_addr,
3253                                           cfg->fc_table, flags, &res);
3254                 /* gw_addr can not require a gateway or resolve to a reject
3255                  * route. If a device is given, it must match the result.
3256                  */
3257                 if (err || res.fib6_flags & RTF_REJECT ||
3258                     res.nh->fib_nh_gw_family ||
3259                     (dev && dev != res.nh->fib_nh_dev))
3260                         err = -EHOSTUNREACH;
3261         }
3262 
3263         if (err < 0) {
3264                 struct flowi6 fl6 = {
3265                         .flowi6_oif = cfg->fc_ifindex,
3266                         .daddr = *gw_addr,
3267                 };
3268 
3269                 err = fib6_lookup(net, cfg->fc_ifindex, &fl6, &res, flags);
3270                 if (err || res.fib6_flags & RTF_REJECT ||
3271                     res.nh->fib_nh_gw_family)
3272                         err = -EHOSTUNREACH;
3273 
3274                 if (err)
3275                         return err;
3276 
3277                 fib6_select_path(net, &res, &fl6, cfg->fc_ifindex,
3278                                  cfg->fc_ifindex != 0, NULL, flags);
3279         }
3280 
3281         err = 0;
3282         if (dev) {
3283                 if (dev != res.nh->fib_nh_dev)
3284                         err = -EHOSTUNREACH;
3285         } else {
3286                 *_dev = dev = res.nh->fib_nh_dev;
3287                 dev_hold(dev);
3288                 *idev = in6_dev_get(dev);
3289         }
3290 
3291         return err;
3292 }
3293 
3294 static int ip6_validate_gw(struct net *net, struct fib6_config *cfg,
3295                            struct net_device **_dev, struct inet6_dev **idev,
3296                            struct netlink_ext_ack *extack)
3297 {
3298         const struct in6_addr *gw_addr = &cfg->fc_gateway;
3299         int gwa_type = ipv6_addr_type(gw_addr);
3300         bool skip_dev = gwa_type & IPV6_ADDR_LINKLOCAL ? false : true;
3301         const struct net_device *dev = *_dev;
3302         bool need_addr_check = !dev;
3303         int err = -EINVAL;
3304 
3305         /* if gw_addr is local we will fail to detect this in case
3306          * address is still TENTATIVE (DAD in progress). rt6_lookup()
3307          * will return already-added prefix route via interface that
3308          * prefix route was assigned to, which might be non-loopback.
3309          */
3310         if (dev &&
3311             ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3312                 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3313                 goto out;
3314         }
3315 
3316         if (gwa_type != (IPV6_ADDR_LINKLOCAL | IPV6_ADDR_UNICAST)) {
3317                 /* IPv6 strictly inhibits using not link-local
3318                  * addresses as nexthop address.
3319                  * Otherwise, router will not able to send redirects.
3320                  * It is very good, but in some (rare!) circumstances
3321                  * (SIT, PtP, NBMA NOARP links) it is handy to allow
3322                  * some exceptions. --ANK
3323                  * We allow IPv4-mapped nexthops to support RFC4798-type
3324                  * addressing
3325                  */
3326                 if (!(gwa_type & (IPV6_ADDR_UNICAST | IPV6_ADDR_MAPPED))) {
3327                         NL_SET_ERR_MSG(extack, "Invalid gateway address");
3328                         goto out;
3329                 }
3330 
3331                 rcu_read_lock();
3332 
3333                 if (cfg->fc_flags & RTNH_F_ONLINK)
3334                         err = ip6_route_check_nh_onlink(net, cfg, dev, extack);
3335                 else
3336                         err = ip6_route_check_nh(net, cfg, _dev, idev);
3337 
3338                 rcu_read_unlock();
3339 
3340                 if (err)
3341                         goto out;
3342         }
3343 
3344         /* reload in case device was changed */
3345         dev = *_dev;
3346 
3347         err = -EINVAL;
3348         if (!dev) {
3349                 NL_SET_ERR_MSG(extack, "Egress device not specified");
3350                 goto out;
3351         } else if (dev->flags & IFF_LOOPBACK) {
3352                 NL_SET_ERR_MSG(extack,
3353                                "Egress device can not be loopback device for this route");
3354                 goto out;
3355         }
3356 
3357         /* if we did not check gw_addr above, do so now that the
3358          * egress device has been resolved.
3359          */
3360         if (need_addr_check &&
3361             ipv6_chk_addr_and_flags(net, gw_addr, dev, skip_dev, 0, 0)) {
3362                 NL_SET_ERR_MSG(extack, "Gateway can not be a local address");
3363                 goto out;
3364         }
3365 
3366         err = 0;
3367 out:
3368         return err;
3369 }
3370 
3371 static bool fib6_is_reject(u32 flags, struct net_device *dev, int addr_type)
3372 {
3373         if ((flags & RTF_REJECT) ||
3374             (dev && (dev->flags & IFF_LOOPBACK) &&
3375              !(addr_type & IPV6_ADDR_LOOPBACK) &&
3376              !(flags & RTF_LOCAL)))
3377                 return true;
3378 
3379         return false;
3380 }
3381 
3382 int fib6_nh_init(struct net *net, struct fib6_nh *fib6_nh,
3383                  struct fib6_config *cfg, gfp_t gfp_flags,
3384                  struct netlink_ext_ack *extack)
3385 {
3386         struct net_device *dev = NULL;
3387         struct inet6_dev *idev = NULL;
3388         int addr_type;
3389         int err;
3390 
3391         fib6_nh->fib_nh_family = AF_INET6;
3392 #ifdef CONFIG_IPV6_ROUTER_PREF
3393         fib6_nh->last_probe = jiffies;
3394 #endif
3395 
3396         err = -ENODEV;
3397         if (cfg->fc_ifindex) {
3398                 dev = dev_get_by_index(net, cfg->fc_ifindex);
3399                 if (!dev)
3400                         goto out;
3401                 idev = in6_dev_get(dev);
3402                 if (!idev)
3403                         goto out;
3404         }
3405 
3406         if (cfg->fc_flags & RTNH_F_ONLINK) {
3407                 if (!dev) {
3408                         NL_SET_ERR_MSG(extack,
3409                                        "Nexthop device required for onlink");
3410                         goto out;
3411                 }
3412 
3413                 if (!(dev->flags & IFF_UP)) {
3414                         NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3415                         err = -ENETDOWN;
3416                         goto out;
3417                 }
3418 
3419                 fib6_nh->fib_nh_flags |= RTNH_F_ONLINK;
3420         }
3421 
3422         fib6_nh->fib_nh_weight = 1;
3423 
3424         /* We cannot add true routes via loopback here,
3425          * they would result in kernel looping; promote them to reject routes
3426          */
3427         addr_type = ipv6_addr_type(&cfg->fc_dst);
3428         if (fib6_is_reject(cfg->fc_flags, dev, addr_type)) {
3429                 /* hold loopback dev/idev if we haven't done so. */
3430                 if (dev != net->loopback_dev) {
3431                         if (dev) {
3432                                 dev_put(dev);
3433                                 in6_dev_put(idev);
3434                         }
3435                         dev = net->loopback_dev;
3436                         dev_hold(dev);
3437                         idev = in6_dev_get(dev);
3438                         if (!idev) {
3439                                 err = -ENODEV;
3440                                 goto out;
3441                         }
3442                 }
3443                 goto pcpu_alloc;
3444         }
3445 
3446         if (cfg->fc_flags & RTF_GATEWAY) {
3447                 err = ip6_validate_gw(net, cfg, &dev, &idev, extack);
3448                 if (err)
3449                         goto out;
3450 
3451                 fib6_nh->fib_nh_gw6 = cfg->fc_gateway;
3452                 fib6_nh->fib_nh_gw_family = AF_INET6;
3453         }
3454 
3455         err = -ENODEV;
3456         if (!dev)
3457                 goto out;
3458 
3459         if (idev->cnf.disable_ipv6) {
3460                 NL_SET_ERR_MSG(extack, "IPv6 is disabled on nexthop device");
3461                 err = -EACCES;
3462                 goto out;
3463         }
3464 
3465         if (!(dev->flags & IFF_UP) && !cfg->fc_ignore_dev_down) {
3466                 NL_SET_ERR_MSG(extack, "Nexthop device is not up");
3467                 err = -ENETDOWN;
3468                 goto out;
3469         }
3470 
3471         if (!(cfg->fc_flags & (RTF_LOCAL | RTF_ANYCAST)) &&
3472             !netif_carrier_ok(dev))
3473                 fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
3474 
3475         err = fib_nh_common_init(&fib6_nh->nh_common, cfg->fc_encap,
3476                                  cfg->fc_encap_type, cfg, gfp_flags, extack);
3477         if (err)
3478                 goto out;
3479 
3480 pcpu_alloc:
3481         fib6_nh->rt6i_pcpu = alloc_percpu_gfp(struct rt6_info *, gfp_flags);
3482         if (!fib6_nh->rt6i_pcpu) {
3483                 err = -ENOMEM;
3484                 goto out;
3485         }
3486 
3487         fib6_nh->fib_nh_dev = dev;
3488         fib6_nh->fib_nh_oif = dev->ifindex;
3489         err = 0;
3490 out:
3491         if (idev)
3492                 in6_dev_put(idev);
3493 
3494         if (err) {
3495                 lwtstate_put(fib6_nh->fib_nh_lws);
3496                 fib6_nh->fib_nh_lws = NULL;
3497                 if (dev)
3498                         dev_put(dev);
3499         }
3500 
3501         return err;
3502 }
3503 
3504 void fib6_nh_release(struct fib6_nh *fib6_nh)
3505 {
3506         struct rt6_exception_bucket *bucket;
3507 
3508         rcu_read_lock();
3509 
3510         fib6_nh_flush_exceptions(fib6_nh, NULL);
3511         bucket = fib6_nh_get_excptn_bucket(fib6_nh, NULL);
3512         if (bucket) {
3513                 rcu_assign_pointer(fib6_nh->rt6i_exception_bucket, NULL);
3514                 kfree(bucket);
3515         }
3516 
3517         rcu_read_unlock();
3518 
3519         if (fib6_nh->rt6i_pcpu) {
3520                 int cpu;
3521 
3522                 for_each_possible_cpu(cpu) {
3523                         struct rt6_info **ppcpu_rt;
3524                         struct rt6_info *pcpu_rt;
3525 
3526                         ppcpu_rt = per_cpu_ptr(fib6_nh->rt6i_pcpu, cpu);
3527                         pcpu_rt = *ppcpu_rt;
3528                         if (pcpu_rt) {
3529                                 dst_dev_put(&pcpu_rt->dst);
3530                                 dst_release(&pcpu_rt->dst);
3531                                 *ppcpu_rt = NULL;
3532                         }
3533                 }
3534 
3535                 free_percpu(fib6_nh->rt6i_pcpu);
3536         }
3537 
3538         fib_nh_common_release(&fib6_nh->nh_common);
3539 }
3540 
3541 static struct fib6_info *ip6_route_info_create(struct fib6_config *cfg,
3542                                               gfp_t gfp_flags,
3543                                               struct netlink_ext_ack *extack)
3544 {
3545         struct net *net = cfg->fc_nlinfo.nl_net;
3546         struct fib6_info *rt = NULL;
3547         struct nexthop *nh = NULL;
3548         struct fib6_table *table;
3549         struct fib6_nh *fib6_nh;
3550         int err = -EINVAL;
3551         int addr_type;
3552 
3553         /* RTF_PCPU is an internal flag; can not be set by userspace */
3554         if (cfg->fc_flags & RTF_PCPU) {
3555                 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_PCPU");
3556                 goto out;
3557         }
3558 
3559         /* RTF_CACHE is an internal flag; can not be set by userspace */
3560         if (cfg->fc_flags & RTF_CACHE) {
3561                 NL_SET_ERR_MSG(extack, "Userspace can not set RTF_CACHE");
3562                 goto out;
3563         }
3564 
3565         if (cfg->fc_type > RTN_MAX) {
3566                 NL_SET_ERR_MSG(extack, "Invalid route type");
3567                 goto out;
3568         }
3569 
3570         if (cfg->fc_dst_len > 128) {
3571                 NL_SET_ERR_MSG(extack, "Invalid prefix length");
3572                 goto out;
3573         }
3574         if (cfg->fc_src_len > 128) {
3575                 NL_SET_ERR_MSG(extack, "Invalid source address length");
3576                 goto out;
3577         }
3578 #ifndef CONFIG_IPV6_SUBTREES
3579         if (cfg->fc_src_len) {
3580                 NL_SET_ERR_MSG(extack,
3581                                "Specifying source address requires IPV6_SUBTREES to be enabled");
3582                 goto out;
3583         }
3584 #endif
3585         if (cfg->fc_nh_id) {
3586                 nh = nexthop_find_by_id(net, cfg->fc_nh_id);
3587                 if (!nh) {
3588                         NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
3589                         goto out;
3590                 }
3591                 err = fib6_check_nexthop(nh, cfg, extack);
3592                 if (err)
3593                         goto out;
3594         }
3595 
3596         err = -ENOBUFS;
3597         if (cfg->fc_nlinfo.nlh &&
3598             !(cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_CREATE)) {
3599                 table = fib6_get_table(net, cfg->fc_table);
3600                 if (!table) {
3601                         pr_warn("NLM_F_CREATE should be specified when creating new route\n");
3602                         table = fib6_new_table(net, cfg->fc_table);
3603                 }
3604         } else {
3605                 table = fib6_new_table(net, cfg->fc_table);
3606         }
3607 
3608         if (!table)
3609                 goto out;
3610 
3611         err = -ENOMEM;
3612         rt = fib6_info_alloc(gfp_flags, !nh);
3613         if (!rt)
3614                 goto out;
3615 
3616         rt->fib6_metrics = ip_fib_metrics_init(net, cfg->fc_mx, cfg->fc_mx_len,
3617                                                extack);
3618         if (IS_ERR(rt->fib6_metrics)) {
3619                 err = PTR_ERR(rt->fib6_metrics);
3620                 /* Do not leave garbage there. */
3621                 rt->fib6_metrics = (struct dst_metrics *)&dst_default_metrics;
3622                 goto out;
3623         }
3624 
3625         if (cfg->fc_flags & RTF_ADDRCONF)
3626                 rt->dst_nocount = true;
3627 
3628         if (cfg->fc_flags & RTF_EXPIRES)
3629                 fib6_set_expires(rt, jiffies +
3630                                 clock_t_to_jiffies(cfg->fc_expires));
3631         else
3632                 fib6_clean_expires(rt);
3633 
3634         if (cfg->fc_protocol == RTPROT_UNSPEC)
3635                 cfg->fc_protocol = RTPROT_BOOT;
3636         rt->fib6_protocol = cfg->fc_protocol;
3637 
3638         rt->fib6_table = table;
3639         rt->fib6_metric = cfg->fc_metric;
3640         rt->fib6_type = cfg->fc_type ? : RTN_UNICAST;
3641         rt->fib6_flags = cfg->fc_flags & ~RTF_GATEWAY;
3642 
3643         ipv6_addr_prefix(&rt->fib6_dst.addr, &cfg->fc_dst, cfg->fc_dst_len);
3644         rt->fib6_dst.plen = cfg->fc_dst_len;
3645         if (rt->fib6_dst.plen == 128)
3646                 rt->dst_host = true;
3647 
3648 #ifdef CONFIG_IPV6_SUBTREES
3649         ipv6_addr_prefix(&rt->fib6_src.addr, &cfg->fc_src, cfg->fc_src_len);
3650         rt->fib6_src.plen = cfg->fc_src_len;
3651 #endif
3652         if (nh) {
3653                 if (!nexthop_get(nh)) {
3654                         NL_SET_ERR_MSG(extack, "Nexthop has been deleted");
3655                         goto out;
3656                 }
3657                 if (rt->fib6_src.plen) {
3658                         NL_SET_ERR_MSG(extack, "Nexthops can not be used with source routing");
3659                         goto out;
3660                 }
3661                 rt->nh = nh;
3662                 fib6_nh = nexthop_fib6_nh(rt->nh);
3663         } else {
3664                 err = fib6_nh_init(net, rt->fib6_nh, cfg, gfp_flags, extack);
3665                 if (err)
3666                         goto out;
3667 
3668                 fib6_nh = rt->fib6_nh;
3669 
3670                 /* We cannot add true routes via loopback here, they would
3671                  * result in kernel looping; promote them to reject routes
3672                  */
3673                 addr_type = ipv6_addr_type(&cfg->fc_dst);
3674                 if (fib6_is_reject(cfg->fc_flags, rt->fib6_nh->fib_nh_dev,
3675                                    addr_type))
3676                         rt->fib6_flags = RTF_REJECT | RTF_NONEXTHOP;
3677         }
3678 
3679         if (!ipv6_addr_any(&cfg->fc_prefsrc)) {
3680                 struct net_device *dev = fib6_nh->fib_nh_dev;
3681 
3682                 if (!ipv6_chk_addr(net, &cfg->fc_prefsrc, dev, 0)) {
3683                         NL_SET_ERR_MSG(extack, "Invalid source address");
3684                         err = -EINVAL;
3685                         goto out;
3686                 }
3687                 rt->fib6_prefsrc.addr = cfg->fc_prefsrc;
3688                 rt->fib6_prefsrc.plen = 128;
3689         } else
3690                 rt->fib6_prefsrc.plen = 0;
3691 
3692         return rt;
3693 out:
3694         fib6_info_release(rt);
3695         return ERR_PTR(err);
3696 }
3697 
3698 int ip6_route_add(struct fib6_config *cfg, gfp_t gfp_flags,
3699                   struct netlink_ext_ack *extack)
3700 {
3701         struct fib6_info *rt;
3702         int err;
3703 
3704         rt = ip6_route_info_create(cfg, gfp_flags, extack);
3705         if (IS_ERR(rt))
3706                 return PTR_ERR(rt);
3707 
3708         err = __ip6_ins_rt(rt, &cfg->fc_nlinfo, extack);
3709         fib6_info_release(rt);
3710 
3711         return err;
3712 }
3713 
3714 static int __ip6_del_rt(struct fib6_info *rt, struct nl_info *info)
3715 {
3716         struct net *net = info->nl_net;
3717         struct fib6_table *table;
3718         int err;
3719 
3720         if (rt == net->ipv6.fib6_null_entry) {
3721                 err = -ENOENT;
3722                 goto out;
3723         }
3724 
3725         table = rt->fib6_table;
3726         spin_lock_bh(&table->tb6_lock);
3727         err = fib6_del(rt, info);
3728         spin_unlock_bh(&table->tb6_lock);
3729 
3730 out:
3731         fib6_info_release(rt);
3732         return err;
3733 }
3734 
3735 int ip6_del_rt(struct net *net, struct fib6_info *rt)
3736 {
3737         struct nl_info info = { .nl_net = net };
3738 
3739         return __ip6_del_rt(rt, &info);
3740 }
3741 
3742 static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
3743 {
3744         struct nl_info *info = &cfg->fc_nlinfo;
3745         struct net *net = info->nl_net;
3746         struct sk_buff *skb = NULL;
3747         struct fib6_table *table;
3748         int err = -ENOENT;
3749 
3750         if (rt == net->ipv6.fib6_null_entry)
3751                 goto out_put;
3752         table = rt->fib6_table;
3753         spin_lock_bh(&table->tb6_lock);
3754 
3755         if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
3756                 struct fib6_info *sibling, *next_sibling;
3757 
3758                 /* prefer to send a single notification with all hops */
3759                 skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
3760                 if (skb) {
3761                         u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
3762 
3763                         if (rt6_fill_node(net, skb, rt, NULL,
3764                                           NULL, NULL, 0, RTM_DELROUTE,
3765                                           info->portid, seq, 0) < 0) {
3766                                 kfree_skb(skb);
3767                                 skb = NULL;
3768                         } else
3769                                 info->skip_notify = 1;
3770                 }
3771 
3772                 info->skip_notify_kernel = 1;
3773                 call_fib6_multipath_entry_notifiers(net,
3774                                                     FIB_EVENT_ENTRY_DEL,
3775                                                     rt,
3776                                                     rt->fib6_nsiblings,
3777                                                     NULL);
3778                 list_for_each_entry_safe(sibling, next_sibling,
3779                                          &rt->fib6_siblings,
3780                                          fib6_siblings) {
3781                         err = fib6_del(sibling, info);
3782                         if (err)
3783                                 goto out_unlock;
3784                 }
3785         }
3786 
3787         err = fib6_del(rt, info);
3788 out_unlock:
3789         spin_unlock_bh(&table->tb6_lock);
3790 out_put:
3791         fib6_info_release(rt);
3792 
3793         if (skb) {
3794                 rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
3795                             info->nlh, gfp_any());
3796         }
3797         return err;
3798 }
3799 
3800 static int __ip6_del_cached_rt(struct rt6_info *rt, struct fib6_config *cfg)
3801 {
3802         int rc = -ESRCH;
3803 
3804         if (cfg->fc_ifindex && rt->dst.dev->ifindex != cfg->fc_ifindex)
3805                 goto out;
3806 
3807         if (cfg->fc_flags & RTF_GATEWAY &&
3808             !ipv6_addr_equal(&cfg->fc_gateway, &rt->rt6i_gateway))
3809                 goto out;
3810 
3811         rc = rt6_remove_exception_rt(rt);
3812 out:
3813         return rc;
3814 }
3815 
3816 static int ip6_del_cached_rt(struct fib6_config *cfg, struct fib6_info *rt,
3817                              struct fib6_nh *nh)
3818 {
3819         struct fib6_result res = {
3820                 .f6i = rt,
3821                 .nh = nh,
3822         };
3823         struct rt6_info *rt_cache;
3824 
3825         rt_cache = rt6_find_cached_rt(&res, &cfg->fc_dst, &cfg->fc_src);
3826         if (rt_cache)
3827                 return __ip6_del_cached_rt(rt_cache, cfg);
3828 
3829         return 0;
3830 }
3831 
3832 struct fib6_nh_del_cached_rt_arg {
3833         struct fib6_config *cfg;
3834         struct fib6_info *f6i;
3835 };
3836 
3837 static int fib6_nh_del_cached_rt(struct fib6_nh *nh, void *_arg)
3838 {
3839         struct fib6_nh_del_cached_rt_arg *arg = _arg;
3840         int rc;
3841 
3842         rc = ip6_del_cached_rt(arg->cfg, arg->f6i, nh);
3843         return rc != -ESRCH ? rc : 0;
3844 }
3845 
3846 static int ip6_del_cached_rt_nh(struct fib6_config *cfg, struct fib6_info *f6i)
3847 {
3848         struct fib6_nh_del_cached_rt_arg arg = {
3849                 .cfg = cfg,
3850                 .f6i = f6i
3851         };
3852 
3853         return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_del_cached_rt, &arg);
3854 }
3855 
3856 static int ip6_route_del(struct fib6_config *cfg,
3857                          struct netlink_ext_ack *extack)
3858 {
3859         struct fib6_table *table;
3860         struct fib6_info *rt;
3861         struct fib6_node *fn;
3862         int err = -ESRCH;
3863 
3864         table = fib6_get_table(cfg->fc_nlinfo.nl_net, cfg->fc_table);
3865         if (!table) {
3866                 NL_SET_ERR_MSG(extack, "FIB table does not exist");
3867                 return err;
3868         }
3869 
3870         rcu_read_lock();
3871 
3872         fn = fib6_locate(&table->tb6_root,
3873                          &cfg->fc_dst, cfg->fc_dst_len,
3874                          &cfg->fc_src, cfg->fc_src_len,
3875                          !(cfg->fc_flags & RTF_CACHE));
3876 
3877         if (fn) {
3878                 for_each_fib6_node_rt_rcu(fn) {
3879                         struct fib6_nh *nh;
3880 
3881                         if (rt->nh && cfg->fc_nh_id &&
3882                             rt->nh->id != cfg->fc_nh_id)
3883                                 continue;
3884 
3885                         if (cfg->fc_flags & RTF_CACHE) {
3886                                 int rc = 0;
3887 
3888                                 if (rt->nh) {
3889                                         rc = ip6_del_cached_rt_nh(cfg, rt);
3890                                 } else if (cfg->fc_nh_id) {
3891                                         continue;
3892                                 } else {
3893                                         nh = rt->fib6_nh;
3894                                         rc = ip6_del_cached_rt(cfg, rt, nh);
3895                                 }
3896                                 if (rc != -ESRCH) {
3897                                         rcu_read_unlock();
3898                                         return rc;
3899                                 }
3900                                 continue;
3901                         }
3902 
3903                         if (cfg->fc_metric && cfg->fc_metric != rt->fib6_metric)
3904                                 continue;
3905                         if (cfg->fc_protocol &&
3906                             cfg->fc_protocol != rt->fib6_protocol)
3907                                 continue;
3908 
3909                         if (rt->nh) {
3910                                 if (!fib6_info_hold_safe(rt))
3911                                         continue;
3912                                 rcu_read_unlock();
3913 
3914                                 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3915                         }
3916                         if (cfg->fc_nh_id)
3917                                 continue;
3918 
3919                         nh = rt->fib6_nh;
3920                         if (cfg->fc_ifindex &&
3921                             (!nh->fib_nh_dev ||
3922                              nh->fib_nh_dev->ifindex != cfg->fc_ifindex))
3923                                 continue;
3924                         if (cfg->fc_flags & RTF_GATEWAY &&
3925                             !ipv6_addr_equal(&cfg->fc_gateway, &nh->fib_nh_gw6))
3926                                 continue;
3927                         if (!fib6_info_hold_safe(rt))
3928                                 continue;
3929                         rcu_read_unlock();
3930 
3931                         /* if gateway was specified only delete the one hop */
3932                         if (cfg->fc_flags & RTF_GATEWAY)
3933                                 return __ip6_del_rt(rt, &cfg->fc_nlinfo);
3934 
3935                         return __ip6_del_rt_siblings(rt, cfg);
3936                 }
3937         }
3938         rcu_read_unlock();
3939 
3940         return err;
3941 }
3942 
3943 static void rt6_do_redirect(struct dst_entry *dst, struct sock *sk, struct sk_buff *skb)
3944 {
3945         struct netevent_redirect netevent;
3946         struct rt6_info *rt, *nrt = NULL;
3947         struct fib6_result res = {};
3948         struct ndisc_options ndopts;
3949         struct inet6_dev *in6_dev;
3950         struct neighbour *neigh;
3951         struct rd_msg *msg;
3952         int optlen, on_link;
3953         u8 *lladdr;
3954 
3955         optlen = skb_tail_pointer(skb) - skb_transport_header(skb);
3956         optlen -= sizeof(*msg);
3957 
3958         if (optlen < 0) {
3959                 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
3960                 return;
3961         }
3962 
3963         msg = (struct rd_msg *)icmp6_hdr(skb);
3964 
3965         if (ipv6_addr_is_multicast(&msg->dest)) {
3966                 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
3967                 return;
3968         }
3969 
3970         on_link = 0;
3971         if (ipv6_addr_equal(&msg->dest, &msg->target)) {
3972                 on_link = 1;
3973         } else if (ipv6_addr_type(&msg->target) !=
3974                    (IPV6_ADDR_UNICAST|IPV6_ADDR_LINKLOCAL)) {
3975                 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
3976                 return;
3977         }
3978 
3979         in6_dev = __in6_dev_get(skb->dev);
3980         if (!in6_dev)
3981                 return;
3982         if (in6_dev->cnf.forwarding || !in6_dev->cnf.accept_redirects)
3983                 return;
3984 
3985         /* RFC2461 8.1:
3986          *      The IP source address of the Redirect MUST be the same as the current
3987          *      first-hop router for the specified ICMP Destination Address.
3988          */
3989 
3990         if (!ndisc_parse_options(skb->dev, msg->opt, optlen, &ndopts)) {
3991                 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3992                 return;
3993         }
3994 
3995         lladdr = NULL;
3996         if (ndopts.nd_opts_tgt_lladdr) {
3997                 lladdr = ndisc_opt_addr_data(ndopts.nd_opts_tgt_lladdr,
3998                                              skb->dev);
3999                 if (!lladdr) {
4000                         net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
4001                         return;
4002                 }
4003         }
4004 
4005         rt = (struct rt6_info *) dst;
4006         if (rt->rt6i_flags & RTF_REJECT) {
4007                 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
4008                 return;
4009         }
4010 
4011         /* Redirect received -> path was valid.
4012          * Look, redirects are sent only in response to data packets,
4013          * so that this nexthop apparently is reachable. --ANK
4014          */
4015         dst_confirm_neigh(&rt->dst, &ipv6_hdr(skb)->saddr);
4016 
4017         neigh = __neigh_lookup(&nd_tbl, &msg->target, skb->dev, 1);
4018         if (!neigh)
4019                 return;
4020 
4021         /*
4022          *      We have finally decided to accept it.
4023          */
4024 
4025         ndisc_update(skb->dev, neigh, lladdr, NUD_STALE,
4026                      NEIGH_UPDATE_F_WEAK_OVERRIDE|
4027                      NEIGH_UPDATE_F_OVERRIDE|
4028                      (on_link ? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER|
4029                                      NEIGH_UPDATE_F_ISROUTER)),
4030                      NDISC_REDIRECT, &ndopts);
4031 
4032         rcu_read_lock();
4033         res.f6i = rcu_dereference(rt->from);
4034         if (!res.f6i)
4035                 goto out;
4036 
4037         if (res.f6i->nh) {
4038                 struct fib6_nh_match_arg arg = {
4039                         .dev = dst->dev,
4040                         .gw = &rt->rt6i_gateway,
4041                 };
4042 
4043                 nexthop_for_each_fib6_nh(res.f6i->nh,
4044                                          fib6_nh_find_match, &arg);
4045 
4046                 /* fib6_info uses a nexthop that does not have fib6_nh
4047                  * using the dst->dev. Should be impossible
4048                  */
4049                 if (!arg.match)
4050                         goto out;
4051                 res.nh = arg.match;
4052         } else {
4053                 res.nh = res.f6i->fib6_nh;
4054         }
4055 
4056         res.fib6_flags = res.f6i->fib6_flags;
4057         res.fib6_type = res.f6i->fib6_type;
4058         nrt = ip6_rt_cache_alloc(&res, &msg->dest, NULL);
4059         if (!nrt)
4060                 goto out;
4061 
4062         nrt->rt6i_flags = RTF_GATEWAY|RTF_UP|RTF_DYNAMIC|RTF_CACHE;
4063         if (on_link)
4064                 nrt->rt6i_flags &= ~RTF_GATEWAY;
4065 
4066         nrt->rt6i_gateway = *(struct in6_addr *)neigh->primary_key;
4067 
4068         /* rt6_insert_exception() will take care of duplicated exceptions */
4069         if (rt6_insert_exception(nrt, &res)) {
4070                 dst_release_immediate(&nrt->dst);
4071                 goto out;
4072         }
4073 
4074         netevent.old = &rt->dst;
4075         netevent.new = &nrt->dst;
4076         netevent.daddr = &msg->dest;
4077         netevent.neigh = neigh;
4078         call_netevent_notifiers(NETEVENT_REDIRECT, &netevent);
4079 
4080 out:
4081         rcu_read_unlock();
4082         neigh_release(neigh);
4083 }
4084 
4085 #ifdef CONFIG_IPV6_ROUTE_INFO
4086 static struct fib6_info *rt6_get_route_info(struct net *net,
4087                                            const struct in6_addr *prefix, int prefixlen,
4088                                            const struct in6_addr *gwaddr,
4089                                            struct net_device *dev)
4090 {
4091         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO;
4092         int ifindex = dev->ifindex;
4093         struct fib6_node *fn;
4094         struct fib6_info *rt = NULL;
4095         struct fib6_table *table;
4096 
4097         table = fib6_get_table(net, tb_id);
4098         if (!table)
4099                 return NULL;
4100 
4101         rcu_read_lock();
4102         fn = fib6_locate(&table->tb6_root, prefix, prefixlen, NULL, 0, true);
4103         if (!fn)
4104                 goto out;
4105 
4106         for_each_fib6_node_rt_rcu(fn) {
4107                 /* these routes do not use nexthops */
4108                 if (rt->nh)
4109                         continue;
4110                 if (rt->fib6_nh->fib_nh_dev->ifindex != ifindex)
4111                         continue;
4112                 if (!(rt->fib6_flags & RTF_ROUTEINFO) ||
4113                     !rt->fib6_nh->fib_nh_gw_family)
4114                         continue;
4115                 if (!ipv6_addr_equal(&rt->fib6_nh->fib_nh_gw6, gwaddr))
4116                         continue;
4117                 if (!fib6_info_hold_safe(rt))
4118                         continue;
4119                 break;
4120         }
4121 out:
4122         rcu_read_unlock();
4123         return rt;
4124 }
4125 
4126 static struct fib6_info *rt6_add_route_info(struct net *net,
4127                                            const struct in6_addr *prefix, int prefixlen,
4128                                            const struct in6_addr *gwaddr,
4129                                            struct net_device *dev,
4130                                            unsigned int pref)
4131 {
4132         struct fib6_config cfg = {
4133                 .fc_metric      = IP6_RT_PRIO_USER,
4134                 .fc_ifindex     = dev->ifindex,
4135                 .fc_dst_len     = prefixlen,
4136                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_ROUTEINFO |
4137                                   RTF_UP | RTF_PREF(pref),
4138                 .fc_protocol = RTPROT_RA,
4139                 .fc_type = RTN_UNICAST,
4140                 .fc_nlinfo.portid = 0,
4141                 .fc_nlinfo.nlh = NULL,
4142                 .fc_nlinfo.nl_net = net,
4143         };
4144 
4145         cfg.fc_table = l3mdev_fib_table(dev) ? : RT6_TABLE_INFO,
4146         cfg.fc_dst = *prefix;
4147         cfg.fc_gateway = *gwaddr;
4148 
4149         /* We should treat it as a default route if prefix length is 0. */
4150         if (!prefixlen)
4151                 cfg.fc_flags |= RTF_DEFAULT;
4152 
4153         ip6_route_add(&cfg, GFP_ATOMIC, NULL);
4154 
4155         return rt6_get_route_info(net, prefix, prefixlen, gwaddr, dev);
4156 }
4157 #endif
4158 
4159 struct fib6_info *rt6_get_dflt_router(struct net *net,
4160                                      const struct in6_addr *addr,
4161                                      struct net_device *dev)
4162 {
4163         u32 tb_id = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT;
4164         struct fib6_info *rt;
4165         struct fib6_table *table;
4166 
4167         table = fib6_get_table(net, tb_id);
4168         if (!table)
4169                 return NULL;
4170 
4171         rcu_read_lock();
4172         for_each_fib6_node_rt_rcu(&table->tb6_root) {
4173                 struct fib6_nh *nh;
4174 
4175                 /* RA routes do not use nexthops */
4176                 if (rt->nh)
4177                         continue;
4178 
4179                 nh = rt->fib6_nh;
4180                 if (dev == nh->fib_nh_dev &&
4181                     ((rt->fib6_flags & (RTF_ADDRCONF | RTF_DEFAULT)) == (RTF_ADDRCONF | RTF_DEFAULT)) &&
4182                     ipv6_addr_equal(&nh->fib_nh_gw6, addr))
4183                         break;
4184         }
4185         if (rt && !fib6_info_hold_safe(rt))
4186                 rt = NULL;
4187         rcu_read_unlock();
4188         return rt;
4189 }
4190 
4191 struct fib6_info *rt6_add_dflt_router(struct net *net,
4192                                      const struct in6_addr *gwaddr,
4193                                      struct net_device *dev,
4194                                      unsigned int pref)
4195 {
4196         struct fib6_config cfg = {
4197                 .fc_table       = l3mdev_fib_table(dev) ? : RT6_TABLE_DFLT,
4198                 .fc_metric      = IP6_RT_PRIO_USER,
4199                 .fc_ifindex     = dev->ifindex,
4200                 .fc_flags       = RTF_GATEWAY | RTF_ADDRCONF | RTF_DEFAULT |
4201                                   RTF_UP | RTF_EXPIRES | RTF_PREF(pref),
4202                 .fc_protocol = RTPROT_RA,
4203                 .fc_type = RTN_UNICAST,
4204                 .fc_nlinfo.portid = 0,
4205                 .fc_nlinfo.nlh = NULL,
4206                 .fc_nlinfo.nl_net = net,
4207         };
4208 
4209         cfg.fc_gateway = *gwaddr;
4210 
4211         if (!ip6_route_add(&cfg, GFP_ATOMIC, NULL)) {
4212                 struct fib6_table *table;
4213 
4214                 table = fib6_get_table(dev_net(dev), cfg.fc_table);
4215                 if (table)
4216                         table->flags |= RT6_TABLE_HAS_DFLT_ROUTER;
4217         }
4218 
4219         return rt6_get_dflt_router(net, gwaddr, dev);
4220 }
4221 
4222 static void __rt6_purge_dflt_routers(struct net *net,
4223                                      struct fib6_table *table)
4224 {
4225         struct fib6_info *rt;
4226 
4227 restart:
4228         rcu_read_lock();
4229         for_each_fib6_node_rt_rcu(&table->tb6_root) {
4230                 struct net_device *dev = fib6_info_nh_dev(rt);
4231                 struct inet6_dev *idev = dev ? __in6_dev_get(dev) : NULL;
4232 
4233                 if (rt->fib6_flags & (RTF_DEFAULT | RTF_ADDRCONF) &&
4234                     (!idev || idev->cnf.accept_ra != 2) &&
4235                     fib6_info_hold_safe(rt)) {
4236                         rcu_read_unlock();
4237                         ip6_del_rt(net, rt);
4238                         goto restart;
4239                 }
4240         }
4241         rcu_read_unlock();
4242 
4243         table->flags &= ~RT6_TABLE_HAS_DFLT_ROUTER;
4244 }
4245 
4246 void rt6_purge_dflt_routers(struct net *net)
4247 {
4248         struct fib6_table *table;
4249         struct hlist_head *head;
4250         unsigned int h;
4251 
4252         rcu_read_lock();
4253 
4254         for (h = 0; h < FIB6_TABLE_HASHSZ; h++) {
4255                 head = &net->ipv6.fib_table_hash[h];
4256                 hlist_for_each_entry_rcu(table, head, tb6_hlist) {
4257                         if (table->flags & RT6_TABLE_HAS_DFLT_ROUTER)
4258                                 __rt6_purge_dflt_routers(net, table);
4259                 }
4260         }
4261 
4262         rcu_read_unlock();
4263 }
4264 
4265 static void rtmsg_to_fib6_config(struct net *net,
4266                                  struct in6_rtmsg *rtmsg,
4267                                  struct fib6_config *cfg)
4268 {
4269         *cfg = (struct fib6_config){
4270                 .fc_table = l3mdev_fib_table_by_index(net, rtmsg->rtmsg_ifindex) ?
4271                          : RT6_TABLE_MAIN,
4272                 .fc_ifindex = rtmsg->rtmsg_ifindex,
4273                 .fc_metric = rtmsg->rtmsg_metric ? : IP6_RT_PRIO_USER,
4274                 .fc_expires = rtmsg->rtmsg_info,
4275                 .fc_dst_len = rtmsg->rtmsg_dst_len,
4276                 .fc_src_len = rtmsg->rtmsg_src_len,
4277                 .fc_flags = rtmsg->rtmsg_flags,
4278                 .fc_type = rtmsg->rtmsg_type,
4279 
4280                 .fc_nlinfo.nl_net = net,
4281 
4282                 .fc_dst = rtmsg->rtmsg_dst,
4283                 .fc_src = rtmsg->rtmsg_src,
4284                 .fc_gateway = rtmsg->rtmsg_gateway,
4285         };
4286 }
4287 
4288 int ipv6_route_ioctl(struct net *net, unsigned int cmd, void __user *arg)
4289 {
4290         struct fib6_config cfg;
4291         struct in6_rtmsg rtmsg;
4292         int err;
4293 
4294         switch (cmd) {
4295         case SIOCADDRT:         /* Add a route */
4296         case SIOCDELRT:         /* Delete a route */
4297                 if (!ns_capable(net->user_ns, CAP_NET_ADMIN))
4298                         return -EPERM;
4299                 err = copy_from_user(&rtmsg, arg,
4300                                      sizeof(struct in6_rtmsg));
4301                 if (err)
4302                         return -EFAULT;
4303 
4304                 rtmsg_to_fib6_config(net, &rtmsg, &cfg);
4305 
4306                 rtnl_lock();
4307                 switch (cmd) {
4308                 case SIOCADDRT:
4309                         err = ip6_route_add(&cfg, GFP_KERNEL, NULL);
4310                         break;
4311                 case SIOCDELRT:
4312                         err = ip6_route_del(&cfg, NULL);
4313                         break;
4314                 default:
4315                         err = -EINVAL;
4316                 }
4317                 rtnl_unlock();
4318 
4319                 return err;
4320         }
4321 
4322         return -EINVAL;
4323 }
4324 
4325 /*
4326  *      Drop the packet on the floor
4327  */
4328 
4329 static int ip6_pkt_drop(struct sk_buff *skb, u8 code, int ipstats_mib_noroutes)
4330 {
4331         struct dst_entry *dst = skb_dst(skb);
4332         struct net *net = dev_net(dst->dev);
4333         struct inet6_dev *idev;
4334         int type;
4335 
4336         if (netif_is_l3_master(skb->dev) &&
4337             dst->dev == net->loopback_dev)
4338                 idev = __in6_dev_get_safely(dev_get_by_index_rcu(net, IP6CB(skb)->iif));
4339         else
4340                 idev = ip6_dst_idev(dst);
4341 
4342         switch (ipstats_mib_noroutes) {
4343         case IPSTATS_MIB_INNOROUTES:
4344                 type = ipv6_addr_type(&ipv6_hdr(skb)->daddr);
4345                 if (type == IPV6_ADDR_ANY) {
4346                         IP6_INC_STATS(net, idev, IPSTATS_MIB_INADDRERRORS);
4347                         break;
4348                 }
4349                 /* FALLTHROUGH */
4350         case IPSTATS_MIB_OUTNOROUTES:
4351                 IP6_INC_STATS(net, idev, ipstats_mib_noroutes);
4352                 break;
4353         }
4354 
4355         /* Start over by dropping the dst for l3mdev case */
4356         if (netif_is_l3_master(skb->dev))
4357                 skb_dst_drop(skb);
4358 
4359         icmpv6_send(skb, ICMPV6_DEST_UNREACH, code, 0);
4360         kfree_skb(skb);
4361         return 0;
4362 }
4363 
4364 static int ip6_pkt_discard(struct sk_buff *skb)
4365 {
4366         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_INNOROUTES);
4367 }
4368 
4369 static int ip6_pkt_discard_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4370 {
4371         skb->dev = skb_dst(skb)->dev;
4372         return ip6_pkt_drop(skb, ICMPV6_NOROUTE, IPSTATS_MIB_OUTNOROUTES);
4373 }
4374 
4375 static int ip6_pkt_prohibit(struct sk_buff *skb)
4376 {
4377         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_INNOROUTES);
4378 }
4379 
4380 static int ip6_pkt_prohibit_out(struct net *net, struct sock *sk, struct sk_buff *skb)
4381 {
4382         skb->dev = skb_dst(skb)->dev;
4383         return ip6_pkt_drop(skb, ICMPV6_ADM_PROHIBITED, IPSTATS_MIB_OUTNOROUTES);
4384 }
4385 
4386 /*
4387  *      Allocate a dst for local (unicast / anycast) address.
4388  */
4389 
4390 struct fib6_info *addrconf_f6i_alloc(struct net *net,
4391                                      struct inet6_dev *idev,
4392                                      const struct in6_addr *addr,
4393                                      bool anycast, gfp_t gfp_flags)
4394 {
4395         struct fib6_config cfg = {
4396                 .fc_table = l3mdev_fib_table(idev->dev) ? : RT6_TABLE_LOCAL,
4397                 .fc_ifindex = idev->dev->ifindex,
4398                 .fc_flags = RTF_UP | RTF_NONEXTHOP,
4399                 .fc_dst = *addr,
4400                 .fc_dst_len = 128,
4401                 .fc_protocol = RTPROT_KERNEL,
4402                 .fc_nlinfo.nl_net = net,
4403                 .fc_ignore_dev_down = true,
4404         };
4405         struct fib6_info *f6i;
4406 
4407         if (anycast) {
4408                 cfg.fc_type = RTN_ANYCAST;
4409                 cfg.fc_flags |= RTF_ANYCAST;
4410         } else {
4411                 cfg.fc_type = RTN_LOCAL;
4412                 cfg.fc_flags |= RTF_LOCAL;
4413         }
4414 
4415         f6i = ip6_route_info_create(&cfg, gfp_flags, NULL);
4416         if (!IS_ERR(f6i))
4417                 f6i->dst_nocount = true;
4418         return f6i;
4419 }
4420 
4421 /* remove deleted ip from prefsrc entries */
4422 struct arg_dev_net_ip {
4423         struct net_device *dev;
4424         struct net *net;
4425         struct in6_addr *addr;
4426 };
4427 
4428 static int fib6_remove_prefsrc(struct fib6_info *rt, void *arg)
4429 {
4430         struct net_device *dev = ((struct arg_dev_net_ip *)arg)->dev;
4431         struct net *net = ((struct arg_dev_net_ip *)arg)->net;
4432         struct in6_addr *addr = ((struct arg_dev_net_ip *)arg)->addr;
4433 
4434         if (!rt->nh &&
4435             ((void *)rt->fib6_nh->fib_nh_dev == dev || !dev) &&
4436             rt != net->ipv6.fib6_null_entry &&
4437             ipv6_addr_equal(addr, &rt->fib6_prefsrc.addr)) {
4438                 spin_lock_bh(&rt6_exception_lock);
4439                 /* remove prefsrc entry */
4440                 rt->fib6_prefsrc.plen = 0;
4441                 spin_unlock_bh(&rt6_exception_lock);
4442         }
4443         return 0;
4444 }
4445 
4446 void rt6_remove_prefsrc(struct inet6_ifaddr *ifp)
4447 {
4448         struct net *net = dev_net(ifp->idev->dev);
4449         struct arg_dev_net_ip adni = {
4450                 .dev = ifp->idev->dev,
4451                 .net = net,
4452                 .addr = &ifp->addr,
4453         };
4454         fib6_clean_all(net, fib6_remove_prefsrc, &adni);
4455 }
4456 
4457 #define RTF_RA_ROUTER           (RTF_ADDRCONF | RTF_DEFAULT)
4458 
4459 /* Remove routers and update dst entries when gateway turn into host. */
4460 static int fib6_clean_tohost(struct fib6_info *rt, void *arg)
4461 {
4462         struct in6_addr *gateway = (struct in6_addr *)arg;
4463         struct fib6_nh *nh;
4464 
4465         /* RA routes do not use nexthops */
4466         if (rt->nh)
4467                 return 0;
4468 
4469         nh = rt->fib6_nh;
4470         if (((rt->fib6_flags & RTF_RA_ROUTER) == RTF_RA_ROUTER) &&
4471             nh->fib_nh_gw_family && ipv6_addr_equal(gateway, &nh->fib_nh_gw6))
4472                 return -1;
4473 
4474         /* Further clean up cached routes in exception table.
4475          * This is needed because cached route may have a different
4476          * gateway than its 'parent' in the case of an ip redirect.
4477          */
4478         fib6_nh_exceptions_clean_tohost(nh, gateway);
4479 
4480         return 0;
4481 }
4482 
4483 void rt6_clean_tohost(struct net *net, struct in6_addr *gateway)
4484 {
4485         fib6_clean_all(net, fib6_clean_tohost, gateway);
4486 }
4487 
4488 struct arg_netdev_event {
4489         const struct net_device *dev;
4490         union {
4491                 unsigned char nh_flags;
4492                 unsigned long event;
4493         };
4494 };
4495 
4496 static struct fib6_info *rt6_multipath_first_sibling(const struct fib6_info *rt)
4497 {
4498         struct fib6_info *iter;
4499         struct fib6_node *fn;
4500 
4501         fn = rcu_dereference_protected(rt->fib6_node,
4502                         lockdep_is_held(&rt->fib6_table->tb6_lock));
4503         iter = rcu_dereference_protected(fn->leaf,
4504                         lockdep_is_held(&rt->fib6_table->tb6_lock));
4505         while (iter) {
4506                 if (iter->fib6_metric == rt->fib6_metric &&
4507                     rt6_qualify_for_ecmp(iter))
4508                         return iter;
4509                 iter = rcu_dereference_protected(iter->fib6_next,
4510                                 lockdep_is_held(&rt->fib6_table->tb6_lock));
4511         }
4512 
4513         return NULL;
4514 }
4515 
4516 /* only called for fib entries with builtin fib6_nh */
4517 static bool rt6_is_dead(const struct fib6_info *rt)
4518 {
4519         if (rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD ||
4520             (rt->fib6_nh->fib_nh_flags & RTNH_F_LINKDOWN &&
4521              ip6_ignore_linkdown(rt->fib6_nh->fib_nh_dev)))
4522                 return true;
4523 
4524         return false;
4525 }
4526 
4527 static int rt6_multipath_total_weight(const struct fib6_info *rt)
4528 {
4529         struct fib6_info *iter;
4530         int total = 0;
4531 
4532         if (!rt6_is_dead(rt))
4533                 total += rt->fib6_nh->fib_nh_weight;
4534 
4535         list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings) {
4536                 if (!rt6_is_dead(iter))
4537                         total += iter->fib6_nh->fib_nh_weight;
4538         }
4539 
4540         return total;
4541 }
4542 
4543 static void rt6_upper_bound_set(struct fib6_info *rt, int *weight, int total)
4544 {
4545         int upper_bound = -1;
4546 
4547         if (!rt6_is_dead(rt)) {
4548                 *weight += rt->fib6_nh->fib_nh_weight;
4549                 upper_bound = DIV_ROUND_CLOSEST_ULL((u64) (*weight) << 31,
4550                                                     total) - 1;
4551         }
4552         atomic_set(&rt->fib6_nh->fib_nh_upper_bound, upper_bound);
4553 }
4554 
4555 static void rt6_multipath_upper_bound_set(struct fib6_info *rt, int total)
4556 {
4557         struct fib6_info *iter;
4558         int weight = 0;
4559 
4560         rt6_upper_bound_set(rt, &weight, total);
4561 
4562         list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4563                 rt6_upper_bound_set(iter, &weight, total);
4564 }
4565 
4566 void rt6_multipath_rebalance(struct fib6_info *rt)
4567 {
4568         struct fib6_info *first;
4569         int total;
4570 
4571         /* In case the entire multipath route was marked for flushing,
4572          * then there is no need to rebalance upon the removal of every
4573          * sibling route.
4574          */
4575         if (!rt->fib6_nsiblings || rt->should_flush)
4576                 return;
4577 
4578         /* During lookup routes are evaluated in order, so we need to
4579          * make sure upper bounds are assigned from the first sibling
4580          * onwards.
4581          */
4582         first = rt6_multipath_first_sibling(rt);
4583         if (WARN_ON_ONCE(!first))
4584                 return;
4585 
4586         total = rt6_multipath_total_weight(first);
4587         rt6_multipath_upper_bound_set(first, total);
4588 }
4589 
4590 static int fib6_ifup(struct fib6_info *rt, void *p_arg)
4591 {
4592         const struct arg_netdev_event *arg = p_arg;
4593         struct net *net = dev_net(arg->dev);
4594 
4595         if (rt != net->ipv6.fib6_null_entry && !rt->nh &&
4596             rt->fib6_nh->fib_nh_dev == arg->dev) {
4597                 rt->fib6_nh->fib_nh_flags &= ~arg->nh_flags;
4598                 fib6_update_sernum_upto_root(net, rt);
4599                 rt6_multipath_rebalance(rt);
4600         }
4601 
4602         return 0;
4603 }
4604 
4605 void rt6_sync_up(struct net_device *dev, unsigned char nh_flags)
4606 {
4607         struct arg_netdev_event arg = {
4608                 .dev = dev,
4609                 {
4610                         .nh_flags = nh_flags,
4611                 },
4612         };
4613 
4614         if (nh_flags & RTNH_F_DEAD && netif_carrier_ok(dev))
4615                 arg.nh_flags |= RTNH_F_LINKDOWN;
4616 
4617         fib6_clean_all(dev_net(dev), fib6_ifup, &arg);
4618 }
4619 
4620 /* only called for fib entries with inline fib6_nh */
4621 static bool rt6_multipath_uses_dev(const struct fib6_info *rt,
4622                                    const struct net_device *dev)
4623 {
4624         struct fib6_info *iter;
4625 
4626         if (rt->fib6_nh->fib_nh_dev == dev)
4627                 return true;
4628         list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4629                 if (iter->fib6_nh->fib_nh_dev == dev)
4630                         return true;
4631 
4632         return false;
4633 }
4634 
4635 static void rt6_multipath_flush(struct fib6_info *rt)
4636 {
4637         struct fib6_info *iter;
4638 
4639         rt->should_flush = 1;
4640         list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4641                 iter->should_flush = 1;
4642 }
4643 
4644 static unsigned int rt6_multipath_dead_count(const struct fib6_info *rt,
4645                                              const struct net_device *down_dev)
4646 {
4647         struct fib6_info *iter;
4648         unsigned int dead = 0;
4649 
4650         if (rt->fib6_nh->fib_nh_dev == down_dev ||
4651             rt->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4652                 dead++;
4653         list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4654                 if (iter->fib6_nh->fib_nh_dev == down_dev ||
4655                     iter->fib6_nh->fib_nh_flags & RTNH_F_DEAD)
4656                         dead++;
4657 
4658         return dead;
4659 }
4660 
4661 static void rt6_multipath_nh_flags_set(struct fib6_info *rt,
4662                                        const struct net_device *dev,
4663                                        unsigned char nh_flags)
4664 {
4665         struct fib6_info *iter;
4666 
4667         if (rt->fib6_nh->fib_nh_dev == dev)
4668                 rt->fib6_nh->fib_nh_flags |= nh_flags;
4669         list_for_each_entry(iter, &rt->fib6_siblings, fib6_siblings)
4670                 if (iter->fib6_nh->fib_nh_dev == dev)
4671                         iter->fib6_nh->fib_nh_flags |= nh_flags;
4672 }
4673 
4674 /* called with write lock held for table with rt */
4675 static int fib6_ifdown(struct fib6_info *rt, void *p_arg)
4676 {
4677         const struct arg_netdev_event *arg = p_arg;
4678         const struct net_device *dev = arg->dev;
4679         struct net *net = dev_net(dev);
4680 
4681         if (rt == net->ipv6.fib6_null_entry || rt->nh)
4682                 return 0;
4683 
4684         switch (arg->event) {
4685         case NETDEV_UNREGISTER:
4686                 return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4687         case NETDEV_DOWN:
4688                 if (rt->should_flush)
4689                         return -1;
4690                 if (!rt->fib6_nsiblings)
4691                         return rt->fib6_nh->fib_nh_dev == dev ? -1 : 0;
4692                 if (rt6_multipath_uses_dev(rt, dev)) {
4693                         unsigned int count;
4694 
4695                         count = rt6_multipath_dead_count(rt, dev);
4696                         if (rt->fib6_nsiblings + 1 == count) {
4697                                 rt6_multipath_flush(rt);
4698                                 return -1;
4699                         }
4700                         rt6_multipath_nh_flags_set(rt, dev, RTNH_F_DEAD |
4701                                                    RTNH_F_LINKDOWN);
4702                         fib6_update_sernum(net, rt);
4703                         rt6_multipath_rebalance(rt);
4704                 }
4705                 return -2;
4706         case NETDEV_CHANGE:
4707                 if (rt->fib6_nh->fib_nh_dev != dev ||
4708                     rt->fib6_flags & (RTF_LOCAL | RTF_ANYCAST))
4709                         break;
4710                 rt->fib6_nh->fib_nh_flags |= RTNH_F_LINKDOWN;
4711                 rt6_multipath_rebalance(rt);
4712                 break;
4713         }
4714 
4715         return 0;
4716 }
4717 
4718 void rt6_sync_down_dev(struct net_device *dev, unsigned long event)
4719 {
4720         struct arg_netdev_event arg = {
4721                 .dev = dev,
4722                 {
4723                         .event = event,
4724                 },
4725         };
4726         struct net *net = dev_net(dev);
4727 
4728         if (net->ipv6.sysctl.skip_notify_on_dev_down)
4729                 fib6_clean_all_skip_notify(net, fib6_ifdown, &arg);
4730         else
4731                 fib6_clean_all(net, fib6_ifdown, &arg);
4732 }
4733 
4734 void rt6_disable_ip(struct net_device *dev, unsigned long event)
4735 {
4736         rt6_sync_down_dev(dev, event);
4737         rt6_uncached_list_flush_dev(dev_net(dev), dev);
4738         neigh_ifdown(&nd_tbl, dev);
4739 }
4740 
4741 struct rt6_mtu_change_arg {
4742         struct net_device *dev;
4743         unsigned int mtu;
4744         struct fib6_info *f6i;
4745 };
4746 
4747 static int fib6_nh_mtu_change(struct fib6_nh *nh, void *_arg)
4748 {
4749         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *)_arg;
4750         struct fib6_info *f6i = arg->f6i;
4751 
4752         /* For administrative MTU increase, there is no way to discover
4753          * IPv6 PMTU increase, so PMTU increase should be updated here.
4754          * Since RFC 1981 doesn't include administrative MTU increase
4755          * update PMTU increase is a MUST. (i.e. jumbo frame)
4756          */
4757         if (nh->fib_nh_dev == arg->dev) {
4758                 struct inet6_dev *idev = __in6_dev_get(arg->dev);
4759                 u32 mtu = f6i->fib6_pmtu;
4760 
4761                 if (mtu >= arg->mtu ||
4762                     (mtu < arg->mtu && mtu == idev->cnf.mtu6))
4763                         fib6_metric_set(f6i, RTAX_MTU, arg->mtu);
4764 
4765                 spin_lock_bh(&rt6_exception_lock);
4766                 rt6_exceptions_update_pmtu(idev, nh, arg->mtu);
4767                 spin_unlock_bh(&rt6_exception_lock);
4768         }
4769 
4770         return 0;
4771 }
4772 
4773 static int rt6_mtu_change_route(struct fib6_info *f6i, void *p_arg)
4774 {
4775         struct rt6_mtu_change_arg *arg = (struct rt6_mtu_change_arg *) p_arg;
4776         struct inet6_dev *idev;
4777 
4778         /* In IPv6 pmtu discovery is not optional,
4779            so that RTAX_MTU lock cannot disable it.
4780            We still use this lock to block changes
4781            caused by addrconf/ndisc.
4782         */
4783 
4784         idev = __in6_dev_get(arg->dev);
4785         if (!idev)
4786                 return 0;
4787 
4788         if (fib6_metric_locked(f6i, RTAX_MTU))
4789                 return 0;
4790 
4791         arg->f6i = f6i;
4792         if (f6i->nh) {
4793                 /* fib6_nh_mtu_change only returns 0, so this is safe */
4794                 return nexthop_for_each_fib6_nh(f6i->nh, fib6_nh_mtu_change,
4795                                                 arg);
4796         }
4797 
4798         return fib6_nh_mtu_change(f6i->fib6_nh, arg);
4799 }
4800 
4801 void rt6_mtu_change(struct net_device *dev, unsigned int mtu)
4802 {
4803         struct rt6_mtu_change_arg arg = {
4804                 .dev = dev,
4805                 .mtu = mtu,
4806         };
4807 
4808         fib6_clean_all(dev_net(dev), rt6_mtu_change_route, &arg);
4809 }
4810 
4811 static const struct nla_policy rtm_ipv6_policy[RTA_MAX+1] = {
4812         [RTA_UNSPEC]            = { .strict_start_type = RTA_DPORT + 1 },
4813         [RTA_GATEWAY]           = { .len = sizeof(struct in6_addr) },
4814         [RTA_PREFSRC]           = { .len = sizeof(struct in6_addr) },
4815         [RTA_OIF]               = { .type = NLA_U32 },
4816         [RTA_IIF]               = { .type = NLA_U32 },
4817         [RTA_PRIORITY]          = { .type = NLA_U32 },
4818         [RTA_METRICS]           = { .type = NLA_NESTED },
4819         [RTA_MULTIPATH]         = { .len = sizeof(struct rtnexthop) },
4820         [RTA_PREF]              = { .type = NLA_U8 },
4821         [RTA_ENCAP_TYPE]        = { .type = NLA_U16 },
4822         [RTA_ENCAP]             = { .type = NLA_NESTED },
4823         [RTA_EXPIRES]           = { .type = NLA_U32 },
4824         [RTA_UID]               = { .type = NLA_U32 },
4825         [RTA_MARK]              = { .type = NLA_U32 },
4826         [RTA_TABLE]             = { .type = NLA_U32 },
4827         [RTA_IP_PROTO]          = { .type = NLA_U8 },
4828         [RTA_SPORT]             = { .type = NLA_U16 },
4829         [RTA_DPORT]             = { .type = NLA_U16 },
4830         [RTA_NH_ID]             = { .type = NLA_U32 },
4831 };
4832 
4833 static int rtm_to_fib6_config(struct sk_buff *skb, struct nlmsghdr *nlh,
4834                               struct fib6_config *cfg,
4835                               struct netlink_ext_ack *extack)
4836 {
4837         struct rtmsg *rtm;
4838         struct nlattr *tb[RTA_MAX+1];
4839         unsigned int pref;
4840         int err;
4841 
4842         err = nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
4843                                      rtm_ipv6_policy, extack);
4844         if (err < 0)
4845                 goto errout;
4846 
4847         err = -EINVAL;
4848         rtm = nlmsg_data(nlh);
4849 
4850         *cfg = (struct fib6_config){
4851                 .fc_table = rtm->rtm_table,
4852                 .fc_dst_len = rtm->rtm_dst_len,
4853                 .fc_src_len = rtm->rtm_src_len,
4854                 .fc_flags = RTF_UP,
4855                 .fc_protocol = rtm->rtm_protocol,
4856                 .fc_type = rtm->rtm_type,
4857 
4858                 .fc_nlinfo.portid = NETLINK_CB(skb).portid,
4859                 .fc_nlinfo.nlh = nlh,
4860                 .fc_nlinfo.nl_net = sock_net(skb->sk),
4861         };
4862 
4863         if (rtm->rtm_type == RTN_UNREACHABLE ||
4864             rtm->rtm_type == RTN_BLACKHOLE ||
4865             rtm->rtm_type == RTN_PROHIBIT ||
4866             rtm->rtm_type == RTN_THROW)
4867                 cfg->fc_flags |= RTF_REJECT;
4868 
4869         if (rtm->rtm_type == RTN_LOCAL)
4870                 cfg->fc_flags |= RTF_LOCAL;
4871 
4872         if (rtm->rtm_flags & RTM_F_CLONED)
4873                 cfg->fc_flags |= RTF_CACHE;
4874 
4875         cfg->fc_flags |= (rtm->rtm_flags & RTNH_F_ONLINK);
4876 
4877         if (tb[RTA_NH_ID]) {
4878                 if (tb[RTA_GATEWAY]   || tb[RTA_OIF] ||
4879                     tb[RTA_MULTIPATH] || tb[RTA_ENCAP]) {
4880                         NL_SET_ERR_MSG(extack,
4881                                        "Nexthop specification and nexthop id are mutually exclusive");
4882                         goto errout;
4883                 }
4884                 cfg->fc_nh_id = nla_get_u32(tb[RTA_NH_ID]);
4885         }
4886 
4887         if (tb[RTA_GATEWAY]) {
4888                 cfg->fc_gateway = nla_get_in6_addr(tb[RTA_GATEWAY]);
4889                 cfg->fc_flags |= RTF_GATEWAY;
4890         }
4891         if (tb[RTA_VIA]) {
4892                 NL_SET_ERR_MSG(extack, "IPv6 does not support RTA_VIA attribute");
4893                 goto errout;
4894         }
4895 
4896         if (tb[RTA_DST]) {
4897                 int plen = (rtm->rtm_dst_len + 7) >> 3;
4898 
4899                 if (nla_len(tb[RTA_DST]) < plen)
4900                         goto errout;
4901 
4902                 nla_memcpy(&cfg->fc_dst, tb[RTA_DST], plen);
4903         }
4904 
4905         if (tb[RTA_SRC]) {
4906                 int plen = (rtm->rtm_src_len + 7) >> 3;
4907 
4908                 if (nla_len(tb[RTA_SRC]) < plen)
4909                         goto errout;
4910 
4911                 nla_memcpy(&cfg->fc_src, tb[RTA_SRC], plen);
4912         }
4913 
4914         if (tb[RTA_PREFSRC])
4915                 cfg->fc_prefsrc = nla_get_in6_addr(tb[RTA_PREFSRC]);
4916 
4917         if (tb[RTA_OIF])
4918                 cfg->fc_ifindex = nla_get_u32(tb[RTA_OIF]);
4919 
4920         if (tb[RTA_PRIORITY])
4921                 cfg->fc_metric = nla_get_u32(tb[RTA_PRIORITY]);
4922 
4923         if (tb[RTA_METRICS]) {
4924                 cfg->fc_mx = nla_data(tb[RTA_METRICS]);
4925                 cfg->fc_mx_len = nla_len(tb[RTA_METRICS]);
4926         }
4927 
4928         if (tb[RTA_TABLE])
4929                 cfg->fc_table = nla_get_u32(tb[RTA_TABLE]);
4930 
4931         if (tb[RTA_MULTIPATH]) {
4932                 cfg->fc_mp = nla_data(tb[RTA_MULTIPATH]);
4933                 cfg->fc_mp_len = nla_len(tb[RTA_MULTIPATH]);
4934 
4935                 err = lwtunnel_valid_encap_type_attr(cfg->fc_mp,
4936                                                      cfg->fc_mp_len, extack);
4937                 if (err < 0)
4938                         goto errout;
4939         }
4940 
4941         if (tb[RTA_PREF]) {
4942                 pref = nla_get_u8(tb[RTA_PREF]);
4943                 if (pref != ICMPV6_ROUTER_PREF_LOW &&
4944                     pref != ICMPV6_ROUTER_PREF_HIGH)
4945                         pref = ICMPV6_ROUTER_PREF_MEDIUM;
4946                 cfg->fc_flags |= RTF_PREF(pref);
4947         }
4948 
4949         if (tb[RTA_ENCAP])
4950                 cfg->fc_encap = tb[RTA_ENCAP];
4951 
4952         if (tb[RTA_ENCAP_TYPE]) {
4953                 cfg->fc_encap_type = nla_get_u16(tb[RTA_ENCAP_TYPE]);
4954 
4955                 err = lwtunnel_valid_encap_type(cfg->fc_encap_type, extack);
4956                 if (err < 0)
4957                         goto errout;
4958         }
4959 
4960         if (tb[RTA_EXPIRES]) {
4961                 unsigned long timeout = addrconf_timeout_fixup(nla_get_u32(tb[RTA_EXPIRES]), HZ);
4962 
4963                 if (addrconf_finite_timeout(timeout)) {
4964                         cfg->fc_expires = jiffies_to_clock_t(timeout * HZ);
4965                         cfg->fc_flags |= RTF_EXPIRES;
4966                 }
4967         }
4968 
4969         err = 0;
4970 errout:
4971         return err;
4972 }
4973 
4974 struct rt6_nh {
4975         struct fib6_info *fib6_info;
4976         struct fib6_config r_cfg;
4977         struct list_head next;
4978 };
4979 
4980 static int ip6_route_info_append(struct net *net,
4981                                  struct list_head *rt6_nh_list,
4982                                  struct fib6_info *rt,
4983                                  struct fib6_config *r_cfg)
4984 {
4985         struct rt6_nh *nh;
4986         int err = -EEXIST;
4987 
4988         list_for_each_entry(nh, rt6_nh_list, next) {
4989                 /* check if fib6_info already exists */
4990                 if (rt6_duplicate_nexthop(nh->fib6_info, rt))
4991                         return err;
4992         }
4993 
4994         nh = kzalloc(sizeof(*nh), GFP_KERNEL);
4995         if (!nh)
4996                 return -ENOMEM;
4997         nh->fib6_info = rt;
4998         memcpy(&nh->r_cfg, r_cfg, sizeof(*r_cfg));
4999         list_add_tail(&nh->next, rt6_nh_list);
5000 
5001         return 0;
5002 }
5003 
5004 static void ip6_route_mpath_notify(struct fib6_info *rt,
5005                                    struct fib6_info *rt_last,
5006                                    struct nl_info *info,
5007                                    __u16 nlflags)
5008 {
5009         /* if this is an APPEND route, then rt points to the first route
5010          * inserted and rt_last points to last route inserted. Userspace
5011          * wants a consistent dump of the route which starts at the first
5012          * nexthop. Since sibling routes are always added at the end of
5013          * the list, find the first sibling of the last route appended
5014          */
5015         if ((nlflags & NLM_F_APPEND) && rt_last && rt_last->fib6_nsiblings) {
5016                 rt = list_first_entry(&rt_last->fib6_siblings,
5017                                       struct fib6_info,
5018                                       fib6_siblings);
5019         }
5020 
5021         if (rt)
5022                 inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
5023 }
5024 
5025 static int ip6_route_multipath_add(struct fib6_config *cfg,
5026                                    struct netlink_ext_ack *extack)
5027 {
5028         struct fib6_info *rt_notif = NULL, *rt_last = NULL;
5029         struct nl_info *info = &cfg->fc_nlinfo;
5030         enum fib_event_type event_type;
5031         struct fib6_config r_cfg;
5032         struct rtnexthop *rtnh;
5033         struct fib6_info *rt;
5034         struct rt6_nh *err_nh;
5035         struct rt6_nh *nh, *nh_safe;
5036         __u16 nlflags;
5037         int remaining;
5038         int attrlen;
5039         int err = 1;
5040         int nhn = 0;
5041         int replace = (cfg->fc_nlinfo.nlh &&
5042                        (cfg->fc_nlinfo.nlh->nlmsg_flags & NLM_F_REPLACE));
5043         LIST_HEAD(rt6_nh_list);
5044 
5045         nlflags = replace ? NLM_F_REPLACE : NLM_F_CREATE;
5046         if (info->nlh && info->nlh->nlmsg_flags & NLM_F_APPEND)
5047                 nlflags |= NLM_F_APPEND;
5048 
5049         remaining = cfg->fc_mp_len;
5050         rtnh = (struct rtnexthop *)cfg->fc_mp;
5051 
5052         /* Parse a Multipath Entry and build a list (rt6_nh_list) of
5053          * fib6_info structs per nexthop
5054          */
5055         while (rtnh_ok(rtnh, remaining)) {
5056                 memcpy(&r_cfg, cfg, sizeof(*cfg));
5057                 if (rtnh->rtnh_ifindex)
5058                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5059 
5060                 attrlen = rtnh_attrlen(rtnh);
5061                 if (attrlen > 0) {
5062                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5063 
5064                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5065                         if (nla) {
5066                                 r_cfg.fc_gateway = nla_get_in6_addr(nla);
5067                                 r_cfg.fc_flags |= RTF_GATEWAY;
5068                         }
5069                         r_cfg.fc_encap = nla_find(attrs, attrlen, RTA_ENCAP);
5070                         nla = nla_find(attrs, attrlen, RTA_ENCAP_TYPE);
5071                         if (nla)
5072                                 r_cfg.fc_encap_type = nla_get_u16(nla);
5073                 }
5074 
5075                 r_cfg.fc_flags |= (rtnh->rtnh_flags & RTNH_F_ONLINK);
5076                 rt = ip6_route_info_create(&r_cfg, GFP_KERNEL, extack);
5077                 if (IS_ERR(rt)) {
5078                         err = PTR_ERR(rt);
5079                         rt = NULL;
5080                         goto cleanup;
5081                 }
5082                 if (!rt6_qualify_for_ecmp(rt)) {
5083                         err = -EINVAL;
5084                         NL_SET_ERR_MSG(extack,
5085                                        "Device only routes can not be added for IPv6 using the multipath API.");
5086                         fib6_info_release(rt);
5087                         goto cleanup;
5088                 }
5089 
5090                 rt->fib6_nh->fib_nh_weight = rtnh->rtnh_hops + 1;
5091 
5092                 err = ip6_route_info_append(info->nl_net, &rt6_nh_list,
5093                                             rt, &r_cfg);
5094                 if (err) {
5095                         fib6_info_release(rt);
5096                         goto cleanup;
5097                 }
5098 
5099                 rtnh = rtnh_next(rtnh, &remaining);
5100         }
5101 
5102         if (list_empty(&rt6_nh_list)) {
5103                 NL_SET_ERR_MSG(extack,
5104                                "Invalid nexthop configuration - no valid nexthops");
5105                 return -EINVAL;
5106         }
5107 
5108         /* for add and replace send one notification with all nexthops.
5109          * Skip the notification in fib6_add_rt2node and send one with
5110          * the full route when done
5111          */
5112         info->skip_notify = 1;
5113 
5114         /* For add and replace, send one notification with all nexthops. For
5115          * append, send one notification with all appended nexthops.
5116          */
5117         info->skip_notify_kernel = 1;
5118 
5119         err_nh = NULL;
5120         list_for_each_entry(nh, &rt6_nh_list, next) {
5121                 err = __ip6_ins_rt(nh->fib6_info, info, extack);
5122                 fib6_info_release(nh->fib6_info);
5123 
5124                 if (!err) {
5125                         /* save reference to last route successfully inserted */
5126                         rt_last = nh->fib6_info;
5127 
5128                         /* save reference to first route for notification */
5129                         if (!rt_notif)
5130                                 rt_notif = nh->fib6_info;
5131                 }
5132 
5133                 /* nh->fib6_info is used or freed at this point, reset to NULL*/
5134                 nh->fib6_info = NULL;
5135                 if (err) {
5136                         if (replace && nhn)
5137                                 NL_SET_ERR_MSG_MOD(extack,
5138                                                    "multipath route replace failed (check consistency of installed routes)");
5139                         err_nh = nh;
5140                         goto add_errout;
5141                 }
5142 
5143                 /* Because each route is added like a single route we remove
5144                  * these flags after the first nexthop: if there is a collision,
5145                  * we have already failed to add the first nexthop:
5146                  * fib6_add_rt2node() has rejected it; when replacing, old
5147                  * nexthops have been replaced by first new, the rest should
5148                  * be added to it.
5149                  */
5150                 cfg->fc_nlinfo.nlh->nlmsg_flags &= ~(NLM_F_EXCL |
5151                                                      NLM_F_REPLACE);
5152                 nhn++;
5153         }
5154 
5155         event_type = replace ? FIB_EVENT_ENTRY_REPLACE : FIB_EVENT_ENTRY_ADD;
5156         err = call_fib6_multipath_entry_notifiers(info->nl_net, event_type,
5157                                                   rt_notif, nhn - 1, extack);
5158         if (err) {
5159                 /* Delete all the siblings that were just added */
5160                 err_nh = NULL;
5161                 goto add_errout;
5162         }
5163 
5164         /* success ... tell user about new route */
5165         ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5166         goto cleanup;
5167 
5168 add_errout:
5169         /* send notification for routes that were added so that
5170          * the delete notifications sent by ip6_route_del are
5171          * coherent
5172          */
5173         if (rt_notif)
5174                 ip6_route_mpath_notify(rt_notif, rt_last, info, nlflags);
5175 
5176         /* Delete routes that were already added */
5177         list_for_each_entry(nh, &rt6_nh_list, next) {
5178                 if (err_nh == nh)
5179                         break;
5180                 ip6_route_del(&nh->r_cfg, extack);
5181         }
5182 
5183 cleanup:
5184         list_for_each_entry_safe(nh, nh_safe, &rt6_nh_list, next) {
5185                 if (nh->fib6_info)
5186                         fib6_info_release(nh->fib6_info);
5187                 list_del(&nh->next);
5188                 kfree(nh);
5189         }
5190 
5191         return err;
5192 }
5193 
5194 static int ip6_route_multipath_del(struct fib6_config *cfg,
5195                                    struct netlink_ext_ack *extack)
5196 {
5197         struct fib6_config r_cfg;
5198         struct rtnexthop *rtnh;
5199         int remaining;
5200         int attrlen;
5201         int err = 1, last_err = 0;
5202 
5203         remaining = cfg->fc_mp_len;
5204         rtnh = (struct rtnexthop *)cfg->fc_mp;
5205 
5206         /* Parse a Multipath Entry */
5207         while (rtnh_ok(rtnh, remaining)) {
5208                 memcpy(&r_cfg, cfg, sizeof(*cfg));
5209                 if (rtnh->rtnh_ifindex)
5210                         r_cfg.fc_ifindex = rtnh->rtnh_ifindex;
5211 
5212                 attrlen = rtnh_attrlen(rtnh);
5213                 if (attrlen > 0) {
5214                         struct nlattr *nla, *attrs = rtnh_attrs(rtnh);
5215 
5216                         nla = nla_find(attrs, attrlen, RTA_GATEWAY);
5217                         if (nla) {
5218                                 nla_memcpy(&r_cfg.fc_gateway, nla, 16);
5219                                 r_cfg.fc_flags |= RTF_GATEWAY;
5220                         }
5221                 }
5222                 err = ip6_route_del(&r_cfg, extack);
5223                 if (err)
5224                         last_err = err;
5225 
5226                 rtnh = rtnh_next(rtnh, &remaining);
5227         }
5228 
5229         return last_err;
5230 }
5231 
5232 static int inet6_rtm_delroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5233                               struct netlink_ext_ack *extack)
5234 {
5235         struct fib6_config cfg;
5236         int err;
5237 
5238         err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5239         if (err < 0)
5240                 return err;
5241 
5242         if (cfg.fc_nh_id &&
5243             !nexthop_find_by_id(sock_net(skb->sk), cfg.fc_nh_id)) {
5244                 NL_SET_ERR_MSG(extack, "Nexthop id does not exist");
5245                 return -EINVAL;
5246         }
5247 
5248         if (cfg.fc_mp)
5249                 return ip6_route_multipath_del(&cfg, extack);
5250         else {
5251                 cfg.fc_delete_all_nh = 1;
5252                 return ip6_route_del(&cfg, extack);
5253         }
5254 }
5255 
5256 static int inet6_rtm_newroute(struct sk_buff *skb, struct nlmsghdr *nlh,
5257                               struct netlink_ext_ack *extack)
5258 {
5259         struct fib6_config cfg;
5260         int err;
5261 
5262         err = rtm_to_fib6_config(skb, nlh, &cfg, extack);
5263         if (err < 0)
5264                 return err;
5265 
5266         if (cfg.fc_metric == 0)
5267                 cfg.fc_metric = IP6_RT_PRIO_USER;
5268 
5269         if (cfg.fc_mp)
5270                 return ip6_route_multipath_add(&cfg, extack);
5271         else
5272                 return ip6_route_add(&cfg, GFP_KERNEL, extack);
5273 }
5274 
5275 /* add the overhead of this fib6_nh to nexthop_len */
5276 static int rt6_nh_nlmsg_size(struct fib6_nh *nh, void *arg)
5277 {
5278         int *nexthop_len = arg;
5279 
5280         *nexthop_len += nla_total_size(0)        /* RTA_MULTIPATH */
5281                      + NLA_ALIGN(sizeof(struct rtnexthop))
5282                      + nla_total_size(16); /* RTA_GATEWAY */
5283 
5284         if (nh->fib_nh_lws) {
5285                 /* RTA_ENCAP_TYPE */
5286                 *nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5287                 /* RTA_ENCAP */
5288                 *nexthop_len += nla_total_size(2);
5289         }
5290 
5291         return 0;
5292 }
5293 
5294 static size_t rt6_nlmsg_size(struct fib6_info *f6i)
5295 {
5296         int nexthop_len;
5297 
5298         if (f6i->nh) {
5299                 nexthop_len = nla_total_size(4); /* RTA_NH_ID */
5300                 nexthop_for_each_fib6_nh(f6i->nh, rt6_nh_nlmsg_size,
5301                                          &nexthop_len);
5302         } else {
5303                 struct fib6_nh *nh = f6i->fib6_nh;
5304 
5305                 nexthop_len = 0;
5306                 if (f6i->fib6_nsiblings) {
5307                         nexthop_len = nla_total_size(0)  /* RTA_MULTIPATH */
5308                                     + NLA_ALIGN(sizeof(struct rtnexthop))
5309                                     + nla_total_size(16) /* RTA_GATEWAY */
5310                                     + lwtunnel_get_encap_size(nh->fib_nh_lws);
5311 
5312                         nexthop_len *= f6i->fib6_nsiblings;
5313                 }
5314                 nexthop_len += lwtunnel_get_encap_size(nh->fib_nh_lws);
5315         }
5316 
5317         return NLMSG_ALIGN(sizeof(struct rtmsg))
5318                + nla_total_size(16) /* RTA_SRC */
5319                + nla_total_size(16) /* RTA_DST */
5320                + nla_total_size(16) /* RTA_GATEWAY */
5321                + nla_total_size(16) /* RTA_PREFSRC */
5322                + nla_total_size(4) /* RTA_TABLE */
5323                + nla_total_size(4) /* RTA_IIF */
5324                + nla_total_size(4) /* RTA_OIF */
5325                + nla_total_size(4) /* RTA_PRIORITY */
5326                + RTAX_MAX * nla_total_size(4) /* RTA_METRICS */
5327                + nla_total_size(sizeof(struct rta_cacheinfo))
5328                + nla_total_size(TCP_CA_NAME_MAX) /* RTAX_CC_ALGO */
5329                + nla_total_size(1) /* RTA_PREF */
5330                + nexthop_len;
5331 }
5332 
5333 static int rt6_fill_node_nexthop(struct sk_buff *skb, struct nexthop *nh,
5334                                  unsigned char *flags)
5335 {
5336         if (nexthop_is_multipath(nh)) {
5337                 struct nlattr *mp;
5338 
5339                 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5340                 if (!mp)
5341                         goto nla_put_failure;
5342 
5343                 if (nexthop_mpath_fill_node(skb, nh, AF_INET6))
5344                         goto nla_put_failure;
5345 
5346                 nla_nest_end(skb, mp);
5347         } else {
5348                 struct fib6_nh *fib6_nh;
5349 
5350                 fib6_nh = nexthop_fib6_nh(nh);
5351                 if (fib_nexthop_info(skb, &fib6_nh->nh_common, AF_INET6,
5352                                      flags, false) < 0)
5353                         goto nla_put_failure;
5354         }
5355 
5356         return 0;
5357 
5358 nla_put_failure:
5359         return -EMSGSIZE;
5360 }
5361 
5362 static int rt6_fill_node(struct net *net, struct sk_buff *skb,
5363                          struct fib6_info *rt, struct dst_entry *dst,
5364                          struct in6_addr *dest, struct in6_addr *src,
5365                          int iif, int type, u32 portid, u32 seq,
5366                          unsigned int flags)
5367 {
5368         struct rt6_info *rt6 = (struct rt6_info *)dst;
5369         struct rt6key *rt6_dst, *rt6_src;
5370         u32 *pmetrics, table, rt6_flags;
5371         unsigned char nh_flags = 0;
5372         struct nlmsghdr *nlh;
5373         struct rtmsg *rtm;
5374         long expires = 0;
5375 
5376         nlh = nlmsg_put(skb, portid, seq, type, sizeof(*rtm), flags);
5377         if (!nlh)
5378                 return -EMSGSIZE;
5379 
5380         if (rt6) {
5381                 rt6_dst = &rt6->rt6i_dst;
5382                 rt6_src = &rt6->rt6i_src;
5383                 rt6_flags = rt6->rt6i_flags;
5384         } else {
5385                 rt6_dst = &rt->fib6_dst;
5386                 rt6_src = &rt->fib6_src;
5387                 rt6_flags = rt->fib6_flags;
5388         }
5389 
5390         rtm = nlmsg_data(nlh);
5391         rtm->rtm_family = AF_INET6;
5392         rtm->rtm_dst_len = rt6_dst->plen;
5393         rtm->rtm_src_len = rt6_src->plen;
5394         rtm->rtm_tos = 0;
5395         if (rt->fib6_table)
5396                 table = rt->fib6_table->tb6_id;
5397         else
5398                 table = RT6_TABLE_UNSPEC;
5399         rtm->rtm_table = table < 256 ? table : RT_TABLE_COMPAT;
5400         if (nla_put_u32(skb, RTA_TABLE, table))
5401                 goto nla_put_failure;
5402 
5403         rtm->rtm_type = rt->fib6_type;
5404         rtm->rtm_flags = 0;
5405         rtm->rtm_scope = RT_SCOPE_UNIVERSE;
5406         rtm->rtm_protocol = rt->fib6_protocol;
5407 
5408         if (rt6_flags & RTF_CACHE)
5409                 rtm->rtm_flags |= RTM_F_CLONED;
5410 
5411         if (dest) {
5412                 if (nla_put_in6_addr(skb, RTA_DST, dest))
5413                         goto nla_put_failure;
5414                 rtm->rtm_dst_len = 128;
5415         } else if (rtm->rtm_dst_len)
5416                 if (nla_put_in6_addr(skb, RTA_DST, &rt6_dst->addr))
5417                         goto nla_put_failure;
5418 #ifdef CONFIG_IPV6_SUBTREES
5419         if (src) {
5420                 if (nla_put_in6_addr(skb, RTA_SRC, src))
5421                         goto nla_put_failure;
5422                 rtm->rtm_src_len = 128;
5423         } else if (rtm->rtm_src_len &&
5424                    nla_put_in6_addr(skb, RTA_SRC, &rt6_src->addr))
5425                 goto nla_put_failure;
5426 #endif
5427         if (iif) {
5428 #ifdef CONFIG_IPV6_MROUTE
5429                 if (ipv6_addr_is_multicast(&rt6_dst->addr)) {
5430                         int err = ip6mr_get_route(net, skb, rtm, portid);
5431 
5432                         if (err == 0)
5433                                 return 0;
5434                         if (err < 0)
5435                                 goto nla_put_failure;
5436                 } else
5437 #endif
5438                         if (nla_put_u32(skb, RTA_IIF, iif))
5439                                 goto nla_put_failure;
5440         } else if (dest) {
5441                 struct in6_addr saddr_buf;
5442                 if (ip6_route_get_saddr(net, rt, dest, 0, &saddr_buf) == 0 &&
5443                     nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5444                         goto nla_put_failure;
5445         }
5446 
5447         if (rt->fib6_prefsrc.plen) {
5448                 struct in6_addr saddr_buf;
5449                 saddr_buf = rt->fib6_prefsrc.addr;
5450                 if (nla_put_in6_addr(skb, RTA_PREFSRC, &saddr_buf))
5451                         goto nla_put_failure;
5452         }
5453 
5454         pmetrics = dst ? dst_metrics_ptr(dst) : rt->fib6_metrics->metrics;
5455         if (rtnetlink_put_metrics(skb, pmetrics) < 0)
5456                 goto nla_put_failure;
5457 
5458         if (nla_put_u32(skb, RTA_PRIORITY, rt->fib6_metric))
5459                 goto nla_put_failure;
5460 
5461         /* For multipath routes, walk the siblings list and add
5462          * each as a nexthop within RTA_MULTIPATH.
5463          */
5464         if (rt6) {
5465                 if (rt6_flags & RTF_GATEWAY &&
5466                     nla_put_in6_addr(skb, RTA_GATEWAY, &rt6->rt6i_gateway))
5467                         goto nla_put_failure;
5468 
5469                 if (dst->dev && nla_put_u32(skb, RTA_OIF, dst->dev->ifindex))
5470                         goto nla_put_failure;
5471         } else if (rt->fib6_nsiblings) {
5472                 struct fib6_info *sibling, *next_sibling;
5473                 struct nlattr *mp;
5474 
5475                 mp = nla_nest_start_noflag(skb, RTA_MULTIPATH);
5476                 if (!mp)
5477                         goto nla_put_failure;
5478 
5479                 if (fib_add_nexthop(skb, &rt->fib6_nh->nh_common,
5480                                     rt->fib6_nh->fib_nh_weight, AF_INET6) < 0)
5481                         goto nla_put_failure;
5482 
5483                 list_for_each_entry_safe(sibling, next_sibling,
5484                                          &rt->fib6_siblings, fib6_siblings) {
5485                         if (fib_add_nexthop(skb, &sibling->fib6_nh->nh_common,
5486                                             sibling->fib6_nh->fib_nh_weight,
5487                                             AF_INET6) < 0)
5488                                 goto nla_put_failure;
5489                 }
5490 
5491                 nla_nest_end(skb, mp);
5492         } else if (rt->nh) {
5493                 if (nla_put_u32(skb, RTA_NH_ID, rt->nh->id))
5494                         goto nla_put_failure;
5495 
5496                 if (nexthop_is_blackhole(rt->nh))
5497                         rtm->rtm_type = RTN_BLACKHOLE;
5498 
5499                 if (rt6_fill_node_nexthop(skb, rt->nh, &nh_flags) < 0)
5500                         goto nla_put_failure;
5501 
5502                 rtm->rtm_flags |= nh_flags;
5503         } else {
5504                 if (fib_nexthop_info(skb, &rt->fib6_nh->nh_common, AF_INET6,
5505                                      &nh_flags, false) < 0)
5506                         goto nla_put_failure;
5507 
5508                 rtm->rtm_flags |= nh_flags;
5509         }
5510 
5511         if (rt6_flags & RTF_EXPIRES) {
5512                 expires = dst ? dst->expires : rt->expires;
5513                 expires -= jiffies;
5514         }
5515 
5516         if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
5517                 goto nla_put_failure;
5518 
5519         if (nla_put_u8(skb, RTA_PREF, IPV6_EXTRACT_PREF(rt6_flags)))
5520                 goto nla_put_failure;
5521 
5522 
5523         nlmsg_end(skb, nlh);
5524         return 0;
5525 
5526 nla_put_failure:
5527         nlmsg_cancel(skb, nlh);
5528         return -EMSGSIZE;
5529 }
5530 
5531 static int fib6_info_nh_uses_dev(struct fib6_nh *nh, void *arg)
5532 {
5533         const struct net_device *dev = arg;
5534 
5535         if (nh->fib_nh_dev == dev)
5536                 return 1;
5537 
5538         return 0;
5539 }
5540 
5541 static bool fib6_info_uses_dev(const struct fib6_info *f6i,
5542                                const struct net_device *dev)
5543 {
5544         if (f6i->nh) {
5545                 struct net_device *_dev = (struct net_device *)dev;
5546 
5547                 return !!nexthop_for_each_fib6_nh(f6i->nh,
5548                                                   fib6_info_nh_uses_dev,
5549                                                   _dev);
5550         }
5551 
5552         if (f6i->fib6_nh->fib_nh_dev == dev)
5553                 return true;
5554 
5555         if (f6i->fib6_nsiblings) {
5556                 struct fib6_info *sibling, *next_sibling;
5557 
5558                 list_for_each_entry_safe(sibling, next_sibling,
5559                                          &f6i->fib6_siblings, fib6_siblings) {
5560                         if (sibling->fib6_nh->fib_nh_dev == dev)
5561                                 return true;
5562                 }
5563         }
5564 
5565         return false;
5566 }
5567 
5568 struct fib6_nh_exception_dump_walker {
5569         struct rt6_rtnl_dump_arg *dump;
5570         struct fib6_info *rt;
5571         unsigned int flags;
5572         unsigned int skip;
5573         unsigned int count;
5574 };
5575 
5576 static int rt6_nh_dump_exceptions(struct fib6_nh *nh, void *arg)
5577 {
5578         struct fib6_nh_exception_dump_walker *w = arg;
5579         struct rt6_rtnl_dump_arg *dump = w->dump;
5580         struct rt6_exception_bucket *bucket;
5581         struct rt6_exception *rt6_ex;
5582         int i, err;
5583 
5584         bucket = fib6_nh_get_excptn_bucket(nh, NULL);
5585         if (!bucket)
5586                 return 0;
5587 
5588         for (i = 0; i < FIB6_EXCEPTION_BUCKET_SIZE; i++) {
5589                 hlist_for_each_entry(rt6_ex, &bucket->chain, hlist) {
5590                         if (w->skip) {
5591                                 w->skip--;
5592                                 continue;
5593                         }
5594 
5595                         /* Expiration of entries doesn't bump sernum, insertion
5596                          * does. Removal is triggered by insertion, so we can
5597                          * rely on the fact that if entries change between two
5598                          * partial dumps, this node is scanned again completely,
5599                          * see rt6_insert_exception() and fib6_dump_table().
5600                          *
5601                          * Count expired entries we go through as handled
5602                          * entries that we'll skip next time, in case of partial
5603                          * node dump. Otherwise, if entries expire meanwhile,
5604                          * we'll skip the wrong amount.
5605                          */
5606                         if (rt6_check_expired(rt6_ex->rt6i)) {
5607                                 w->count++;
5608                                 continue;
5609                         }
5610 
5611                         err = rt6_fill_node(dump->net, dump->skb, w->rt,
5612                                             &rt6_ex->rt6i->dst, NULL, NULL, 0,
5613                                             RTM_NEWROUTE,
5614                                             NETLINK_CB(dump->cb->skb).portid,
5615                                             dump->cb->nlh->nlmsg_seq, w->flags);
5616                         if (err)
5617                                 return err;
5618 
5619                         w->count++;
5620                 }
5621                 bucket++;
5622         }
5623 
5624         return 0;
5625 }
5626 
5627 /* Return -1 if done with node, number of handled routes on partial dump */
5628 int rt6_dump_route(struct fib6_info *rt, void *p_arg, unsigned int skip)
5629 {
5630         struct rt6_rtnl_dump_arg *arg = (struct rt6_rtnl_dump_arg *) p_arg;
5631         struct fib_dump_filter *filter = &arg->filter;
5632         unsigned int flags = NLM_F_MULTI;
5633         struct net *net = arg->net;
5634         int count = 0;
5635 
5636         if (rt == net->ipv6.fib6_null_entry)
5637                 return -1;
5638 
5639         if ((filter->flags & RTM_F_PREFIX) &&
5640             !(rt->fib6_flags & RTF_PREFIX_RT)) {
5641                 /* success since this is not a prefix route */
5642                 return -1;
5643         }
5644         if (filter->filter_set &&
5645             ((filter->rt_type  && rt->fib6_type != filter->rt_type) ||
5646              (filter->dev      && !fib6_info_uses_dev(rt, filter->dev)) ||
5647              (filter->protocol && rt->fib6_protocol != filter->protocol))) {
5648                 return -1;
5649         }
5650 
5651         if (filter->filter_set ||
5652             !filter->dump_routes || !filter->dump_exceptions) {
5653                 flags |= NLM_F_DUMP_FILTERED;
5654         }
5655 
5656         if (filter->dump_routes) {
5657                 if (skip) {
5658                         skip--;
5659                 } else {
5660                         if (rt6_fill_node(net, arg->skb, rt, NULL, NULL, NULL,
5661                                           0, RTM_NEWROUTE,
5662                                           NETLINK_CB(arg->cb->skb).portid,
5663                                           arg->cb->nlh->nlmsg_seq, flags)) {
5664                                 return 0;
5665                         }
5666                         count++;
5667                 }
5668         }
5669 
5670         if (filter->dump_exceptions) {
5671                 struct fib6_nh_exception_dump_walker w = { .dump = arg,
5672                                                            .rt = rt,
5673                                                            .flags = flags,
5674                                                            .skip = skip,
5675                                                            .count = 0 };
5676                 int err;
5677 
5678                 rcu_read_lock();
5679                 if (rt->nh) {
5680                         err = nexthop_for_each_fib6_nh(rt->nh,
5681                                                        rt6_nh_dump_exceptions,
5682                                                        &w);
5683                 } else {
5684                         err = rt6_nh_dump_exceptions(rt->fib6_nh, &w);
5685                 }
5686                 rcu_read_unlock();
5687 
5688                 if (err)
5689                         return count += w.count;
5690         }
5691 
5692         return -1;
5693 }
5694 
5695 static int inet6_rtm_valid_getroute_req(struct sk_buff *skb,
5696                                         const struct nlmsghdr *nlh,
5697                                         struct nlattr **tb,
5698                                         struct netlink_ext_ack *extack)
5699 {
5700         struct rtmsg *rtm;
5701         int i, err;
5702 
5703         if (nlh->nlmsg_len < nlmsg_msg_size(sizeof(*rtm))) {
5704                 NL_SET_ERR_MSG_MOD(extack,
5705                                    "Invalid header for get route request");
5706                 return -EINVAL;
5707         }
5708 
5709         if (!netlink_strict_get_check(skb))
5710                 return nlmsg_parse_deprecated(nlh, sizeof(*rtm), tb, RTA_MAX,
5711                                               rtm_ipv6_policy, extack);
5712 
5713         rtm = nlmsg_data(nlh);
5714         if ((rtm->rtm_src_len && rtm->rtm_src_len != 128) ||
5715             (rtm->rtm_dst_len && rtm->rtm_dst_len != 128) ||
5716             rtm->rtm_table || rtm->rtm_protocol || rtm->rtm_scope ||
5717             rtm->rtm_type) {
5718                 NL_SET_ERR_MSG_MOD(extack, "Invalid values in header for get route request");
5719                 return -EINVAL;
5720         }
5721         if (rtm->rtm_flags & ~RTM_F_FIB_MATCH) {
5722                 NL_SET_ERR_MSG_MOD(extack,
5723                                    "Invalid flags for get route request");
5724                 return -EINVAL;
5725         }
5726 
5727         err = nlmsg_parse_deprecated_strict(nlh, sizeof(*rtm), tb, RTA_MAX,
5728                                             rtm_ipv6_policy, extack);
5729         if (err)
5730                 return err;
5731 
5732         if ((tb[RTA_SRC] && !rtm->rtm_src_len) ||
5733             (tb[RTA_DST] && !rtm->rtm_dst_len)) {
5734                 NL_SET_ERR_MSG_MOD(extack, "rtm_src_len and rtm_dst_len must be 128 for IPv6");
5735                 return -EINVAL;
5736         }
5737 
5738         for (i = 0; i <= RTA_MAX; i++) {
5739                 if (!tb[i])
5740                         continue;
5741 
5742                 switch (i) {
5743                 case RTA_SRC:
5744                 case RTA_DST:
5745                 case RTA_IIF:
5746                 case RTA_OIF:
5747                 case RTA_MARK:
5748                 case RTA_UID:
5749                 case RTA_SPORT:
5750                 case RTA_DPORT:
5751                 case RTA_IP_PROTO:
5752                         break;
5753                 default:
5754                         NL_SET_ERR_MSG_MOD(extack, "Unsupported attribute in get route request");
5755                         return -EINVAL;
5756                 }
5757         }
5758 
5759         return 0;
5760 }
5761 
5762 static int inet6_rtm_getroute(struct sk_buff *in_skb, struct nlmsghdr *nlh,
5763                               struct netlink_ext_ack *extack)
5764 {
5765         struct net *net = sock_net(in_skb->sk);
5766         struct nlattr *tb[RTA_MAX+1];
5767         int err, iif = 0, oif = 0;
5768         struct fib6_info *from;
5769         struct dst_entry *dst;
5770         struct rt6_info *rt;
5771         struct sk_buff *skb;
5772         struct rtmsg *rtm;
5773         struct flowi6 fl6 = {};
5774         bool fibmatch;
5775 
5776         err = inet6_rtm_valid_getroute_req(in_skb, nlh, tb, extack);
5777         if (err < 0)
5778                 goto errout;
5779 
5780         err = -EINVAL;
5781         rtm = nlmsg_data(nlh);
5782         fl6.flowlabel = ip6_make_flowinfo(rtm->rtm_tos, 0);
5783         fibmatch = !!(rtm->rtm_flags & RTM_F_FIB_MATCH);
5784 
5785         if (tb[RTA_SRC]) {
5786                 if (nla_len(tb[RTA_SRC]) < sizeof(struct in6_addr))
5787                         goto errout;
5788 
5789                 fl6.saddr = *(struct in6_addr *)nla_data(tb[RTA_SRC]);
5790         }
5791 
5792         if (tb[RTA_DST]) {
5793                 if (nla_len(tb[RTA_DST]) < sizeof(struct in6_addr))
5794                         goto errout;
5795 
5796                 fl6.daddr = *(struct in6_addr *)nla_data(tb[RTA_DST]);
5797         }
5798 
5799         if (tb[RTA_IIF])
5800                 iif = nla_get_u32(tb[RTA_IIF]);
5801 
5802         if (tb[RTA_OIF])
5803                 oif = nla_get_u32(tb[RTA_OIF]);
5804 
5805         if (tb[RTA_MARK])
5806                 fl6.flowi6_mark = nla_get_u32(tb[RTA_MARK]);
5807 
5808         if (tb[RTA_UID])
5809                 fl6.flowi6_uid = make_kuid(current_user_ns(),
5810                                            nla_get_u32(tb[RTA_UID]));
5811         else
5812                 fl6.flowi6_uid = iif ? INVALID_UID : current_uid();
5813 
5814         if (tb[RTA_SPORT])
5815                 fl6.fl6_sport = nla_get_be16(tb[RTA_SPORT]);
5816 
5817         if (tb[RTA_DPORT])
5818                 fl6.fl6_dport = nla_get_be16(tb[RTA_DPORT]);
5819 
5820         if (tb[RTA_IP_PROTO]) {
5821                 err = rtm_getroute_parse_ip_proto(tb[RTA_IP_PROTO],
5822                                                   &fl6.flowi6_proto, AF_INET6,
5823                                                   extack);
5824                 if (err)
5825                         goto errout;
5826         }
5827 
5828         if (iif) {
5829                 struct net_device *dev;
5830                 int flags = 0;
5831 
5832                 rcu_read_lock();
5833 
5834                 dev = dev_get_by_index_rcu(net, iif);
5835                 if (!dev) {
5836                         rcu_read_unlock();
5837                         err = -ENODEV;
5838                         goto errout;
5839                 }
5840 
5841                 fl6.flowi6_iif = iif;
5842 
5843                 if (!ipv6_addr_any(&fl6.saddr))
5844                         flags |= RT6_LOOKUP_F_HAS_SADDR;
5845 
5846                 dst = ip6_route_input_lookup(net, dev, &fl6, NULL, flags);
5847 
5848                 rcu_read_unlock();
5849         } else {
5850                 fl6.flowi6_oif = oif;
5851 
5852                 dst = ip6_route_output(net, NULL, &fl6);
5853         }
5854 
5855 
5856         rt = container_of(dst, struct rt6_info, dst);
5857         if (rt->dst.error) {
5858                 err = rt->dst.error;
5859                 ip6_rt_put(rt);
5860                 goto errout;
5861         }
5862 
5863         if (rt == net->ipv6.ip6_null_entry) {
5864                 err = rt->dst.error;
5865                 ip6_rt_put(rt);
5866                 goto errout;
5867         }
5868 
5869         skb = alloc_skb(NLMSG_GOODSIZE, GFP_KERNEL);
5870         if (!skb) {
5871                 ip6_rt_put(rt);
5872                 err = -ENOBUFS;
5873                 goto errout;
5874         }
5875 
5876         skb_dst_set(skb, &rt->dst);
5877 
5878         rcu_read_lock();
5879         from = rcu_dereference(rt->from);
5880         if (from) {
5881                 if (fibmatch)
5882                         err = rt6_fill_node(net, skb, from, NULL, NULL, NULL,
5883                                             iif, RTM_NEWROUTE,
5884                                             NETLINK_CB(in_skb).portid,
5885                                             nlh->nlmsg_seq, 0);
5886                 else
5887                         err = rt6_fill_node(net, skb, from, dst, &fl6.daddr,
5888                                             &fl6.saddr, iif, RTM_NEWROUTE,
5889                                             NETLINK_CB(in_skb).portid,
5890                                             nlh->nlmsg_seq, 0);
5891         } else {
5892                 err = -ENETUNREACH;
5893         }
5894         rcu_read_unlock();
5895 
5896         if (err < 0) {
5897                 kfree_skb(skb);
5898                 goto errout;
5899         }
5900 
5901         err = rtnl_unicast(skb, net, NETLINK_CB(in_skb).portid);
5902 errout:
5903         return err;
5904 }
5905 
5906 void inet6_rt_notify(int event, struct fib6_info *rt, struct nl_info *info,
5907                      unsigned int nlm_flags)
5908 {
5909         struct sk_buff *skb;
5910         struct net *net = info->nl_net;
5911         u32 seq;
5912         int err;
5913 
5914         err = -ENOBUFS;
5915         seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5916 
5917         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5918         if (!skb)
5919                 goto errout;
5920 
5921         err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5922                             event, info->portid, seq, nlm_flags);
5923         if (err < 0) {
5924                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5925                 WARN_ON(err == -EMSGSIZE);
5926                 kfree_skb(skb);
5927                 goto errout;
5928         }
5929         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5930                     info->nlh, gfp_any());
5931         return;
5932 errout:
5933         if (err < 0)
5934                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5935 }
5936 
5937 void fib6_rt_update(struct net *net, struct fib6_info *rt,
5938                     struct nl_info *info)
5939 {
5940         u32 seq = info->nlh ? info->nlh->nlmsg_seq : 0;
5941         struct sk_buff *skb;
5942         int err = -ENOBUFS;
5943 
5944         /* call_fib6_entry_notifiers will be removed when in-kernel notifier
5945          * is implemented and supported for nexthop objects
5946          */
5947         call_fib6_entry_notifiers(net, FIB_EVENT_ENTRY_REPLACE, rt, NULL);
5948 
5949         skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
5950         if (!skb)
5951                 goto errout;
5952 
5953         err = rt6_fill_node(net, skb, rt, NULL, NULL, NULL, 0,
5954                             RTM_NEWROUTE, info->portid, seq, NLM_F_REPLACE);
5955         if (err < 0) {
5956                 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
5957                 WARN_ON(err == -EMSGSIZE);
5958                 kfree_skb(skb);
5959                 goto errout;
5960         }
5961         rtnl_notify(skb, net, info->portid, RTNLGRP_IPV6_ROUTE,
5962                     info->nlh, gfp_any());
5963         return;
5964 errout:
5965         if (err < 0)
5966                 rtnl_set_sk_err(net, RTNLGRP_IPV6_ROUTE, err);
5967 }
5968 
5969 static int ip6_route_dev_notify(struct notifier_block *this,
5970                                 unsigned long event, void *ptr)
5971 {
5972         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
5973         struct net *net = dev_net(dev);
5974 
5975         if (!(dev->flags & IFF_LOOPBACK))
5976                 return NOTIFY_OK;
5977 
5978         if (event == NETDEV_REGISTER) {
5979                 net->ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = dev;
5980                 net->ipv6.ip6_null_entry->dst.dev = dev;
5981                 net->ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(dev);
5982 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5983                 net->ipv6.ip6_prohibit_entry->dst.dev = dev;
5984                 net->ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(dev);
5985                 net->ipv6.ip6_blk_hole_entry->dst.dev = dev;
5986                 net->ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(dev);
5987 #endif
5988          } else if (event == NETDEV_UNREGISTER &&
5989                     dev->reg_state != NETREG_UNREGISTERED) {
5990                 /* NETDEV_UNREGISTER could be fired for multiple times by
5991                  * netdev_wait_allrefs(). Make sure we only call this once.
5992                  */
5993                 in6_dev_put_clear(&net->ipv6.ip6_null_entry->rt6i_idev);
5994 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
5995                 in6_dev_put_clear(&net->ipv6.ip6_prohibit_entry->rt6i_idev);
5996                 in6_dev_put_clear(&net->ipv6.ip6_blk_hole_entry->rt6i_idev);
5997 #endif
5998         }
5999 
6000         return NOTIFY_OK;
6001 }
6002 
6003 /*
6004  *      /proc
6005  */
6006 
6007 #ifdef CONFIG_PROC_FS
6008 static int rt6_stats_seq_show(struct seq_file *seq, void *v)
6009 {
6010         struct net *net = (struct net *)seq->private;
6011         seq_printf(seq, "%04x %04x %04x %04x %04x %04x %04x\n",
6012                    net->ipv6.rt6_stats->fib_nodes,
6013                    net->ipv6.rt6_stats->fib_route_nodes,
6014                    atomic_read(&net->ipv6.rt6_stats->fib_rt_alloc),
6015                    net->ipv6.rt6_stats->fib_rt_entries,
6016                    net->ipv6.rt6_stats->fib_rt_cache,
6017                    dst_entries_get_slow(&net->ipv6.ip6_dst_ops),
6018                    net->ipv6.rt6_stats->fib_discarded_routes);
6019 
6020         return 0;
6021 }
6022 #endif  /* CONFIG_PROC_FS */
6023 
6024 #ifdef CONFIG_SYSCTL
6025 
6026 static
6027 int ipv6_sysctl_rtcache_flush(struct ctl_table *ctl, int write,
6028                               void __user *buffer, size_t *lenp, loff_t *ppos)
6029 {
6030         struct net *net;
6031         int delay;
6032         int ret;
6033         if (!write)
6034                 return -EINVAL;
6035 
6036         net = (struct net *)ctl->extra1;
6037         delay = net->ipv6.sysctl.flush_delay;
6038         ret = proc_dointvec(ctl, write, buffer, lenp, ppos);
6039         if (ret)
6040                 return ret;
6041 
6042         fib6_run_gc(delay <= 0 ? 0 : (unsigned long)delay, net, delay > 0);
6043         return 0;
6044 }
6045 
6046 static struct ctl_table ipv6_route_table_template[] = {
6047         {
6048                 .procname       =       "flush",
6049                 .data           =       &init_net.ipv6.sysctl.flush_delay,
6050                 .maxlen         =       sizeof(int),
6051                 .mode           =       0200,
6052                 .proc_handler   =       ipv6_sysctl_rtcache_flush
6053         },
6054         {
6055                 .procname       =       "gc_thresh",
6056                 .data           =       &ip6_dst_ops_template.gc_thresh,
6057                 .maxlen         =       sizeof(int),
6058                 .mode           =       0644,
6059                 .proc_handler   =       proc_dointvec,
6060         },
6061         {
6062                 .procname       =       "max_size",
6063                 .data           =       &init_net.ipv6.sysctl.ip6_rt_max_size,
6064                 .maxlen         =       sizeof(int),
6065                 .mode           =       0644,
6066                 .proc_handler   =       proc_dointvec,
6067         },
6068         {
6069                 .procname       =       "gc_min_interval",
6070                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6071                 .maxlen         =       sizeof(int),
6072                 .mode           =       0644,
6073                 .proc_handler   =       proc_dointvec_jiffies,
6074         },
6075         {
6076                 .procname       =       "gc_timeout",
6077                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_timeout,
6078                 .maxlen         =       sizeof(int),
6079                 .mode           =       0644,
6080                 .proc_handler   =       proc_dointvec_jiffies,
6081         },
6082         {
6083                 .procname       =       "gc_interval",
6084                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_interval,
6085                 .maxlen         =       sizeof(int),
6086                 .mode           =       0644,
6087                 .proc_handler   =       proc_dointvec_jiffies,
6088         },
6089         {
6090                 .procname       =       "gc_elasticity",
6091                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_elasticity,
6092                 .maxlen         =       sizeof(int),
6093                 .mode           =       0644,
6094                 .proc_handler   =       proc_dointvec,
6095         },
6096         {
6097                 .procname       =       "mtu_expires",
6098                 .data           =       &init_net.ipv6.sysctl.ip6_rt_mtu_expires,
6099                 .maxlen         =       sizeof(int),
6100                 .mode           =       0644,
6101                 .proc_handler   =       proc_dointvec_jiffies,
6102         },
6103         {
6104                 .procname       =       "min_adv_mss",
6105                 .data           =       &init_net.ipv6.sysctl.ip6_rt_min_advmss,
6106                 .maxlen         =       sizeof(int),
6107                 .mode           =       0644,
6108                 .proc_handler   =       proc_dointvec,
6109         },
6110         {
6111                 .procname       =       "gc_min_interval_ms",
6112                 .data           =       &init_net.ipv6.sysctl.ip6_rt_gc_min_interval,
6113                 .maxlen         =       sizeof(int),
6114                 .mode           =       0644,
6115                 .proc_handler   =       proc_dointvec_ms_jiffies,
6116         },
6117         {
6118                 .procname       =       "skip_notify_on_dev_down",
6119                 .data           =       &init_net.ipv6.sysctl.skip_notify_on_dev_down,
6120                 .maxlen         =       sizeof(int),
6121                 .mode           =       0644,
6122                 .proc_handler   =       proc_dointvec_minmax,
6123                 .extra1         =       SYSCTL_ZERO,
6124                 .extra2         =       SYSCTL_ONE,
6125         },
6126         { }
6127 };
6128 
6129 struct ctl_table * __net_init ipv6_route_sysctl_init(struct net *net)
6130 {
6131         struct ctl_table *table;
6132 
6133         table = kmemdup(ipv6_route_table_template,
6134                         sizeof(ipv6_route_table_template),
6135                         GFP_KERNEL);
6136 
6137         if (table) {
6138                 table[0].data = &net->ipv6.sysctl.flush_delay;
6139                 table[0].extra1 = net;
6140                 table[1].data = &net->ipv6.ip6_dst_ops.gc_thresh;
6141                 table[2].data = &net->ipv6.sysctl.ip6_rt_max_size;
6142                 table[3].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6143                 table[4].data = &net->ipv6.sysctl.ip6_rt_gc_timeout;
6144                 table[5].data = &net->ipv6.sysctl.ip6_rt_gc_interval;
6145                 table[6].data = &net->ipv6.sysctl.ip6_rt_gc_elasticity;
6146                 table[7].data = &net->ipv6.sysctl.ip6_rt_mtu_expires;
6147                 table[8].data = &net->ipv6.sysctl.ip6_rt_min_advmss;
6148                 table[9].data = &net->ipv6.sysctl.ip6_rt_gc_min_interval;
6149                 table[10].data = &net->ipv6.sysctl.skip_notify_on_dev_down;
6150 
6151                 /* Don't export sysctls to unprivileged users */
6152                 if (net->user_ns != &init_user_ns)
6153                         table[0].procname = NULL;
6154         }
6155 
6156         return table;
6157 }
6158 #endif
6159 
6160 static int __net_init ip6_route_net_init(struct net *net)
6161 {
6162         int ret = -ENOMEM;
6163 
6164         memcpy(&net->ipv6.ip6_dst_ops, &ip6_dst_ops_template,
6165                sizeof(net->ipv6.ip6_dst_ops));
6166 
6167         if (dst_entries_init(&net->ipv6.ip6_dst_ops) < 0)
6168                 goto out_ip6_dst_ops;
6169 
6170         net->ipv6.fib6_null_entry = fib6_info_alloc(GFP_KERNEL, true);
6171         if (!net->ipv6.fib6_null_entry)
6172                 goto out_ip6_dst_entries;
6173         memcpy(net->ipv6.fib6_null_entry, &fib6_null_entry_template,
6174                sizeof(*net->ipv6.fib6_null_entry));
6175 
6176         net->ipv6.ip6_null_entry = kmemdup(&ip6_null_entry_template,
6177                                            sizeof(*net->ipv6.ip6_null_entry),
6178                                            GFP_KERNEL);
6179         if (!net->ipv6.ip6_null_entry)
6180                 goto out_fib6_null_entry;
6181         net->ipv6.ip6_null_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6182         dst_init_metrics(&net->ipv6.ip6_null_entry->dst,
6183                          ip6_template_metrics, true);
6184         INIT_LIST_HEAD(&net->ipv6.ip6_null_entry->rt6i_uncached);
6185 
6186 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6187         net->ipv6.fib6_has_custom_rules = false;
6188         net->ipv6.ip6_prohibit_entry = kmemdup(&ip6_prohibit_entry_template,
6189                                                sizeof(*net->ipv6.ip6_prohibit_entry),
6190                                                GFP_KERNEL);
6191         if (!net->ipv6.ip6_prohibit_entry)
6192                 goto out_ip6_null_entry;
6193         net->ipv6.ip6_prohibit_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6194         dst_init_metrics(&net->ipv6.ip6_prohibit_entry->dst,
6195                          ip6_template_metrics, true);
6196         INIT_LIST_HEAD(&net->ipv6.ip6_prohibit_entry->rt6i_uncached);
6197 
6198         net->ipv6.ip6_blk_hole_entry = kmemdup(&ip6_blk_hole_entry_template,
6199                                                sizeof(*net->ipv6.ip6_blk_hole_entry),
6200                                                GFP_KERNEL);
6201         if (!net->ipv6.ip6_blk_hole_entry)
6202                 goto out_ip6_prohibit_entry;
6203         net->ipv6.ip6_blk_hole_entry->dst.ops = &net->ipv6.ip6_dst_ops;
6204         dst_init_metrics(&net->ipv6.ip6_blk_hole_entry->dst,
6205                          ip6_template_metrics, true);
6206         INIT_LIST_HEAD(&net->ipv6.ip6_blk_hole_entry->rt6i_uncached);
6207 #endif
6208 
6209         net->ipv6.sysctl.flush_delay = 0;
6210         net->ipv6.sysctl.ip6_rt_max_size = 4096;
6211         net->ipv6.sysctl.ip6_rt_gc_min_interval = HZ / 2;
6212         net->ipv6.sysctl.ip6_rt_gc_timeout = 60*HZ;
6213         net->ipv6.sysctl.ip6_rt_gc_interval = 30*HZ;
6214         net->ipv6.sysctl.ip6_rt_gc_elasticity = 9;
6215         net->ipv6.sysctl.ip6_rt_mtu_expires = 10*60*HZ;
6216         net->ipv6.sysctl.ip6_rt_min_advmss = IPV6_MIN_MTU - 20 - 40;
6217         net->ipv6.sysctl.skip_notify_on_dev_down = 0;
6218 
6219         net->ipv6.ip6_rt_gc_expire = 30*HZ;
6220 
6221         ret = 0;
6222 out:
6223         return ret;
6224 
6225 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6226 out_ip6_prohibit_entry:
6227         kfree(net->ipv6.ip6_prohibit_entry);
6228 out_ip6_null_entry:
6229         kfree(net->ipv6.ip6_null_entry);
6230 #endif
6231 out_fib6_null_entry:
6232         kfree(net->ipv6.fib6_null_entry);
6233 out_ip6_dst_entries:
6234         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6235 out_ip6_dst_ops:
6236         goto out;
6237 }
6238 
6239 static void __net_exit ip6_route_net_exit(struct net *net)
6240 {
6241         kfree(net->ipv6.fib6_null_entry);
6242         kfree(net->ipv6.ip6_null_entry);
6243 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6244         kfree(net->ipv6.ip6_prohibit_entry);
6245         kfree(net->ipv6.ip6_blk_hole_entry);
6246 #endif
6247         dst_entries_destroy(&net->ipv6.ip6_dst_ops);
6248 }
6249 
6250 static int __net_init ip6_route_net_init_late(struct net *net)
6251 {
6252 #ifdef CONFIG_PROC_FS
6253         proc_create_net("ipv6_route", 0, net->proc_net, &ipv6_route_seq_ops,
6254                         sizeof(struct ipv6_route_iter));
6255         proc_create_net_single("rt6_stats", 0444, net->proc_net,
6256                         rt6_stats_seq_show, NULL);
6257 #endif
6258         return 0;
6259 }
6260 
6261 static void __net_exit ip6_route_net_exit_late(struct net *net)
6262 {
6263 #ifdef CONFIG_PROC_FS
6264         remove_proc_entry("ipv6_route", net->proc_net);
6265         remove_proc_entry("rt6_stats", net->proc_net);
6266 #endif
6267 }
6268 
6269 static struct pernet_operations ip6_route_net_ops = {
6270         .init = ip6_route_net_init,
6271         .exit = ip6_route_net_exit,
6272 };
6273 
6274 static int __net_init ipv6_inetpeer_init(struct net *net)
6275 {
6276         struct inet_peer_base *bp = kmalloc(sizeof(*bp), GFP_KERNEL);
6277 
6278         if (!bp)
6279                 return -ENOMEM;
6280         inet_peer_base_init(bp);
6281         net->ipv6.peers = bp;
6282         return 0;
6283 }
6284 
6285 static void __net_exit ipv6_inetpeer_exit(struct net *net)
6286 {
6287         struct inet_peer_base *bp = net->ipv6.peers;
6288 
6289         net->ipv6.peers = NULL;
6290         inetpeer_invalidate_tree(bp);
6291         kfree(bp);
6292 }
6293 
6294 static struct pernet_operations ipv6_inetpeer_ops = {
6295         .init   =       ipv6_inetpeer_init,
6296         .exit   =       ipv6_inetpeer_exit,
6297 };
6298 
6299 static struct pernet_operations ip6_route_net_late_ops = {
6300         .init = ip6_route_net_init_late,
6301         .exit = ip6_route_net_exit_late,
6302 };
6303 
6304 static struct notifier_block ip6_route_dev_notifier = {
6305         .notifier_call = ip6_route_dev_notify,
6306         .priority = ADDRCONF_NOTIFY_PRIORITY - 10,
6307 };
6308 
6309 void __init ip6_route_init_special_entries(void)
6310 {
6311         /* Registering of the loopback is done before this portion of code,
6312          * the loopback reference in rt6_info will not be taken, do it
6313          * manually for init_net */
6314         init_net.ipv6.fib6_null_entry->fib6_nh->fib_nh_dev = init_net.loopback_dev;
6315         init_net.ipv6.ip6_null_entry->dst.dev = init_net.loopback_dev;
6316         init_net.ipv6.ip6_null_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6317   #ifdef CONFIG_IPV6_MULTIPLE_TABLES
6318         init_net.ipv6.ip6_prohibit_entry->dst.dev = init_net.loopback_dev;
6319         init_net.ipv6.ip6_prohibit_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6320         init_net.ipv6.ip6_blk_hole_entry->dst.dev = init_net.loopback_dev;
6321         init_net.ipv6.ip6_blk_hole_entry->rt6i_idev = in6_dev_get(init_net.loopback_dev);
6322   #endif
6323 }
6324 
6325 int __init ip6_route_init(void)
6326 {
6327         int ret;
6328         int cpu;
6329 
6330         ret = -ENOMEM;
6331         ip6_dst_ops_template.kmem_cachep =
6332                 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info), 0,
6333                                   SLAB_HWCACHE_ALIGN, NULL);
6334         if (!ip6_dst_ops_template.kmem_cachep)
6335                 goto out;
6336 
6337         ret = dst_entries_init(&ip6_dst_blackhole_ops);
6338         if (ret)
6339                 goto out_kmem_cache;
6340 
6341         ret = register_pernet_subsys(&ipv6_inetpeer_ops);
6342         if (ret)
6343                 goto out_dst_entries;
6344 
6345         ret = register_pernet_subsys(&ip6_route_net_ops);
6346         if (ret)
6347                 goto out_register_inetpeer;
6348 
6349         ip6_dst_blackhole_ops.kmem_cachep = ip6_dst_ops_template.kmem_cachep;
6350 
6351         ret = fib6_init();
6352         if (ret)
6353                 goto out_register_subsys;
6354 
6355         ret = xfrm6_init();
6356         if (ret)
6357                 goto out_fib6_init;
6358 
6359         ret = fib6_rules_init();
6360         if (ret)
6361                 goto xfrm6_init;
6362 
6363         ret = register_pernet_subsys(&ip6_route_net_late_ops);
6364         if (ret)
6365                 goto fib6_rules_init;
6366 
6367         ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_NEWROUTE,
6368                                    inet6_rtm_newroute, NULL, 0);
6369         if (ret < 0)
6370                 goto out_register_late_subsys;
6371 
6372         ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_DELROUTE,
6373                                    inet6_rtm_delroute, NULL, 0);
6374         if (ret < 0)
6375                 goto out_register_late_subsys;
6376 
6377         ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE,
6378                                    inet6_rtm_getroute, NULL,
6379                                    RTNL_FLAG_DOIT_UNLOCKED);
6380         if (ret < 0)
6381                 goto out_register_late_subsys;
6382 
6383         ret = register_netdevice_notifier(&ip6_route_dev_notifier);
6384         if (ret)
6385                 goto out_register_late_subsys;
6386 
6387         for_each_possible_cpu(cpu) {
6388                 struct uncached_list *ul = per_cpu_ptr(&rt6_uncached_list, cpu);
6389 
6390                 INIT_LIST_HEAD(&ul->head);
6391                 spin_lock_init(&ul->lock);
6392         }
6393 
6394 out:
6395         return ret;
6396 
6397 out_register_late_subsys:
6398         rtnl_unregister_all(PF_INET6);
6399         unregister_pernet_subsys(&ip6_route_net_late_ops);
6400 fib6_rules_init:
6401         fib6_rules_cleanup();
6402 xfrm6_init:
6403         xfrm6_fini();
6404 out_fib6_init:
6405         fib6_gc_cleanup();
6406 out_register_subsys:
6407         unregister_pernet_subsys(&ip6_route_net_ops);
6408 out_register_inetpeer:
6409         unregister_pernet_subsys(&ipv6_inetpeer_ops);
6410 out_dst_entries:
6411         dst_entries_destroy(&ip6_dst_blackhole_ops);
6412 out_kmem_cache:
6413         kmem_cache_destroy(ip6_dst_ops_template.kmem_cachep);
6414         goto out;
6415 }
6416 
6417 void ip6_route_cleanup(void)
6418 {
6419         unregister_netdevice_notifier(&ip6_route_dev_notifier);
6420         unregister_pernet_subsys(&ip6_route_net_late_ops);
6421         fib6_rules_cleanup();
6422         xfrm6_fini();
6423         fib6_gc_cleanup();
6424         unregister_pernet_subsys(&ipv6_inetpeer_ops);
6425         unregister_pernet_subsys(&ip6_route_net_ops);
6426         dst_entries_destroy(&ip6_dst_blackhole_ops);
6427         kmem_cache_destroy(