2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <linux/jhash.h>
48 #include <net/net_namespace.h>
51 #include <net/ip6_fib.h>
52 #include <net/ip6_route.h>
53 #include <net/ndisc.h>
54 #include <net/addrconf.h>
56 #include <linux/rtnetlink.h>
58 #include <net/dst_metadata.h>
60 #include <net/netevent.h>
61 #include <net/netlink.h>
62 #include <net/nexthop.h>
63 #include <net/lwtunnel.h>
64 #include <net/ip_tunnels.h>
65 #include <net/l3mdev.h>
66 #include <trace/events/fib6.h>
68 #include <linux/uaccess.h>
71 #include <linux/sysctl.h>
75 RT6_NUD_FAIL_HARD
= -3,
76 RT6_NUD_FAIL_PROBE
= -2,
77 RT6_NUD_FAIL_DO_RR
= -1,
81 static void ip6_rt_copy_init(struct rt6_info
*rt
, struct rt6_info
*ort
);
82 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
83 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
84 static unsigned int ip6_mtu(const struct dst_entry
*dst
);
85 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
86 static void ip6_dst_destroy(struct dst_entry
*);
87 static void ip6_dst_ifdown(struct dst_entry
*,
88 struct net_device
*dev
, int how
);
89 static int ip6_dst_gc(struct dst_ops
*ops
);
91 static int ip6_pkt_discard(struct sk_buff
*skb
);
92 static int ip6_pkt_discard_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
93 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
94 static int ip6_pkt_prohibit_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
95 static void ip6_link_failure(struct sk_buff
*skb
);
96 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
97 struct sk_buff
*skb
, u32 mtu
);
98 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
100 static void rt6_dst_from_metrics_check(struct rt6_info
*rt
);
101 static int rt6_score_route(struct rt6_info
*rt
, int oif
, int strict
);
102 static size_t rt6_nlmsg_size(struct rt6_info
*rt
);
103 static int rt6_fill_node(struct net
*net
,
104 struct sk_buff
*skb
, struct rt6_info
*rt
,
105 struct in6_addr
*dst
, struct in6_addr
*src
,
106 int iif
, int type
, u32 portid
, u32 seq
,
108 static struct rt6_info
*rt6_find_cached_rt(struct rt6_info
*rt
,
109 struct in6_addr
*daddr
,
110 struct in6_addr
*saddr
);
112 #ifdef CONFIG_IPV6_ROUTE_INFO
113 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
114 const struct in6_addr
*prefix
, int prefixlen
,
115 const struct in6_addr
*gwaddr
,
116 struct net_device
*dev
,
118 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
119 const struct in6_addr
*prefix
, int prefixlen
,
120 const struct in6_addr
*gwaddr
,
121 struct net_device
*dev
);
124 struct uncached_list
{
126 struct list_head head
;
129 static DEFINE_PER_CPU_ALIGNED(struct uncached_list
, rt6_uncached_list
);
131 static void rt6_uncached_list_add(struct rt6_info
*rt
)
133 struct uncached_list
*ul
= raw_cpu_ptr(&rt6_uncached_list
);
135 rt
->rt6i_uncached_list
= ul
;
137 spin_lock_bh(&ul
->lock
);
138 list_add_tail(&rt
->rt6i_uncached
, &ul
->head
);
139 spin_unlock_bh(&ul
->lock
);
142 static void rt6_uncached_list_del(struct rt6_info
*rt
)
144 if (!list_empty(&rt
->rt6i_uncached
)) {
145 struct uncached_list
*ul
= rt
->rt6i_uncached_list
;
146 struct net
*net
= dev_net(rt
->dst
.dev
);
148 spin_lock_bh(&ul
->lock
);
149 list_del(&rt
->rt6i_uncached
);
150 atomic_dec(&net
->ipv6
.rt6_stats
->fib_rt_uncache
);
151 spin_unlock_bh(&ul
->lock
);
155 static void rt6_uncached_list_flush_dev(struct net
*net
, struct net_device
*dev
)
157 struct net_device
*loopback_dev
= net
->loopback_dev
;
160 if (dev
== loopback_dev
)
163 for_each_possible_cpu(cpu
) {
164 struct uncached_list
*ul
= per_cpu_ptr(&rt6_uncached_list
, cpu
);
167 spin_lock_bh(&ul
->lock
);
168 list_for_each_entry(rt
, &ul
->head
, rt6i_uncached
) {
169 struct inet6_dev
*rt_idev
= rt
->rt6i_idev
;
170 struct net_device
*rt_dev
= rt
->dst
.dev
;
172 if (rt_idev
->dev
== dev
) {
173 rt
->rt6i_idev
= in6_dev_get(loopback_dev
);
174 in6_dev_put(rt_idev
);
178 rt
->dst
.dev
= loopback_dev
;
179 dev_hold(rt
->dst
.dev
);
183 spin_unlock_bh(&ul
->lock
);
187 static u32
*rt6_pcpu_cow_metrics(struct rt6_info
*rt
)
189 return dst_metrics_write_ptr(rt
->dst
.from
);
192 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
194 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
196 if (rt
->rt6i_flags
& RTF_PCPU
)
197 return rt6_pcpu_cow_metrics(rt
);
198 else if (rt
->rt6i_flags
& RTF_CACHE
)
201 return dst_cow_metrics_generic(dst
, old
);
204 static inline const void *choose_neigh_daddr(struct rt6_info
*rt
,
208 struct in6_addr
*p
= &rt
->rt6i_gateway
;
210 if (!ipv6_addr_any(p
))
211 return (const void *) p
;
213 return &ipv6_hdr(skb
)->daddr
;
217 static struct neighbour
*ip6_neigh_lookup(const struct dst_entry
*dst
,
221 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
224 daddr
= choose_neigh_daddr(rt
, skb
, daddr
);
225 n
= __ipv6_neigh_lookup(dst
->dev
, daddr
);
228 return neigh_create(&nd_tbl
, daddr
, dst
->dev
);
231 static void ip6_confirm_neigh(const struct dst_entry
*dst
, const void *daddr
)
233 struct net_device
*dev
= dst
->dev
;
234 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
236 daddr
= choose_neigh_daddr(rt
, NULL
, daddr
);
239 if (dev
->flags
& (IFF_NOARP
| IFF_LOOPBACK
))
241 if (ipv6_addr_is_multicast((const struct in6_addr
*)daddr
))
243 __ipv6_confirm_neigh(dev
, daddr
);
246 static struct dst_ops ip6_dst_ops_template
= {
250 .check
= ip6_dst_check
,
251 .default_advmss
= ip6_default_advmss
,
253 .cow_metrics
= ipv6_cow_metrics
,
254 .destroy
= ip6_dst_destroy
,
255 .ifdown
= ip6_dst_ifdown
,
256 .negative_advice
= ip6_negative_advice
,
257 .link_failure
= ip6_link_failure
,
258 .update_pmtu
= ip6_rt_update_pmtu
,
259 .redirect
= rt6_do_redirect
,
260 .local_out
= __ip6_local_out
,
261 .neigh_lookup
= ip6_neigh_lookup
,
262 .confirm_neigh
= ip6_confirm_neigh
,
265 static unsigned int ip6_blackhole_mtu(const struct dst_entry
*dst
)
267 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
269 return mtu
? : dst
->dev
->mtu
;
272 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
273 struct sk_buff
*skb
, u32 mtu
)
277 static void ip6_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
282 static struct dst_ops ip6_dst_blackhole_ops
= {
284 .destroy
= ip6_dst_destroy
,
285 .check
= ip6_dst_check
,
286 .mtu
= ip6_blackhole_mtu
,
287 .default_advmss
= ip6_default_advmss
,
288 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
289 .redirect
= ip6_rt_blackhole_redirect
,
290 .cow_metrics
= dst_cow_metrics_generic
,
291 .neigh_lookup
= ip6_neigh_lookup
,
294 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
295 [RTAX_HOPLIMIT
- 1] = 0,
298 static const struct rt6_info ip6_null_entry_template
= {
300 .__refcnt
= ATOMIC_INIT(1),
302 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
303 .error
= -ENETUNREACH
,
304 .input
= ip6_pkt_discard
,
305 .output
= ip6_pkt_discard_out
,
307 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
308 .rt6i_protocol
= RTPROT_KERNEL
,
309 .rt6i_metric
= ~(u32
) 0,
310 .rt6i_ref
= ATOMIC_INIT(1),
313 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
315 static const struct rt6_info ip6_prohibit_entry_template
= {
317 .__refcnt
= ATOMIC_INIT(1),
319 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
321 .input
= ip6_pkt_prohibit
,
322 .output
= ip6_pkt_prohibit_out
,
324 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
325 .rt6i_protocol
= RTPROT_KERNEL
,
326 .rt6i_metric
= ~(u32
) 0,
327 .rt6i_ref
= ATOMIC_INIT(1),
330 static const struct rt6_info ip6_blk_hole_entry_template
= {
332 .__refcnt
= ATOMIC_INIT(1),
334 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
336 .input
= dst_discard
,
337 .output
= dst_discard_out
,
339 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
340 .rt6i_protocol
= RTPROT_KERNEL
,
341 .rt6i_metric
= ~(u32
) 0,
342 .rt6i_ref
= ATOMIC_INIT(1),
347 static void rt6_info_init(struct rt6_info
*rt
)
349 struct dst_entry
*dst
= &rt
->dst
;
351 memset(dst
+ 1, 0, sizeof(*rt
) - sizeof(*dst
));
352 INIT_LIST_HEAD(&rt
->rt6i_siblings
);
353 INIT_LIST_HEAD(&rt
->rt6i_uncached
);
356 /* allocate dst with ip6_dst_ops */
357 static struct rt6_info
*__ip6_dst_alloc(struct net
*net
,
358 struct net_device
*dev
,
361 struct rt6_info
*rt
= dst_alloc(&net
->ipv6
.ip6_dst_ops
, dev
,
362 1, DST_OBSOLETE_FORCE_CHK
, flags
);
366 atomic_inc(&net
->ipv6
.rt6_stats
->fib_rt_alloc
);
372 struct rt6_info
*ip6_dst_alloc(struct net
*net
,
373 struct net_device
*dev
,
376 struct rt6_info
*rt
= __ip6_dst_alloc(net
, dev
, flags
);
379 rt
->rt6i_pcpu
= alloc_percpu_gfp(struct rt6_info
*, GFP_ATOMIC
);
380 if (!rt
->rt6i_pcpu
) {
381 dst_release_immediate(&rt
->dst
);
388 EXPORT_SYMBOL(ip6_dst_alloc
);
390 static void ip6_dst_destroy(struct dst_entry
*dst
)
392 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
393 struct rt6_exception_bucket
*bucket
;
394 struct dst_entry
*from
= dst
->from
;
395 struct inet6_dev
*idev
;
397 dst_destroy_metrics_generic(dst
);
398 free_percpu(rt
->rt6i_pcpu
);
399 rt6_uncached_list_del(rt
);
401 idev
= rt
->rt6i_idev
;
403 rt
->rt6i_idev
= NULL
;
406 bucket
= rcu_dereference_protected(rt
->rt6i_exception_bucket
, 1);
408 rt
->rt6i_exception_bucket
= NULL
;
416 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
419 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
420 struct inet6_dev
*idev
= rt
->rt6i_idev
;
421 struct net_device
*loopback_dev
=
422 dev_net(dev
)->loopback_dev
;
424 if (idev
&& idev
->dev
!= loopback_dev
) {
425 struct inet6_dev
*loopback_idev
= in6_dev_get(loopback_dev
);
427 rt
->rt6i_idev
= loopback_idev
;
433 static bool __rt6_check_expired(const struct rt6_info
*rt
)
435 if (rt
->rt6i_flags
& RTF_EXPIRES
)
436 return time_after(jiffies
, rt
->dst
.expires
);
441 static bool rt6_check_expired(const struct rt6_info
*rt
)
443 if (rt
->rt6i_flags
& RTF_EXPIRES
) {
444 if (time_after(jiffies
, rt
->dst
.expires
))
446 } else if (rt
->dst
.from
) {
447 return rt
->dst
.obsolete
!= DST_OBSOLETE_FORCE_CHK
||
448 rt6_check_expired((struct rt6_info
*)rt
->dst
.from
);
453 static struct rt6_info
*rt6_multipath_select(struct rt6_info
*match
,
454 struct flowi6
*fl6
, int oif
,
457 struct rt6_info
*sibling
, *next_sibling
;
460 /* We might have already computed the hash for ICMPv6 errors. In such
461 * case it will always be non-zero. Otherwise now is the time to do it.
464 fl6
->mp_hash
= rt6_multipath_hash(fl6
, NULL
);
466 route_choosen
= fl6
->mp_hash
% (match
->rt6i_nsiblings
+ 1);
467 /* Don't change the route, if route_choosen == 0
468 * (siblings does not include ourself)
471 list_for_each_entry_safe(sibling
, next_sibling
,
472 &match
->rt6i_siblings
, rt6i_siblings
) {
474 if (route_choosen
== 0) {
475 struct inet6_dev
*idev
= sibling
->rt6i_idev
;
477 if (!netif_carrier_ok(sibling
->dst
.dev
) &&
478 idev
->cnf
.ignore_routes_with_linkdown
)
480 if (rt6_score_route(sibling
, oif
, strict
) < 0)
490 * Route lookup. rcu_read_lock() should be held.
493 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
495 const struct in6_addr
*saddr
,
499 struct rt6_info
*local
= NULL
;
500 struct rt6_info
*sprt
;
502 if (!oif
&& ipv6_addr_any(saddr
))
505 for (sprt
= rt
; sprt
; sprt
= rcu_dereference(sprt
->dst
.rt6_next
)) {
506 struct net_device
*dev
= sprt
->dst
.dev
;
509 if (dev
->ifindex
== oif
)
511 if (dev
->flags
& IFF_LOOPBACK
) {
512 if (!sprt
->rt6i_idev
||
513 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
514 if (flags
& RT6_LOOKUP_F_IFACE
)
517 local
->rt6i_idev
->dev
->ifindex
== oif
)
523 if (ipv6_chk_addr(net
, saddr
, dev
,
524 flags
& RT6_LOOKUP_F_IFACE
))
533 if (flags
& RT6_LOOKUP_F_IFACE
)
534 return net
->ipv6
.ip6_null_entry
;
540 #ifdef CONFIG_IPV6_ROUTER_PREF
541 struct __rt6_probe_work
{
542 struct work_struct work
;
543 struct in6_addr target
;
544 struct net_device
*dev
;
547 static void rt6_probe_deferred(struct work_struct
*w
)
549 struct in6_addr mcaddr
;
550 struct __rt6_probe_work
*work
=
551 container_of(w
, struct __rt6_probe_work
, work
);
553 addrconf_addr_solict_mult(&work
->target
, &mcaddr
);
554 ndisc_send_ns(work
->dev
, &work
->target
, &mcaddr
, NULL
, 0);
559 static void rt6_probe(struct rt6_info
*rt
)
561 struct __rt6_probe_work
*work
;
562 struct neighbour
*neigh
;
564 * Okay, this does not seem to be appropriate
565 * for now, however, we need to check if it
566 * is really so; aka Router Reachability Probing.
568 * Router Reachability Probe MUST be rate-limited
569 * to no more than one per minute.
571 if (!rt
|| !(rt
->rt6i_flags
& RTF_GATEWAY
))
574 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
576 if (neigh
->nud_state
& NUD_VALID
)
580 write_lock(&neigh
->lock
);
581 if (!(neigh
->nud_state
& NUD_VALID
) &&
584 rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
585 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
587 __neigh_set_probe_once(neigh
);
589 write_unlock(&neigh
->lock
);
591 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
595 INIT_WORK(&work
->work
, rt6_probe_deferred
);
596 work
->target
= rt
->rt6i_gateway
;
597 dev_hold(rt
->dst
.dev
);
598 work
->dev
= rt
->dst
.dev
;
599 schedule_work(&work
->work
);
603 rcu_read_unlock_bh();
606 static inline void rt6_probe(struct rt6_info
*rt
)
612 * Default Router Selection (RFC 2461 6.3.6)
614 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
616 struct net_device
*dev
= rt
->dst
.dev
;
617 if (!oif
|| dev
->ifindex
== oif
)
619 if ((dev
->flags
& IFF_LOOPBACK
) &&
620 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
625 static inline enum rt6_nud_state
rt6_check_neigh(struct rt6_info
*rt
)
627 struct neighbour
*neigh
;
628 enum rt6_nud_state ret
= RT6_NUD_FAIL_HARD
;
630 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
631 !(rt
->rt6i_flags
& RTF_GATEWAY
))
632 return RT6_NUD_SUCCEED
;
635 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
637 read_lock(&neigh
->lock
);
638 if (neigh
->nud_state
& NUD_VALID
)
639 ret
= RT6_NUD_SUCCEED
;
640 #ifdef CONFIG_IPV6_ROUTER_PREF
641 else if (!(neigh
->nud_state
& NUD_FAILED
))
642 ret
= RT6_NUD_SUCCEED
;
644 ret
= RT6_NUD_FAIL_PROBE
;
646 read_unlock(&neigh
->lock
);
648 ret
= IS_ENABLED(CONFIG_IPV6_ROUTER_PREF
) ?
649 RT6_NUD_SUCCEED
: RT6_NUD_FAIL_DO_RR
;
651 rcu_read_unlock_bh();
656 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
661 m
= rt6_check_dev(rt
, oif
);
662 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
663 return RT6_NUD_FAIL_HARD
;
664 #ifdef CONFIG_IPV6_ROUTER_PREF
665 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
667 if (strict
& RT6_LOOKUP_F_REACHABLE
) {
668 int n
= rt6_check_neigh(rt
);
675 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
676 int *mpri
, struct rt6_info
*match
,
680 bool match_do_rr
= false;
681 struct inet6_dev
*idev
= rt
->rt6i_idev
;
682 struct net_device
*dev
= rt
->dst
.dev
;
684 if (dev
&& !netif_carrier_ok(dev
) &&
685 idev
->cnf
.ignore_routes_with_linkdown
&&
686 !(strict
& RT6_LOOKUP_F_IGNORE_LINKSTATE
))
689 if (rt6_check_expired(rt
))
692 m
= rt6_score_route(rt
, oif
, strict
);
693 if (m
== RT6_NUD_FAIL_DO_RR
) {
695 m
= 0; /* lowest valid score */
696 } else if (m
== RT6_NUD_FAIL_HARD
) {
700 if (strict
& RT6_LOOKUP_F_REACHABLE
)
703 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
705 *do_rr
= match_do_rr
;
713 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
714 struct rt6_info
*leaf
,
715 struct rt6_info
*rr_head
,
716 u32 metric
, int oif
, int strict
,
719 struct rt6_info
*rt
, *match
, *cont
;
724 for (rt
= rr_head
; rt
; rt
= rcu_dereference(rt
->dst
.rt6_next
)) {
725 if (rt
->rt6i_metric
!= metric
) {
730 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
733 for (rt
= leaf
; rt
&& rt
!= rr_head
;
734 rt
= rcu_dereference(rt
->dst
.rt6_next
)) {
735 if (rt
->rt6i_metric
!= metric
) {
740 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
746 for (rt
= cont
; rt
; rt
= rcu_dereference(rt
->dst
.rt6_next
))
747 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
752 static struct rt6_info
*rt6_select(struct net
*net
, struct fib6_node
*fn
,
755 struct rt6_info
*leaf
= rcu_dereference(fn
->leaf
);
756 struct rt6_info
*match
, *rt0
;
760 if (!leaf
|| leaf
== net
->ipv6
.ip6_null_entry
)
761 return net
->ipv6
.ip6_null_entry
;
763 rt0
= rcu_dereference(fn
->rr_ptr
);
767 /* Double check to make sure fn is not an intermediate node
768 * and fn->leaf does not points to its child's leaf
769 * (This might happen if all routes under fn are deleted from
770 * the tree and fib6_repair_tree() is called on the node.)
772 key_plen
= rt0
->rt6i_dst
.plen
;
773 #ifdef CONFIG_IPV6_SUBTREES
774 if (rt0
->rt6i_src
.plen
)
775 key_plen
= rt0
->rt6i_src
.plen
;
777 if (fn
->fn_bit
!= key_plen
)
778 return net
->ipv6
.ip6_null_entry
;
780 match
= find_rr_leaf(fn
, leaf
, rt0
, rt0
->rt6i_metric
, oif
, strict
,
784 struct rt6_info
*next
= rcu_dereference(rt0
->dst
.rt6_next
);
786 /* no entries matched; do round-robin */
787 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
791 spin_lock_bh(&leaf
->rt6i_table
->tb6_lock
);
792 /* make sure next is not being deleted from the tree */
794 rcu_assign_pointer(fn
->rr_ptr
, next
);
795 spin_unlock_bh(&leaf
->rt6i_table
->tb6_lock
);
799 return match
? match
: net
->ipv6
.ip6_null_entry
;
802 static bool rt6_is_gw_or_nonexthop(const struct rt6_info
*rt
)
804 return (rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
));
807 #ifdef CONFIG_IPV6_ROUTE_INFO
808 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
809 const struct in6_addr
*gwaddr
)
811 struct net
*net
= dev_net(dev
);
812 struct route_info
*rinfo
= (struct route_info
*) opt
;
813 struct in6_addr prefix_buf
, *prefix
;
815 unsigned long lifetime
;
818 if (len
< sizeof(struct route_info
)) {
822 /* Sanity check for prefix_len and length */
823 if (rinfo
->length
> 3) {
825 } else if (rinfo
->prefix_len
> 128) {
827 } else if (rinfo
->prefix_len
> 64) {
828 if (rinfo
->length
< 2) {
831 } else if (rinfo
->prefix_len
> 0) {
832 if (rinfo
->length
< 1) {
837 pref
= rinfo
->route_pref
;
838 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
841 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
843 if (rinfo
->length
== 3)
844 prefix
= (struct in6_addr
*)rinfo
->prefix
;
846 /* this function is safe */
847 ipv6_addr_prefix(&prefix_buf
,
848 (struct in6_addr
*)rinfo
->prefix
,
850 prefix
= &prefix_buf
;
853 if (rinfo
->prefix_len
== 0)
854 rt
= rt6_get_dflt_router(gwaddr
, dev
);
856 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
,
859 if (rt
&& !lifetime
) {
865 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
,
868 rt
->rt6i_flags
= RTF_ROUTEINFO
|
869 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
872 if (!addrconf_finite_timeout(lifetime
))
873 rt6_clean_expires(rt
);
875 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
883 static struct fib6_node
* fib6_backtrack(struct fib6_node
*fn
,
884 struct in6_addr
*saddr
)
886 struct fib6_node
*pn
, *sn
;
888 if (fn
->fn_flags
& RTN_TL_ROOT
)
890 pn
= rcu_dereference(fn
->parent
);
891 sn
= FIB6_SUBTREE(pn
);
893 fn
= fib6_lookup(sn
, NULL
, saddr
);
896 if (fn
->fn_flags
& RTN_RTINFO
)
901 static bool ip6_hold_safe(struct net
*net
, struct rt6_info
**prt
,
904 struct rt6_info
*rt
= *prt
;
906 if (dst_hold_safe(&rt
->dst
))
909 rt
= net
->ipv6
.ip6_null_entry
;
918 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
919 struct fib6_table
*table
,
920 struct flowi6
*fl6
, int flags
)
922 struct rt6_info
*rt
, *rt_cache
;
923 struct fib6_node
*fn
;
926 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
928 rt
= rcu_dereference(fn
->leaf
);
930 rt
= net
->ipv6
.ip6_null_entry
;
932 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
,
933 fl6
->flowi6_oif
, flags
);
934 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
935 rt
= rt6_multipath_select(rt
, fl6
,
936 fl6
->flowi6_oif
, flags
);
938 if (rt
== net
->ipv6
.ip6_null_entry
) {
939 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
943 /* Search through exception table */
944 rt_cache
= rt6_find_cached_rt(rt
, &fl6
->daddr
, &fl6
->saddr
);
948 if (ip6_hold_safe(net
, &rt
, true))
949 dst_use_noref(&rt
->dst
, jiffies
);
953 trace_fib6_table_lookup(net
, rt
, table
, fl6
);
959 struct dst_entry
*ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
962 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
964 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
966 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
967 const struct in6_addr
*saddr
, int oif
, int strict
)
969 struct flowi6 fl6
= {
973 struct dst_entry
*dst
;
974 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
977 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
978 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
981 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
983 return (struct rt6_info
*) dst
;
989 EXPORT_SYMBOL(rt6_lookup
);
991 /* ip6_ins_rt is called with FREE table->tb6_lock.
992 * It takes new route entry, the addition fails by any reason the
994 * Caller must hold dst before calling it.
997 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
,
998 struct mx6_config
*mxc
,
999 struct netlink_ext_ack
*extack
)
1002 struct fib6_table
*table
;
1004 table
= rt
->rt6i_table
;
1005 spin_lock_bh(&table
->tb6_lock
);
1006 err
= fib6_add(&table
->tb6_root
, rt
, info
, mxc
, extack
);
1007 spin_unlock_bh(&table
->tb6_lock
);
1012 int ip6_ins_rt(struct rt6_info
*rt
)
1014 struct nl_info info
= { .nl_net
= dev_net(rt
->dst
.dev
), };
1015 struct mx6_config mxc
= { .mx
= NULL
, };
1017 /* Hold dst to account for the reference from the fib6 tree */
1019 return __ip6_ins_rt(rt
, &info
, &mxc
, NULL
);
1022 /* called with rcu_lock held */
1023 static struct net_device
*ip6_rt_get_dev_rcu(struct rt6_info
*rt
)
1025 struct net_device
*dev
= rt
->dst
.dev
;
1027 if (rt
->rt6i_flags
& (RTF_LOCAL
| RTF_ANYCAST
)) {
1028 /* for copies of local routes, dst->dev needs to be the
1029 * device if it is a master device, the master device if
1030 * device is enslaved, and the loopback as the default
1032 if (netif_is_l3_slave(dev
) &&
1033 !rt6_need_strict(&rt
->rt6i_dst
.addr
))
1034 dev
= l3mdev_master_dev_rcu(dev
);
1035 else if (!netif_is_l3_master(dev
))
1036 dev
= dev_net(dev
)->loopback_dev
;
1037 /* last case is netif_is_l3_master(dev) is true in which
1038 * case we want dev returned to be dev
1045 static struct rt6_info
*ip6_rt_cache_alloc(struct rt6_info
*ort
,
1046 const struct in6_addr
*daddr
,
1047 const struct in6_addr
*saddr
)
1049 struct net_device
*dev
;
1050 struct rt6_info
*rt
;
1056 if (ort
->rt6i_flags
& (RTF_CACHE
| RTF_PCPU
))
1057 ort
= (struct rt6_info
*)ort
->dst
.from
;
1060 dev
= ip6_rt_get_dev_rcu(ort
);
1061 rt
= __ip6_dst_alloc(dev_net(dev
), dev
, 0);
1066 ip6_rt_copy_init(rt
, ort
);
1067 rt
->rt6i_flags
|= RTF_CACHE
;
1068 rt
->rt6i_metric
= 0;
1069 rt
->dst
.flags
|= DST_HOST
;
1070 rt
->rt6i_dst
.addr
= *daddr
;
1071 rt
->rt6i_dst
.plen
= 128;
1073 if (!rt6_is_gw_or_nonexthop(ort
)) {
1074 if (ort
->rt6i_dst
.plen
!= 128 &&
1075 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
1076 rt
->rt6i_flags
|= RTF_ANYCAST
;
1077 #ifdef CONFIG_IPV6_SUBTREES
1078 if (rt
->rt6i_src
.plen
&& saddr
) {
1079 rt
->rt6i_src
.addr
= *saddr
;
1080 rt
->rt6i_src
.plen
= 128;
1088 static struct rt6_info
*ip6_rt_pcpu_alloc(struct rt6_info
*rt
)
1090 struct net_device
*dev
;
1091 struct rt6_info
*pcpu_rt
;
1094 dev
= ip6_rt_get_dev_rcu(rt
);
1095 pcpu_rt
= __ip6_dst_alloc(dev_net(dev
), dev
, rt
->dst
.flags
);
1099 ip6_rt_copy_init(pcpu_rt
, rt
);
1100 pcpu_rt
->rt6i_protocol
= rt
->rt6i_protocol
;
1101 pcpu_rt
->rt6i_flags
|= RTF_PCPU
;
1105 /* It should be called with rcu_read_lock() acquired */
1106 static struct rt6_info
*rt6_get_pcpu_route(struct rt6_info
*rt
)
1108 struct rt6_info
*pcpu_rt
, **p
;
1110 p
= this_cpu_ptr(rt
->rt6i_pcpu
);
1113 if (pcpu_rt
&& ip6_hold_safe(NULL
, &pcpu_rt
, false))
1114 rt6_dst_from_metrics_check(pcpu_rt
);
1119 static struct rt6_info
*rt6_make_pcpu_route(struct rt6_info
*rt
)
1121 struct rt6_info
*pcpu_rt
, *prev
, **p
;
1123 pcpu_rt
= ip6_rt_pcpu_alloc(rt
);
1125 struct net
*net
= dev_net(rt
->dst
.dev
);
1127 dst_hold(&net
->ipv6
.ip6_null_entry
->dst
);
1128 return net
->ipv6
.ip6_null_entry
;
1131 dst_hold(&pcpu_rt
->dst
);
1132 p
= this_cpu_ptr(rt
->rt6i_pcpu
);
1133 prev
= cmpxchg(p
, NULL
, pcpu_rt
);
1136 rt6_dst_from_metrics_check(pcpu_rt
);
1140 /* exception hash table implementation
1142 static DEFINE_SPINLOCK(rt6_exception_lock
);
1144 /* Remove rt6_ex from hash table and free the memory
1145 * Caller must hold rt6_exception_lock
1147 static void rt6_remove_exception(struct rt6_exception_bucket
*bucket
,
1148 struct rt6_exception
*rt6_ex
)
1152 if (!bucket
|| !rt6_ex
)
1155 net
= dev_net(rt6_ex
->rt6i
->dst
.dev
);
1156 rt6_ex
->rt6i
->rt6i_node
= NULL
;
1157 hlist_del_rcu(&rt6_ex
->hlist
);
1158 rt6_release(rt6_ex
->rt6i
);
1159 kfree_rcu(rt6_ex
, rcu
);
1160 WARN_ON_ONCE(!bucket
->depth
);
1162 net
->ipv6
.rt6_stats
->fib_rt_cache
--;
1165 /* Remove oldest rt6_ex in bucket and free the memory
1166 * Caller must hold rt6_exception_lock
1168 static void rt6_exception_remove_oldest(struct rt6_exception_bucket
*bucket
)
1170 struct rt6_exception
*rt6_ex
, *oldest
= NULL
;
1175 hlist_for_each_entry(rt6_ex
, &bucket
->chain
, hlist
) {
1176 if (!oldest
|| time_before(rt6_ex
->stamp
, oldest
->stamp
))
1179 rt6_remove_exception(bucket
, oldest
);
1182 static u32
rt6_exception_hash(const struct in6_addr
*dst
,
1183 const struct in6_addr
*src
)
1185 static u32 seed __read_mostly
;
1188 net_get_random_once(&seed
, sizeof(seed
));
1189 val
= jhash(dst
, sizeof(*dst
), seed
);
1191 #ifdef CONFIG_IPV6_SUBTREES
1193 val
= jhash(src
, sizeof(*src
), val
);
1195 return hash_32(val
, FIB6_EXCEPTION_BUCKET_SIZE_SHIFT
);
1198 /* Helper function to find the cached rt in the hash table
1199 * and update bucket pointer to point to the bucket for this
1200 * (daddr, saddr) pair
1201 * Caller must hold rt6_exception_lock
1203 static struct rt6_exception
*
1204 __rt6_find_exception_spinlock(struct rt6_exception_bucket
**bucket
,
1205 const struct in6_addr
*daddr
,
1206 const struct in6_addr
*saddr
)
1208 struct rt6_exception
*rt6_ex
;
1211 if (!(*bucket
) || !daddr
)
1214 hval
= rt6_exception_hash(daddr
, saddr
);
1217 hlist_for_each_entry(rt6_ex
, &(*bucket
)->chain
, hlist
) {
1218 struct rt6_info
*rt6
= rt6_ex
->rt6i
;
1219 bool matched
= ipv6_addr_equal(daddr
, &rt6
->rt6i_dst
.addr
);
1221 #ifdef CONFIG_IPV6_SUBTREES
1222 if (matched
&& saddr
)
1223 matched
= ipv6_addr_equal(saddr
, &rt6
->rt6i_src
.addr
);
1231 /* Helper function to find the cached rt in the hash table
1232 * and update bucket pointer to point to the bucket for this
1233 * (daddr, saddr) pair
1234 * Caller must hold rcu_read_lock()
1236 static struct rt6_exception
*
1237 __rt6_find_exception_rcu(struct rt6_exception_bucket
**bucket
,
1238 const struct in6_addr
*daddr
,
1239 const struct in6_addr
*saddr
)
1241 struct rt6_exception
*rt6_ex
;
1244 WARN_ON_ONCE(!rcu_read_lock_held());
1246 if (!(*bucket
) || !daddr
)
1249 hval
= rt6_exception_hash(daddr
, saddr
);
1252 hlist_for_each_entry_rcu(rt6_ex
, &(*bucket
)->chain
, hlist
) {
1253 struct rt6_info
*rt6
= rt6_ex
->rt6i
;
1254 bool matched
= ipv6_addr_equal(daddr
, &rt6
->rt6i_dst
.addr
);
1256 #ifdef CONFIG_IPV6_SUBTREES
1257 if (matched
&& saddr
)
1258 matched
= ipv6_addr_equal(saddr
, &rt6
->rt6i_src
.addr
);
1266 static int rt6_insert_exception(struct rt6_info
*nrt
,
1267 struct rt6_info
*ort
)
1269 struct net
*net
= dev_net(ort
->dst
.dev
);
1270 struct rt6_exception_bucket
*bucket
;
1271 struct in6_addr
*src_key
= NULL
;
1272 struct rt6_exception
*rt6_ex
;
1275 /* ort can't be a cache or pcpu route */
1276 if (ort
->rt6i_flags
& (RTF_CACHE
| RTF_PCPU
))
1277 ort
= (struct rt6_info
*)ort
->dst
.from
;
1278 WARN_ON_ONCE(ort
->rt6i_flags
& (RTF_CACHE
| RTF_PCPU
));
1280 spin_lock_bh(&rt6_exception_lock
);
1282 if (ort
->exception_bucket_flushed
) {
1287 bucket
= rcu_dereference_protected(ort
->rt6i_exception_bucket
,
1288 lockdep_is_held(&rt6_exception_lock
));
1290 bucket
= kcalloc(FIB6_EXCEPTION_BUCKET_SIZE
, sizeof(*bucket
),
1296 rcu_assign_pointer(ort
->rt6i_exception_bucket
, bucket
);
1299 #ifdef CONFIG_IPV6_SUBTREES
1300 /* rt6i_src.plen != 0 indicates ort is in subtree
1301 * and exception table is indexed by a hash of
1302 * both rt6i_dst and rt6i_src.
1303 * Otherwise, the exception table is indexed by
1304 * a hash of only rt6i_dst.
1306 if (ort
->rt6i_src
.plen
)
1307 src_key
= &nrt
->rt6i_src
.addr
;
1310 /* Update rt6i_prefsrc as it could be changed
1311 * in rt6_remove_prefsrc()
1313 nrt
->rt6i_prefsrc
= ort
->rt6i_prefsrc
;
1314 /* rt6_mtu_change() might lower mtu on ort.
1315 * Only insert this exception route if its mtu
1316 * is less than ort's mtu value.
1318 if (nrt
->rt6i_pmtu
>= dst_mtu(&ort
->dst
)) {
1323 rt6_ex
= __rt6_find_exception_spinlock(&bucket
, &nrt
->rt6i_dst
.addr
,
1326 rt6_remove_exception(bucket
, rt6_ex
);
1328 rt6_ex
= kzalloc(sizeof(*rt6_ex
), GFP_ATOMIC
);
1334 rt6_ex
->stamp
= jiffies
;
1335 atomic_inc(&nrt
->rt6i_ref
);
1336 nrt
->rt6i_node
= ort
->rt6i_node
;
1337 hlist_add_head_rcu(&rt6_ex
->hlist
, &bucket
->chain
);
1339 net
->ipv6
.rt6_stats
->fib_rt_cache
++;
1341 if (bucket
->depth
> FIB6_MAX_DEPTH
)
1342 rt6_exception_remove_oldest(bucket
);
1345 spin_unlock_bh(&rt6_exception_lock
);
1347 /* Update fn->fn_sernum to invalidate all cached dst */
1349 fib6_update_sernum(ort
);
1350 fib6_force_start_gc(net
);
1356 void rt6_flush_exceptions(struct rt6_info
*rt
)
1358 struct rt6_exception_bucket
*bucket
;
1359 struct rt6_exception
*rt6_ex
;
1360 struct hlist_node
*tmp
;
1363 spin_lock_bh(&rt6_exception_lock
);
1364 /* Prevent rt6_insert_exception() to recreate the bucket list */
1365 rt
->exception_bucket_flushed
= 1;
1367 bucket
= rcu_dereference_protected(rt
->rt6i_exception_bucket
,
1368 lockdep_is_held(&rt6_exception_lock
));
1372 for (i
= 0; i
< FIB6_EXCEPTION_BUCKET_SIZE
; i
++) {
1373 hlist_for_each_entry_safe(rt6_ex
, tmp
, &bucket
->chain
, hlist
)
1374 rt6_remove_exception(bucket
, rt6_ex
);
1375 WARN_ON_ONCE(bucket
->depth
);
1380 spin_unlock_bh(&rt6_exception_lock
);
1383 /* Find cached rt in the hash table inside passed in rt
1384 * Caller has to hold rcu_read_lock()
1386 static struct rt6_info
*rt6_find_cached_rt(struct rt6_info
*rt
,
1387 struct in6_addr
*daddr
,
1388 struct in6_addr
*saddr
)
1390 struct rt6_exception_bucket
*bucket
;
1391 struct in6_addr
*src_key
= NULL
;
1392 struct rt6_exception
*rt6_ex
;
1393 struct rt6_info
*res
= NULL
;
1395 bucket
= rcu_dereference(rt
->rt6i_exception_bucket
);
1397 #ifdef CONFIG_IPV6_SUBTREES
1398 /* rt6i_src.plen != 0 indicates rt is in subtree
1399 * and exception table is indexed by a hash of
1400 * both rt6i_dst and rt6i_src.
1401 * Otherwise, the exception table is indexed by
1402 * a hash of only rt6i_dst.
1404 if (rt
->rt6i_src
.plen
)
1407 rt6_ex
= __rt6_find_exception_rcu(&bucket
, daddr
, src_key
);
1409 if (rt6_ex
&& !rt6_check_expired(rt6_ex
->rt6i
))
1415 /* Remove the passed in cached rt from the hash table that contains it */
1416 int rt6_remove_exception_rt(struct rt6_info
*rt
)
1418 struct rt6_info
*from
= (struct rt6_info
*)rt
->dst
.from
;
1419 struct rt6_exception_bucket
*bucket
;
1420 struct in6_addr
*src_key
= NULL
;
1421 struct rt6_exception
*rt6_ex
;
1425 !(rt
->rt6i_flags
& RTF_CACHE
))
1428 if (!rcu_access_pointer(from
->rt6i_exception_bucket
))
1431 spin_lock_bh(&rt6_exception_lock
);
1432 bucket
= rcu_dereference_protected(from
->rt6i_exception_bucket
,
1433 lockdep_is_held(&rt6_exception_lock
));
1434 #ifdef CONFIG_IPV6_SUBTREES
1435 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1436 * and exception table is indexed by a hash of
1437 * both rt6i_dst and rt6i_src.
1438 * Otherwise, the exception table is indexed by
1439 * a hash of only rt6i_dst.
1441 if (from
->rt6i_src
.plen
)
1442 src_key
= &rt
->rt6i_src
.addr
;
1444 rt6_ex
= __rt6_find_exception_spinlock(&bucket
,
1448 rt6_remove_exception(bucket
, rt6_ex
);
1454 spin_unlock_bh(&rt6_exception_lock
);
1458 /* Find rt6_ex which contains the passed in rt cache and
1461 static void rt6_update_exception_stamp_rt(struct rt6_info
*rt
)
1463 struct rt6_info
*from
= (struct rt6_info
*)rt
->dst
.from
;
1464 struct rt6_exception_bucket
*bucket
;
1465 struct in6_addr
*src_key
= NULL
;
1466 struct rt6_exception
*rt6_ex
;
1469 !(rt
->rt6i_flags
& RTF_CACHE
))
1473 bucket
= rcu_dereference(from
->rt6i_exception_bucket
);
1475 #ifdef CONFIG_IPV6_SUBTREES
1476 /* rt6i_src.plen != 0 indicates 'from' is in subtree
1477 * and exception table is indexed by a hash of
1478 * both rt6i_dst and rt6i_src.
1479 * Otherwise, the exception table is indexed by
1480 * a hash of only rt6i_dst.
1482 if (from
->rt6i_src
.plen
)
1483 src_key
= &rt
->rt6i_src
.addr
;
1485 rt6_ex
= __rt6_find_exception_rcu(&bucket
,
1489 rt6_ex
->stamp
= jiffies
;
1494 static void rt6_exceptions_remove_prefsrc(struct rt6_info
*rt
)
1496 struct rt6_exception_bucket
*bucket
;
1497 struct rt6_exception
*rt6_ex
;
1500 bucket
= rcu_dereference_protected(rt
->rt6i_exception_bucket
,
1501 lockdep_is_held(&rt6_exception_lock
));
1504 for (i
= 0; i
< FIB6_EXCEPTION_BUCKET_SIZE
; i
++) {
1505 hlist_for_each_entry(rt6_ex
, &bucket
->chain
, hlist
) {
1506 rt6_ex
->rt6i
->rt6i_prefsrc
.plen
= 0;
1513 static bool rt6_mtu_change_route_allowed(struct inet6_dev
*idev
,
1514 struct rt6_info
*rt
, int mtu
)
1516 /* If the new MTU is lower than the route PMTU, this new MTU will be the
1517 * lowest MTU in the path: always allow updating the route PMTU to
1518 * reflect PMTU decreases.
1520 * If the new MTU is higher, and the route PMTU is equal to the local
1521 * MTU, this means the old MTU is the lowest in the path, so allow
1522 * updating it: if other nodes now have lower MTUs, PMTU discovery will
1526 if (dst_mtu(&rt
->dst
) >= mtu
)
1529 if (dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
)
1535 static void rt6_exceptions_update_pmtu(struct inet6_dev
*idev
,
1536 struct rt6_info
*rt
, int mtu
)
1538 struct rt6_exception_bucket
*bucket
;
1539 struct rt6_exception
*rt6_ex
;
1542 bucket
= rcu_dereference_protected(rt
->rt6i_exception_bucket
,
1543 lockdep_is_held(&rt6_exception_lock
));
1548 for (i
= 0; i
< FIB6_EXCEPTION_BUCKET_SIZE
; i
++) {
1549 hlist_for_each_entry(rt6_ex
, &bucket
->chain
, hlist
) {
1550 struct rt6_info
*entry
= rt6_ex
->rt6i
;
1552 /* For RTF_CACHE with rt6i_pmtu == 0 (i.e. a redirected
1553 * route), the metrics of its rt->dst.from have already
1556 if (entry
->rt6i_pmtu
&&
1557 rt6_mtu_change_route_allowed(idev
, entry
, mtu
))
1558 entry
->rt6i_pmtu
= mtu
;
1564 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
1566 static void rt6_exceptions_clean_tohost(struct rt6_info
*rt
,
1567 struct in6_addr
*gateway
)
1569 struct rt6_exception_bucket
*bucket
;
1570 struct rt6_exception
*rt6_ex
;
1571 struct hlist_node
*tmp
;
1574 if (!rcu_access_pointer(rt
->rt6i_exception_bucket
))
1577 spin_lock_bh(&rt6_exception_lock
);
1578 bucket
= rcu_dereference_protected(rt
->rt6i_exception_bucket
,
1579 lockdep_is_held(&rt6_exception_lock
));
1582 for (i
= 0; i
< FIB6_EXCEPTION_BUCKET_SIZE
; i
++) {
1583 hlist_for_each_entry_safe(rt6_ex
, tmp
,
1584 &bucket
->chain
, hlist
) {
1585 struct rt6_info
*entry
= rt6_ex
->rt6i
;
1587 if ((entry
->rt6i_flags
& RTF_CACHE_GATEWAY
) ==
1588 RTF_CACHE_GATEWAY
&&
1589 ipv6_addr_equal(gateway
,
1590 &entry
->rt6i_gateway
)) {
1591 rt6_remove_exception(bucket
, rt6_ex
);
1598 spin_unlock_bh(&rt6_exception_lock
);
1601 static void rt6_age_examine_exception(struct rt6_exception_bucket
*bucket
,
1602 struct rt6_exception
*rt6_ex
,
1603 struct fib6_gc_args
*gc_args
,
1606 struct rt6_info
*rt
= rt6_ex
->rt6i
;
1608 /* we are pruning and obsoleting aged-out and non gateway exceptions
1609 * even if others have still references to them, so that on next
1610 * dst_check() such references can be dropped.
1611 * EXPIRES exceptions - e.g. pmtu-generated ones are pruned when
1612 * expired, independently from their aging, as per RFC 8201 section 4
1614 if (!(rt
->rt6i_flags
& RTF_EXPIRES
)) {
1615 if (time_after_eq(now
, rt
->dst
.lastuse
+ gc_args
->timeout
)) {
1616 RT6_TRACE("aging clone %p\n", rt
);
1617 rt6_remove_exception(bucket
, rt6_ex
);
1620 } else if (time_after(jiffies
, rt
->dst
.expires
)) {
1621 RT6_TRACE("purging expired route %p\n", rt
);
1622 rt6_remove_exception(bucket
, rt6_ex
);
1626 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
1627 struct neighbour
*neigh
;
1628 __u8 neigh_flags
= 0;
1630 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
1632 neigh_flags
= neigh
->flags
;
1634 if (!(neigh_flags
& NTF_ROUTER
)) {
1635 RT6_TRACE("purging route %p via non-router but gateway\n",
1637 rt6_remove_exception(bucket
, rt6_ex
);
1645 void rt6_age_exceptions(struct rt6_info
*rt
,
1646 struct fib6_gc_args
*gc_args
,
1649 struct rt6_exception_bucket
*bucket
;
1650 struct rt6_exception
*rt6_ex
;
1651 struct hlist_node
*tmp
;
1654 if (!rcu_access_pointer(rt
->rt6i_exception_bucket
))
1658 spin_lock(&rt6_exception_lock
);
1659 bucket
= rcu_dereference_protected(rt
->rt6i_exception_bucket
,
1660 lockdep_is_held(&rt6_exception_lock
));
1663 for (i
= 0; i
< FIB6_EXCEPTION_BUCKET_SIZE
; i
++) {
1664 hlist_for_each_entry_safe(rt6_ex
, tmp
,
1665 &bucket
->chain
, hlist
) {
1666 rt6_age_examine_exception(bucket
, rt6_ex
,
1672 spin_unlock(&rt6_exception_lock
);
1673 rcu_read_unlock_bh();
1676 struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
,
1677 int oif
, struct flowi6
*fl6
, int flags
)
1679 struct fib6_node
*fn
, *saved_fn
;
1680 struct rt6_info
*rt
, *rt_cache
;
1683 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
1684 strict
|= flags
& RT6_LOOKUP_F_IGNORE_LINKSTATE
;
1685 if (net
->ipv6
.devconf_all
->forwarding
== 0)
1686 strict
|= RT6_LOOKUP_F_REACHABLE
;
1690 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1693 if (fl6
->flowi6_flags
& FLOWI_FLAG_SKIP_NH_OIF
)
1697 rt
= rt6_select(net
, fn
, oif
, strict
);
1698 if (rt
->rt6i_nsiblings
)
1699 rt
= rt6_multipath_select(rt
, fl6
, oif
, strict
);
1700 if (rt
== net
->ipv6
.ip6_null_entry
) {
1701 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
1703 goto redo_rt6_select
;
1704 else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
1705 /* also consider unreachable route */
1706 strict
&= ~RT6_LOOKUP_F_REACHABLE
;
1708 goto redo_rt6_select
;
1712 /*Search through exception table */
1713 rt_cache
= rt6_find_cached_rt(rt
, &fl6
->daddr
, &fl6
->saddr
);
1717 if (rt
== net
->ipv6
.ip6_null_entry
) {
1720 trace_fib6_table_lookup(net
, rt
, table
, fl6
);
1722 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1723 if (ip6_hold_safe(net
, &rt
, true)) {
1724 dst_use_noref(&rt
->dst
, jiffies
);
1725 rt6_dst_from_metrics_check(rt
);
1728 trace_fib6_table_lookup(net
, rt
, table
, fl6
);
1730 } else if (unlikely((fl6
->flowi6_flags
& FLOWI_FLAG_KNOWN_NH
) &&
1731 !(rt
->rt6i_flags
& RTF_GATEWAY
))) {
1732 /* Create a RTF_CACHE clone which will not be
1733 * owned by the fib6 tree. It is for the special case where
1734 * the daddr in the skb during the neighbor look-up is different
1735 * from the fl6->daddr used to look-up route here.
1738 struct rt6_info
*uncached_rt
;
1740 if (ip6_hold_safe(net
, &rt
, true)) {
1741 dst_use_noref(&rt
->dst
, jiffies
);
1745 goto uncached_rt_out
;
1749 uncached_rt
= ip6_rt_cache_alloc(rt
, &fl6
->daddr
, NULL
);
1750 dst_release(&rt
->dst
);
1753 /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
1754 * No need for another dst_hold()
1756 rt6_uncached_list_add(uncached_rt
);
1757 atomic_inc(&net
->ipv6
.rt6_stats
->fib_rt_uncache
);
1759 uncached_rt
= net
->ipv6
.ip6_null_entry
;
1760 dst_hold(&uncached_rt
->dst
);
1764 trace_fib6_table_lookup(net
, uncached_rt
, table
, fl6
);
1768 /* Get a percpu copy */
1770 struct rt6_info
*pcpu_rt
;
1772 dst_use_noref(&rt
->dst
, jiffies
);
1774 pcpu_rt
= rt6_get_pcpu_route(rt
);
1777 /* atomic_inc_not_zero() is needed when using rcu */
1778 if (atomic_inc_not_zero(&rt
->rt6i_ref
)) {
1779 /* No dst_hold() on rt is needed because grabbing
1780 * rt->rt6i_ref makes sure rt can't be released.
1782 pcpu_rt
= rt6_make_pcpu_route(rt
);
1785 /* rt is already removed from tree */
1786 pcpu_rt
= net
->ipv6
.ip6_null_entry
;
1787 dst_hold(&pcpu_rt
->dst
);
1792 trace_fib6_table_lookup(net
, pcpu_rt
, table
, fl6
);
1796 EXPORT_SYMBOL_GPL(ip6_pol_route
);
1798 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
1799 struct flowi6
*fl6
, int flags
)
1801 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
1804 struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
1805 struct net_device
*dev
,
1806 struct flowi6
*fl6
, int flags
)
1808 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
1809 flags
|= RT6_LOOKUP_F_IFACE
;
1811 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
1813 EXPORT_SYMBOL_GPL(ip6_route_input_lookup
);
1815 static void ip6_multipath_l3_keys(const struct sk_buff
*skb
,
1816 struct flow_keys
*keys
)
1818 const struct ipv6hdr
*outer_iph
= ipv6_hdr(skb
);
1819 const struct ipv6hdr
*key_iph
= outer_iph
;
1820 const struct ipv6hdr
*inner_iph
;
1821 const struct icmp6hdr
*icmph
;
1822 struct ipv6hdr _inner_iph
;
1824 if (likely(outer_iph
->nexthdr
!= IPPROTO_ICMPV6
))
1827 icmph
= icmp6_hdr(skb
);
1828 if (icmph
->icmp6_type
!= ICMPV6_DEST_UNREACH
&&
1829 icmph
->icmp6_type
!= ICMPV6_PKT_TOOBIG
&&
1830 icmph
->icmp6_type
!= ICMPV6_TIME_EXCEED
&&
1831 icmph
->icmp6_type
!= ICMPV6_PARAMPROB
)
1834 inner_iph
= skb_header_pointer(skb
,
1835 skb_transport_offset(skb
) + sizeof(*icmph
),
1836 sizeof(_inner_iph
), &_inner_iph
);
1840 key_iph
= inner_iph
;
1842 memset(keys
, 0, sizeof(*keys
));
1843 keys
->control
.addr_type
= FLOW_DISSECTOR_KEY_IPV6_ADDRS
;
1844 keys
->addrs
.v6addrs
.src
= key_iph
->saddr
;
1845 keys
->addrs
.v6addrs
.dst
= key_iph
->daddr
;
1846 keys
->tags
.flow_label
= ip6_flowinfo(key_iph
);
1847 keys
->basic
.ip_proto
= key_iph
->nexthdr
;
1850 /* if skb is set it will be used and fl6 can be NULL */
1851 u32
rt6_multipath_hash(const struct flowi6
*fl6
, const struct sk_buff
*skb
)
1853 struct flow_keys hash_keys
;
1856 ip6_multipath_l3_keys(skb
, &hash_keys
);
1857 return flow_hash_from_keys(&hash_keys
);
1860 return get_hash_from_flowi6(fl6
);
1863 void ip6_route_input(struct sk_buff
*skb
)
1865 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1866 struct net
*net
= dev_net(skb
->dev
);
1867 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1868 struct ip_tunnel_info
*tun_info
;
1869 struct flowi6 fl6
= {
1870 .flowi6_iif
= skb
->dev
->ifindex
,
1871 .daddr
= iph
->daddr
,
1872 .saddr
= iph
->saddr
,
1873 .flowlabel
= ip6_flowinfo(iph
),
1874 .flowi6_mark
= skb
->mark
,
1875 .flowi6_proto
= iph
->nexthdr
,
1878 tun_info
= skb_tunnel_info(skb
);
1879 if (tun_info
&& !(tun_info
->mode
& IP_TUNNEL_INFO_TX
))
1880 fl6
.flowi6_tun_key
.tun_id
= tun_info
->key
.tun_id
;
1881 if (unlikely(fl6
.flowi6_proto
== IPPROTO_ICMPV6
))
1882 fl6
.mp_hash
= rt6_multipath_hash(&fl6
, skb
);
1884 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
1887 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
1888 struct flowi6
*fl6
, int flags
)
1890 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1893 struct dst_entry
*ip6_route_output_flags(struct net
*net
, const struct sock
*sk
,
1894 struct flowi6
*fl6
, int flags
)
1898 if (rt6_need_strict(&fl6
->daddr
)) {
1899 struct dst_entry
*dst
;
1901 dst
= l3mdev_link_scope_lookup(net
, fl6
);
1906 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1908 any_src
= ipv6_addr_any(&fl6
->saddr
);
1909 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
) ||
1910 (fl6
->flowi6_oif
&& any_src
))
1911 flags
|= RT6_LOOKUP_F_IFACE
;
1914 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1916 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1918 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1920 EXPORT_SYMBOL_GPL(ip6_route_output_flags
);
1922 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1924 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1925 struct net_device
*loopback_dev
= net
->loopback_dev
;
1926 struct dst_entry
*new = NULL
;
1928 rt
= dst_alloc(&ip6_dst_blackhole_ops
, loopback_dev
, 1,
1929 DST_OBSOLETE_DEAD
, 0);
1932 atomic_inc(&net
->ipv6
.rt6_stats
->fib_rt_alloc
);
1936 new->input
= dst_discard
;
1937 new->output
= dst_discard_out
;
1939 dst_copy_metrics(new, &ort
->dst
);
1941 rt
->rt6i_idev
= in6_dev_get(loopback_dev
);
1942 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1943 rt
->rt6i_flags
= ort
->rt6i_flags
& ~RTF_PCPU
;
1944 rt
->rt6i_metric
= 0;
1946 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1947 #ifdef CONFIG_IPV6_SUBTREES
1948 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1952 dst_release(dst_orig
);
1953 return new ? new : ERR_PTR(-ENOMEM
);
1957 * Destination cache support functions
1960 static void rt6_dst_from_metrics_check(struct rt6_info
*rt
)
1963 dst_metrics_ptr(&rt
->dst
) != dst_metrics_ptr(rt
->dst
.from
))
1964 dst_init_metrics(&rt
->dst
, dst_metrics_ptr(rt
->dst
.from
), true);
1967 static struct dst_entry
*rt6_check(struct rt6_info
*rt
, u32 cookie
)
1971 if (!rt6_get_cookie_safe(rt
, &rt_cookie
) || rt_cookie
!= cookie
)
1974 if (rt6_check_expired(rt
))
1980 static struct dst_entry
*rt6_dst_from_check(struct rt6_info
*rt
, u32 cookie
)
1982 if (!__rt6_check_expired(rt
) &&
1983 rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1984 rt6_check((struct rt6_info
*)(rt
->dst
.from
), cookie
))
1990 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1992 struct rt6_info
*rt
;
1994 rt
= (struct rt6_info
*) dst
;
1996 /* All IPV6 dsts are created with ->obsolete set to the value
1997 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1998 * into this function always.
2001 rt6_dst_from_metrics_check(rt
);
2003 if (rt
->rt6i_flags
& RTF_PCPU
||
2004 (unlikely(!list_empty(&rt
->rt6i_uncached
)) && rt
->dst
.from
))
2005 return rt6_dst_from_check(rt
, cookie
);
2007 return rt6_check(rt
, cookie
);
2010 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
2012 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
2015 if (rt
->rt6i_flags
& RTF_CACHE
) {
2016 if (rt6_check_expired(rt
)) {
2028 static void ip6_link_failure(struct sk_buff
*skb
)
2030 struct rt6_info
*rt
;
2032 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
2034 rt
= (struct rt6_info
*) skb_dst(skb
);
2036 if (rt
->rt6i_flags
& RTF_CACHE
) {
2037 if (dst_hold_safe(&rt
->dst
))
2040 struct fib6_node
*fn
;
2043 fn
= rcu_dereference(rt
->rt6i_node
);
2044 if (fn
&& (rt
->rt6i_flags
& RTF_DEFAULT
))
2051 static void rt6_do_update_pmtu(struct rt6_info
*rt
, u32 mtu
)
2053 struct net
*net
= dev_net(rt
->dst
.dev
);
2055 rt
->rt6i_flags
|= RTF_MODIFIED
;
2056 rt
->rt6i_pmtu
= mtu
;
2057 rt6_update_expires(rt
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
2060 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info
*rt
)
2062 return !(rt
->rt6i_flags
& RTF_CACHE
) &&
2063 (rt
->rt6i_flags
& RTF_PCPU
||
2064 rcu_access_pointer(rt
->rt6i_node
));
2067 static void __ip6_rt_update_pmtu(struct dst_entry
*dst
, const struct sock
*sk
,
2068 const struct ipv6hdr
*iph
, u32 mtu
)
2070 const struct in6_addr
*daddr
, *saddr
;
2071 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
2073 if (rt6
->rt6i_flags
& RTF_LOCAL
)
2076 if (dst_metric_locked(dst
, RTAX_MTU
))
2080 daddr
= &iph
->daddr
;
2081 saddr
= &iph
->saddr
;
2083 daddr
= &sk
->sk_v6_daddr
;
2084 saddr
= &inet6_sk(sk
)->saddr
;
2089 dst_confirm_neigh(dst
, daddr
);
2090 mtu
= max_t(u32
, mtu
, IPV6_MIN_MTU
);
2091 if (mtu
>= dst_mtu(dst
))
2094 if (!rt6_cache_allowed_for_pmtu(rt6
)) {
2095 rt6_do_update_pmtu(rt6
, mtu
);
2096 /* update rt6_ex->stamp for cache */
2097 if (rt6
->rt6i_flags
& RTF_CACHE
)
2098 rt6_update_exception_stamp_rt(rt6
);
2100 struct rt6_info
*nrt6
;
2102 nrt6
= ip6_rt_cache_alloc(rt6
, daddr
, saddr
);
2104 rt6_do_update_pmtu(nrt6
, mtu
);
2105 if (rt6_insert_exception(nrt6
, rt6
))
2106 dst_release_immediate(&nrt6
->dst
);
2111 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
2112 struct sk_buff
*skb
, u32 mtu
)
2114 __ip6_rt_update_pmtu(dst
, sk
, skb
? ipv6_hdr(skb
) : NULL
, mtu
);
2117 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
2118 int oif
, u32 mark
, kuid_t uid
)
2120 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
2121 struct dst_entry
*dst
;
2124 memset(&fl6
, 0, sizeof(fl6
));
2125 fl6
.flowi6_oif
= oif
;
2126 fl6
.flowi6_mark
= mark
? mark
: IP6_REPLY_MARK(net
, skb
->mark
);
2127 fl6
.daddr
= iph
->daddr
;
2128 fl6
.saddr
= iph
->saddr
;
2129 fl6
.flowlabel
= ip6_flowinfo(iph
);
2130 fl6
.flowi6_uid
= uid
;
2132 dst
= ip6_route_output(net
, NULL
, &fl6
);
2134 __ip6_rt_update_pmtu(dst
, NULL
, iph
, ntohl(mtu
));
2137 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
2139 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
2141 struct dst_entry
*dst
;
2143 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
2144 sk
->sk_bound_dev_if
, sk
->sk_mark
, sk
->sk_uid
);
2146 dst
= __sk_dst_get(sk
);
2147 if (!dst
|| !dst
->obsolete
||
2148 dst
->ops
->check(dst
, inet6_sk(sk
)->dst_cookie
))
2152 if (!sock_owned_by_user(sk
) && !ipv6_addr_v4mapped(&sk
->sk_v6_daddr
))
2153 ip6_datagram_dst_update(sk
, false);
2156 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
2158 /* Handle redirects */
2159 struct ip6rd_flowi
{
2161 struct in6_addr gateway
;
2164 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
2165 struct fib6_table
*table
,
2169 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl6
;
2170 struct rt6_info
*rt
, *rt_cache
;
2171 struct fib6_node
*fn
;
2173 /* Get the "current" route for this destination and
2174 * check if the redirect has come from appropriate router.
2176 * RFC 4861 specifies that redirects should only be
2177 * accepted if they come from the nexthop to the target.
2178 * Due to the way the routes are chosen, this notion
2179 * is a bit fuzzy and one might need to check all possible
2184 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
2186 for_each_fib6_node_rt_rcu(fn
) {
2187 if (rt6_check_expired(rt
))
2191 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
2193 if (fl6
->flowi6_oif
!= rt
->dst
.dev
->ifindex
)
2195 /* rt_cache's gateway might be different from its 'parent'
2196 * in the case of an ip redirect.
2197 * So we keep searching in the exception table if the gateway
2200 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
)) {
2201 rt_cache
= rt6_find_cached_rt(rt
,
2205 ipv6_addr_equal(&rdfl
->gateway
,
2206 &rt_cache
->rt6i_gateway
)) {
2216 rt
= net
->ipv6
.ip6_null_entry
;
2217 else if (rt
->dst
.error
) {
2218 rt
= net
->ipv6
.ip6_null_entry
;
2222 if (rt
== net
->ipv6
.ip6_null_entry
) {
2223 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
2229 ip6_hold_safe(net
, &rt
, true);
2233 trace_fib6_table_lookup(net
, rt
, table
, fl6
);
2237 static struct dst_entry
*ip6_route_redirect(struct net
*net
,
2238 const struct flowi6
*fl6
,
2239 const struct in6_addr
*gateway
)
2241 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
2242 struct ip6rd_flowi rdfl
;
2245 rdfl
.gateway
= *gateway
;
2247 return fib6_rule_lookup(net
, &rdfl
.fl6
,
2248 flags
, __ip6_route_redirect
);
2251 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
,
2254 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
2255 struct dst_entry
*dst
;
2258 memset(&fl6
, 0, sizeof(fl6
));
2259 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
2260 fl6
.flowi6_oif
= oif
;
2261 fl6
.flowi6_mark
= mark
;
2262 fl6
.daddr
= iph
->daddr
;
2263 fl6
.saddr
= iph
->saddr
;
2264 fl6
.flowlabel
= ip6_flowinfo(iph
);
2265 fl6
.flowi6_uid
= uid
;
2267 dst
= ip6_route_redirect(net
, &fl6
, &ipv6_hdr(skb
)->saddr
);
2268 rt6_do_redirect(dst
, NULL
, skb
);
2271 EXPORT_SYMBOL_GPL(ip6_redirect
);
2273 void ip6_redirect_no_header(struct sk_buff
*skb
, struct net
*net
, int oif
,
2276 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
2277 const struct rd_msg
*msg
= (struct rd_msg
*)icmp6_hdr(skb
);
2278 struct dst_entry
*dst
;
2281 memset(&fl6
, 0, sizeof(fl6
));
2282 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
2283 fl6
.flowi6_oif
= oif
;
2284 fl6
.flowi6_mark
= mark
;
2285 fl6
.daddr
= msg
->dest
;
2286 fl6
.saddr
= iph
->daddr
;
2287 fl6
.flowi6_uid
= sock_net_uid(net
, NULL
);
2289 dst
= ip6_route_redirect(net
, &fl6
, &iph
->saddr
);
2290 rt6_do_redirect(dst
, NULL
, skb
);
2294 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
2296 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
,
2299 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
2301 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
2303 struct net_device
*dev
= dst
->dev
;
2304 unsigned int mtu
= dst_mtu(dst
);
2305 struct net
*net
= dev_net(dev
);
2307 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
2309 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
2310 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2313 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
2314 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
2315 * IPV6_MAXPLEN is also valid and means: "any MSS,
2316 * rely only on pmtu discovery"
2318 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
2323 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
2325 const struct rt6_info
*rt
= (const struct rt6_info
*)dst
;
2326 unsigned int mtu
= rt
->rt6i_pmtu
;
2327 struct inet6_dev
*idev
;
2332 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
2339 idev
= __in6_dev_get(dst
->dev
);
2341 mtu
= idev
->cnf
.mtu6
;
2345 mtu
= min_t(unsigned int, mtu
, IP6_MAX_MTU
);
2347 return mtu
- lwtunnel_headroom(dst
->lwtstate
, mtu
);
2350 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
2353 struct dst_entry
*dst
;
2354 struct rt6_info
*rt
;
2355 struct inet6_dev
*idev
= in6_dev_get(dev
);
2356 struct net
*net
= dev_net(dev
);
2358 if (unlikely(!idev
))
2359 return ERR_PTR(-ENODEV
);
2361 rt
= ip6_dst_alloc(net
, dev
, 0);
2362 if (unlikely(!rt
)) {
2364 dst
= ERR_PTR(-ENOMEM
);
2368 rt
->dst
.flags
|= DST_HOST
;
2369 rt
->dst
.input
= ip6_input
;
2370 rt
->dst
.output
= ip6_output
;
2371 rt
->rt6i_gateway
= fl6
->daddr
;
2372 rt
->rt6i_dst
.addr
= fl6
->daddr
;
2373 rt
->rt6i_dst
.plen
= 128;
2374 rt
->rt6i_idev
= idev
;
2375 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
2377 /* Add this dst into uncached_list so that rt6_ifdown() can
2378 * do proper release of the net_device
2380 rt6_uncached_list_add(rt
);
2381 atomic_inc(&net
->ipv6
.rt6_stats
->fib_rt_uncache
);
2383 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
2389 static int ip6_dst_gc(struct dst_ops
*ops
)
2391 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
2392 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2393 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
2394 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2395 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2396 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
2399 entries
= dst_entries_get_fast(ops
);
2400 if (time_after(rt_last_gc
+ rt_min_interval
, jiffies
) &&
2401 entries
<= rt_max_size
)
2404 net
->ipv6
.ip6_rt_gc_expire
++;
2405 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
, true);
2406 entries
= dst_entries_get_slow(ops
);
2407 if (entries
< ops
->gc_thresh
)
2408 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
2410 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
2411 return entries
> rt_max_size
;
2414 static int ip6_convert_metrics(struct mx6_config
*mxc
,
2415 const struct fib6_config
*cfg
)
2417 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
2418 bool ecn_ca
= false;
2426 mp
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_KERNEL
);
2430 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
2431 int type
= nla_type(nla
);
2436 if (unlikely(type
> RTAX_MAX
))
2439 if (type
== RTAX_CC_ALGO
) {
2440 char tmp
[TCP_CA_NAME_MAX
];
2442 nla_strlcpy(tmp
, nla
, sizeof(tmp
));
2443 val
= tcp_ca_get_key_by_name(net
, tmp
, &ecn_ca
);
2444 if (val
== TCP_CA_UNSPEC
)
2447 val
= nla_get_u32(nla
);
2449 if (type
== RTAX_HOPLIMIT
&& val
> 255)
2451 if (type
== RTAX_FEATURES
&& (val
& ~RTAX_FEATURE_MASK
))
2455 __set_bit(type
- 1, mxc
->mx_valid
);
2459 __set_bit(RTAX_FEATURES
- 1, mxc
->mx_valid
);
2460 mp
[RTAX_FEATURES
- 1] |= DST_FEATURE_ECN_CA
;
2470 static struct rt6_info
*ip6_nh_lookup_table(struct net
*net
,
2471 struct fib6_config
*cfg
,
2472 const struct in6_addr
*gw_addr
)
2474 struct flowi6 fl6
= {
2475 .flowi6_oif
= cfg
->fc_ifindex
,
2477 .saddr
= cfg
->fc_prefsrc
,
2479 struct fib6_table
*table
;
2480 struct rt6_info
*rt
;
2481 int flags
= RT6_LOOKUP_F_IFACE
| RT6_LOOKUP_F_IGNORE_LINKSTATE
;
2483 table
= fib6_get_table(net
, cfg
->fc_table
);
2487 if (!ipv6_addr_any(&cfg
->fc_prefsrc
))
2488 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
2490 rt
= ip6_pol_route(net
, table
, cfg
->fc_ifindex
, &fl6
, flags
);
2492 /* if table lookup failed, fall back to full lookup */
2493 if (rt
== net
->ipv6
.ip6_null_entry
) {
2501 static struct rt6_info
*ip6_route_info_create(struct fib6_config
*cfg
,
2502 struct netlink_ext_ack
*extack
)
2504 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
2505 struct rt6_info
*rt
= NULL
;
2506 struct net_device
*dev
= NULL
;
2507 struct inet6_dev
*idev
= NULL
;
2508 struct fib6_table
*table
;
2512 /* RTF_PCPU is an internal flag; can not be set by userspace */
2513 if (cfg
->fc_flags
& RTF_PCPU
) {
2514 NL_SET_ERR_MSG(extack
, "Userspace can not set RTF_PCPU");
2518 /* RTF_CACHE is an internal flag; can not be set by userspace */
2519 if (cfg
->fc_flags
& RTF_CACHE
) {
2520 NL_SET_ERR_MSG(extack
, "Userspace can not set RTF_CACHE");
2524 if (cfg
->fc_dst_len
> 128) {
2525 NL_SET_ERR_MSG(extack
, "Invalid prefix length");
2528 if (cfg
->fc_src_len
> 128) {
2529 NL_SET_ERR_MSG(extack
, "Invalid source address length");
2532 #ifndef CONFIG_IPV6_SUBTREES
2533 if (cfg
->fc_src_len
) {
2534 NL_SET_ERR_MSG(extack
,
2535 "Specifying source address requires IPV6_SUBTREES to be enabled");
2539 if (cfg
->fc_ifindex
) {
2541 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
2544 idev
= in6_dev_get(dev
);
2549 if (cfg
->fc_metric
== 0)
2550 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
2553 if (cfg
->fc_nlinfo
.nlh
&&
2554 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
2555 table
= fib6_get_table(net
, cfg
->fc_table
);
2557 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
2558 table
= fib6_new_table(net
, cfg
->fc_table
);
2561 table
= fib6_new_table(net
, cfg
->fc_table
);
2567 rt
= ip6_dst_alloc(net
, NULL
,
2568 (cfg
->fc_flags
& RTF_ADDRCONF
) ? 0 : DST_NOCOUNT
);
2575 if (cfg
->fc_flags
& RTF_EXPIRES
)
2576 rt6_set_expires(rt
, jiffies
+
2577 clock_t_to_jiffies(cfg
->fc_expires
));
2579 rt6_clean_expires(rt
);
2581 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
2582 cfg
->fc_protocol
= RTPROT_BOOT
;
2583 rt
->rt6i_protocol
= cfg
->fc_protocol
;
2585 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
2587 if (addr_type
& IPV6_ADDR_MULTICAST
)
2588 rt
->dst
.input
= ip6_mc_input
;
2589 else if (cfg
->fc_flags
& RTF_LOCAL
)
2590 rt
->dst
.input
= ip6_input
;
2592 rt
->dst
.input
= ip6_forward
;
2594 rt
->dst
.output
= ip6_output
;
2596 if (cfg
->fc_encap
) {
2597 struct lwtunnel_state
*lwtstate
;
2599 err
= lwtunnel_build_state(cfg
->fc_encap_type
,
2600 cfg
->fc_encap
, AF_INET6
, cfg
,
2604 rt
->dst
.lwtstate
= lwtstate_get(lwtstate
);
2605 if (lwtunnel_output_redirect(rt
->dst
.lwtstate
)) {
2606 rt
->dst
.lwtstate
->orig_output
= rt
->dst
.output
;
2607 rt
->dst
.output
= lwtunnel_output
;
2609 if (lwtunnel_input_redirect(rt
->dst
.lwtstate
)) {
2610 rt
->dst
.lwtstate
->orig_input
= rt
->dst
.input
;
2611 rt
->dst
.input
= lwtunnel_input
;
2615 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
2616 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
2617 if (rt
->rt6i_dst
.plen
== 128)
2618 rt
->dst
.flags
|= DST_HOST
;
2620 #ifdef CONFIG_IPV6_SUBTREES
2621 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
2622 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
2625 rt
->rt6i_metric
= cfg
->fc_metric
;
2627 /* We cannot add true routes via loopback here,
2628 they would result in kernel looping; promote them to reject routes
2630 if ((cfg
->fc_flags
& RTF_REJECT
) ||
2631 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
2632 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
2633 !(cfg
->fc_flags
& RTF_LOCAL
))) {
2634 /* hold loopback dev/idev if we haven't done so. */
2635 if (dev
!= net
->loopback_dev
) {
2640 dev
= net
->loopback_dev
;
2642 idev
= in6_dev_get(dev
);
2648 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
2649 switch (cfg
->fc_type
) {
2651 rt
->dst
.error
= -EINVAL
;
2652 rt
->dst
.output
= dst_discard_out
;
2653 rt
->dst
.input
= dst_discard
;
2656 rt
->dst
.error
= -EACCES
;
2657 rt
->dst
.output
= ip6_pkt_prohibit_out
;
2658 rt
->dst
.input
= ip6_pkt_prohibit
;
2661 case RTN_UNREACHABLE
:
2663 rt
->dst
.error
= (cfg
->fc_type
== RTN_THROW
) ? -EAGAIN
2664 : (cfg
->fc_type
== RTN_UNREACHABLE
)
2665 ? -EHOSTUNREACH
: -ENETUNREACH
;
2666 rt
->dst
.output
= ip6_pkt_discard_out
;
2667 rt
->dst
.input
= ip6_pkt_discard
;
2673 if (cfg
->fc_flags
& RTF_GATEWAY
) {
2674 const struct in6_addr
*gw_addr
;
2677 gw_addr
= &cfg
->fc_gateway
;
2678 gwa_type
= ipv6_addr_type(gw_addr
);
2680 /* if gw_addr is local we will fail to detect this in case
2681 * address is still TENTATIVE (DAD in progress). rt6_lookup()
2682 * will return already-added prefix route via interface that
2683 * prefix route was assigned to, which might be non-loopback.
2686 if (ipv6_chk_addr_and_flags(net
, gw_addr
,
2687 gwa_type
& IPV6_ADDR_LINKLOCAL
?
2688 dev
: NULL
, 0, 0)) {
2689 NL_SET_ERR_MSG(extack
, "Invalid gateway address");
2692 rt
->rt6i_gateway
= *gw_addr
;
2694 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
2695 struct rt6_info
*grt
= NULL
;
2697 /* IPv6 strictly inhibits using not link-local
2698 addresses as nexthop address.
2699 Otherwise, router will not able to send redirects.
2700 It is very good, but in some (rare!) circumstances
2701 (SIT, PtP, NBMA NOARP links) it is handy to allow
2702 some exceptions. --ANK
2703 We allow IPv4-mapped nexthops to support RFC4798-type
2706 if (!(gwa_type
& (IPV6_ADDR_UNICAST
|
2707 IPV6_ADDR_MAPPED
))) {
2708 NL_SET_ERR_MSG(extack
,
2709 "Invalid gateway address");
2713 if (cfg
->fc_table
) {
2714 grt
= ip6_nh_lookup_table(net
, cfg
, gw_addr
);
2717 if (grt
->rt6i_flags
& RTF_GATEWAY
||
2718 (dev
&& dev
!= grt
->dst
.dev
)) {
2726 grt
= rt6_lookup(net
, gw_addr
, NULL
,
2727 cfg
->fc_ifindex
, 1);
2729 err
= -EHOSTUNREACH
;
2733 if (dev
!= grt
->dst
.dev
) {
2739 idev
= grt
->rt6i_idev
;
2741 in6_dev_hold(grt
->rt6i_idev
);
2743 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
2752 NL_SET_ERR_MSG(extack
, "Egress device not specified");
2754 } else if (dev
->flags
& IFF_LOOPBACK
) {
2755 NL_SET_ERR_MSG(extack
,
2756 "Egress device can not be loopback device for this route");
2765 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
2766 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
2767 NL_SET_ERR_MSG(extack
, "Invalid source address");
2771 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
2772 rt
->rt6i_prefsrc
.plen
= 128;
2774 rt
->rt6i_prefsrc
.plen
= 0;
2776 rt
->rt6i_flags
= cfg
->fc_flags
;
2780 rt
->rt6i_idev
= idev
;
2781 rt
->rt6i_table
= table
;
2783 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
2792 dst_release_immediate(&rt
->dst
);
2794 return ERR_PTR(err
);
2797 int ip6_route_add(struct fib6_config
*cfg
,
2798 struct netlink_ext_ack
*extack
)
2800 struct mx6_config mxc
= { .mx
= NULL
, };
2801 struct rt6_info
*rt
;
2804 rt
= ip6_route_info_create(cfg
, extack
);
2811 err
= ip6_convert_metrics(&mxc
, cfg
);
2815 err
= __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
, &mxc
, extack
);
2822 dst_release_immediate(&rt
->dst
);
2827 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
2830 struct fib6_table
*table
;
2831 struct net
*net
= dev_net(rt
->dst
.dev
);
2833 if (rt
== net
->ipv6
.ip6_null_entry
) {
2838 table
= rt
->rt6i_table
;
2839 spin_lock_bh(&table
->tb6_lock
);
2840 err
= fib6_del(rt
, info
);
2841 spin_unlock_bh(&table
->tb6_lock
);
2848 int ip6_del_rt(struct rt6_info
*rt
)
2850 struct nl_info info
= {
2851 .nl_net
= dev_net(rt
->dst
.dev
),
2853 return __ip6_del_rt(rt
, &info
);
2856 static int __ip6_del_rt_siblings(struct rt6_info
*rt
, struct fib6_config
*cfg
)
2858 struct nl_info
*info
= &cfg
->fc_nlinfo
;
2859 struct net
*net
= info
->nl_net
;
2860 struct sk_buff
*skb
= NULL
;
2861 struct fib6_table
*table
;
2864 if (rt
== net
->ipv6
.ip6_null_entry
)
2866 table
= rt
->rt6i_table
;
2867 spin_lock_bh(&table
->tb6_lock
);
2869 if (rt
->rt6i_nsiblings
&& cfg
->fc_delete_all_nh
) {
2870 struct rt6_info
*sibling
, *next_sibling
;
2872 /* prefer to send a single notification with all hops */
2873 skb
= nlmsg_new(rt6_nlmsg_size(rt
), gfp_any());
2875 u32 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
2877 if (rt6_fill_node(net
, skb
, rt
,
2878 NULL
, NULL
, 0, RTM_DELROUTE
,
2879 info
->portid
, seq
, 0) < 0) {
2883 info
->skip_notify
= 1;
2886 list_for_each_entry_safe(sibling
, next_sibling
,
2889 err
= fib6_del(sibling
, info
);
2895 err
= fib6_del(rt
, info
);
2897 spin_unlock_bh(&table
->tb6_lock
);
2902 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
2903 info
->nlh
, gfp_any());
2908 static int ip6_route_del(struct fib6_config
*cfg
,
2909 struct netlink_ext_ack
*extack
)
2911 struct rt6_info
*rt
, *rt_cache
;
2912 struct fib6_table
*table
;
2913 struct fib6_node
*fn
;
2916 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
2918 NL_SET_ERR_MSG(extack
, "FIB table does not exist");
2924 fn
= fib6_locate(&table
->tb6_root
,
2925 &cfg
->fc_dst
, cfg
->fc_dst_len
,
2926 &cfg
->fc_src
, cfg
->fc_src_len
,
2927 !(cfg
->fc_flags
& RTF_CACHE
));
2930 for_each_fib6_node_rt_rcu(fn
) {
2931 if (cfg
->fc_flags
& RTF_CACHE
) {
2932 rt_cache
= rt6_find_cached_rt(rt
, &cfg
->fc_dst
,
2938 if (cfg
->fc_ifindex
&&
2940 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
2942 if (cfg
->fc_flags
& RTF_GATEWAY
&&
2943 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
2945 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
2947 if (cfg
->fc_protocol
&& cfg
->fc_protocol
!= rt
->rt6i_protocol
)
2949 if (!dst_hold_safe(&rt
->dst
))
2953 /* if gateway was specified only delete the one hop */
2954 if (cfg
->fc_flags
& RTF_GATEWAY
)
2955 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
2957 return __ip6_del_rt_siblings(rt
, cfg
);
2965 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
2967 struct netevent_redirect netevent
;
2968 struct rt6_info
*rt
, *nrt
= NULL
;
2969 struct ndisc_options ndopts
;
2970 struct inet6_dev
*in6_dev
;
2971 struct neighbour
*neigh
;
2973 int optlen
, on_link
;
2976 optlen
= skb_tail_pointer(skb
) - skb_transport_header(skb
);
2977 optlen
-= sizeof(*msg
);
2980 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2984 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
2986 if (ipv6_addr_is_multicast(&msg
->dest
)) {
2987 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2992 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
2994 } else if (ipv6_addr_type(&msg
->target
) !=
2995 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
2996 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
3000 in6_dev
= __in6_dev_get(skb
->dev
);
3003 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
3007 * The IP source address of the Redirect MUST be the same as the current
3008 * first-hop router for the specified ICMP Destination Address.
3011 if (!ndisc_parse_options(skb
->dev
, msg
->opt
, optlen
, &ndopts
)) {
3012 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
3017 if (ndopts
.nd_opts_tgt_lladdr
) {
3018 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
3021 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
3026 rt
= (struct rt6_info
*) dst
;
3027 if (rt
->rt6i_flags
& RTF_REJECT
) {
3028 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
3032 /* Redirect received -> path was valid.
3033 * Look, redirects are sent only in response to data packets,
3034 * so that this nexthop apparently is reachable. --ANK
3036 dst_confirm_neigh(&rt
->dst
, &ipv6_hdr(skb
)->saddr
);
3038 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
3043 * We have finally decided to accept it.
3046 ndisc_update(skb
->dev
, neigh
, lladdr
, NUD_STALE
,
3047 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
3048 NEIGH_UPDATE_F_OVERRIDE
|
3049 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
3050 NEIGH_UPDATE_F_ISROUTER
)),
3051 NDISC_REDIRECT
, &ndopts
);
3053 nrt
= ip6_rt_cache_alloc(rt
, &msg
->dest
, NULL
);
3057 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
3059 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
3061 nrt
->rt6i_protocol
= RTPROT_REDIRECT
;
3062 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
3064 /* No need to remove rt from the exception table if rt is
3065 * a cached route because rt6_insert_exception() will
3068 if (rt6_insert_exception(nrt
, rt
)) {
3069 dst_release_immediate(&nrt
->dst
);
3073 netevent
.old
= &rt
->dst
;
3074 netevent
.new = &nrt
->dst
;
3075 netevent
.daddr
= &msg
->dest
;
3076 netevent
.neigh
= neigh
;
3077 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
3080 neigh_release(neigh
);
3084 * Misc support functions
3087 static void rt6_set_from(struct rt6_info
*rt
, struct rt6_info
*from
)
3089 BUG_ON(from
->dst
.from
);
3091 rt
->rt6i_flags
&= ~RTF_EXPIRES
;
3092 dst_hold(&from
->dst
);
3093 rt
->dst
.from
= &from
->dst
;
3094 dst_init_metrics(&rt
->dst
, dst_metrics_ptr(&from
->dst
), true);
3097 static void ip6_rt_copy_init(struct rt6_info
*rt
, struct rt6_info
*ort
)
3099 rt
->dst
.input
= ort
->dst
.input
;
3100 rt
->dst
.output
= ort
->dst
.output
;
3101 rt
->rt6i_dst
= ort
->rt6i_dst
;
3102 rt
->dst
.error
= ort
->dst
.error
;
3103 rt
->rt6i_idev
= ort
->rt6i_idev
;
3105 in6_dev_hold(rt
->rt6i_idev
);
3106 rt
->dst
.lastuse
= jiffies
;
3107 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
3108 rt
->rt6i_flags
= ort
->rt6i_flags
;
3109 rt6_set_from(rt
, ort
);
3110 rt
->rt6i_metric
= ort
->rt6i_metric
;
3111 #ifdef CONFIG_IPV6_SUBTREES
3112 rt
->rt6i_src
= ort
->rt6i_src
;
3114 rt
->rt6i_prefsrc
= ort
->rt6i_prefsrc
;
3115 rt
->rt6i_table
= ort
->rt6i_table
;
3116 rt
->dst
.lwtstate
= lwtstate_get(ort
->dst
.lwtstate
);
3119 #ifdef CONFIG_IPV6_ROUTE_INFO
3120 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
3121 const struct in6_addr
*prefix
, int prefixlen
,
3122 const struct in6_addr
*gwaddr
,
3123 struct net_device
*dev
)
3125 u32 tb_id
= l3mdev_fib_table(dev
) ? : RT6_TABLE_INFO
;
3126 int ifindex
= dev
->ifindex
;
3127 struct fib6_node
*fn
;
3128 struct rt6_info
*rt
= NULL
;
3129 struct fib6_table
*table
;
3131 table
= fib6_get_table(net
, tb_id
);
3136 fn
= fib6_locate(&table
->tb6_root
, prefix
, prefixlen
, NULL
, 0, true);
3140 for_each_fib6_node_rt_rcu(fn
) {
3141 if (rt
->dst
.dev
->ifindex
!= ifindex
)
3143 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
3145 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
3147 ip6_hold_safe(NULL
, &rt
, false);
3155 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
3156 const struct in6_addr
*prefix
, int prefixlen
,
3157 const struct in6_addr
*gwaddr
,
3158 struct net_device
*dev
,
3161 struct fib6_config cfg
= {
3162 .fc_metric
= IP6_RT_PRIO_USER
,
3163 .fc_ifindex
= dev
->ifindex
,
3164 .fc_dst_len
= prefixlen
,
3165 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
3166 RTF_UP
| RTF_PREF(pref
),
3167 .fc_protocol
= RTPROT_RA
,
3168 .fc_nlinfo
.portid
= 0,
3169 .fc_nlinfo
.nlh
= NULL
,
3170 .fc_nlinfo
.nl_net
= net
,
3173 cfg
.fc_table
= l3mdev_fib_table(dev
) ? : RT6_TABLE_INFO
,
3174 cfg
.fc_dst
= *prefix
;
3175 cfg
.fc_gateway
= *gwaddr
;
3177 /* We should treat it as a default route if prefix length is 0. */
3179 cfg
.fc_flags
|= RTF_DEFAULT
;
3181 ip6_route_add(&cfg
, NULL
);
3183 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, dev
);
3187 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
3189 u32 tb_id
= l3mdev_fib_table(dev
) ? : RT6_TABLE_DFLT
;
3190 struct rt6_info
*rt
;
3191 struct fib6_table
*table
;
3193 table
= fib6_get_table(dev_net(dev
), tb_id
);
3198 for_each_fib6_node_rt_rcu(&table
->tb6_root
) {
3199 if (dev
== rt
->dst
.dev
&&
3200 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
3201 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
3205 ip6_hold_safe(NULL
, &rt
, false);
3210 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
3211 struct net_device
*dev
,
3214 struct fib6_config cfg
= {
3215 .fc_table
= l3mdev_fib_table(dev
) ? : RT6_TABLE_DFLT
,
3216 .fc_metric
= IP6_RT_PRIO_USER
,
3217 .fc_ifindex
= dev
->ifindex
,
3218 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
3219 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
3220 .fc_protocol
= RTPROT_RA
,
3221 .fc_nlinfo
.portid
= 0,
3222 .fc_nlinfo
.nlh
= NULL
,
3223 .fc_nlinfo
.nl_net
= dev_net(dev
),
3226 cfg
.fc_gateway
= *gwaddr
;
3228 if (!ip6_route_add(&cfg
, NULL
)) {
3229 struct fib6_table
*table
;
3231 table
= fib6_get_table(dev_net(dev
), cfg
.fc_table
);
3233 table
->flags
|= RT6_TABLE_HAS_DFLT_ROUTER
;
3236 return rt6_get_dflt_router(gwaddr
, dev
);
3239 static void __rt6_purge_dflt_routers(struct fib6_table
*table
)
3241 struct rt6_info
*rt
;
3245 for_each_fib6_node_rt_rcu(&table
->tb6_root
) {
3246 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
) &&
3247 (!rt
->rt6i_idev
|| rt
->rt6i_idev
->cnf
.accept_ra
!= 2)) {
3248 if (dst_hold_safe(&rt
->dst
)) {
3259 table
->flags
&= ~RT6_TABLE_HAS_DFLT_ROUTER
;
3262 void rt6_purge_dflt_routers(struct net
*net
)
3264 struct fib6_table
*table
;
3265 struct hlist_head
*head
;
3270 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
3271 head
= &net
->ipv6
.fib_table_hash
[h
];
3272 hlist_for_each_entry_rcu(table
, head
, tb6_hlist
) {
3273 if (table
->flags
& RT6_TABLE_HAS_DFLT_ROUTER
)
3274 __rt6_purge_dflt_routers(table
);
3281 static void rtmsg_to_fib6_config(struct net
*net
,
3282 struct in6_rtmsg
*rtmsg
,
3283 struct fib6_config
*cfg
)
3285 memset(cfg
, 0, sizeof(*cfg
));
3287 cfg
->fc_table
= l3mdev_fib_table_by_index(net
, rtmsg
->rtmsg_ifindex
) ?
3289 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
3290 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
3291 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
3292 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
3293 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
3294 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
3296 cfg
->fc_nlinfo
.nl_net
= net
;
3298 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
3299 cfg
->fc_src
= rtmsg
->rtmsg_src
;
3300 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
3303 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
3305 struct fib6_config cfg
;
3306 struct in6_rtmsg rtmsg
;
3310 case SIOCADDRT
: /* Add a route */
3311 case SIOCDELRT
: /* Delete a route */
3312 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
3314 err
= copy_from_user(&rtmsg
, arg
,
3315 sizeof(struct in6_rtmsg
));
3319 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
3324 err
= ip6_route_add(&cfg
, NULL
);
3327 err
= ip6_route_del(&cfg
, NULL
);
3341 * Drop the packet on the floor
3344 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
3347 struct dst_entry
*dst
= skb_dst(skb
);
3348 switch (ipstats_mib_noroutes
) {
3349 case IPSTATS_MIB_INNOROUTES
:
3350 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
3351 if (type
== IPV6_ADDR_ANY
) {
3352 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
3353 IPSTATS_MIB_INADDRERRORS
);
3357 case IPSTATS_MIB_OUTNOROUTES
:
3358 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
3359 ipstats_mib_noroutes
);
3362 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
3367 static int ip6_pkt_discard(struct sk_buff
*skb
)
3369 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
3372 static int ip6_pkt_discard_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
3374 skb
->dev
= skb_dst(skb
)->dev
;
3375 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
3378 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
3380 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
3383 static int ip6_pkt_prohibit_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
3385 skb
->dev
= skb_dst(skb
)->dev
;
3386 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
3390 * Allocate a dst for local (unicast / anycast) address.
3393 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
3394 const struct in6_addr
*addr
,
3398 struct net
*net
= dev_net(idev
->dev
);
3399 struct net_device
*dev
= idev
->dev
;
3400 struct rt6_info
*rt
;
3402 rt
= ip6_dst_alloc(net
, dev
, DST_NOCOUNT
);
3404 return ERR_PTR(-ENOMEM
);
3408 rt
->dst
.flags
|= DST_HOST
;
3409 rt
->dst
.input
= ip6_input
;
3410 rt
->dst
.output
= ip6_output
;
3411 rt
->rt6i_idev
= idev
;
3413 rt
->rt6i_protocol
= RTPROT_KERNEL
;
3414 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
3416 rt
->rt6i_flags
|= RTF_ANYCAST
;
3418 rt
->rt6i_flags
|= RTF_LOCAL
;
3420 rt
->rt6i_gateway
= *addr
;
3421 rt
->rt6i_dst
.addr
= *addr
;
3422 rt
->rt6i_dst
.plen
= 128;
3423 tb_id
= l3mdev_fib_table(idev
->dev
) ? : RT6_TABLE_LOCAL
;
3424 rt
->rt6i_table
= fib6_get_table(net
, tb_id
);
3429 /* remove deleted ip from prefsrc entries */
3430 struct arg_dev_net_ip
{
3431 struct net_device
*dev
;
3433 struct in6_addr
*addr
;
3436 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
3438 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
3439 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
3440 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
3442 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
3443 rt
!= net
->ipv6
.ip6_null_entry
&&
3444 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
3445 spin_lock_bh(&rt6_exception_lock
);
3446 /* remove prefsrc entry */
3447 rt
->rt6i_prefsrc
.plen
= 0;
3448 /* need to update cache as well */
3449 rt6_exceptions_remove_prefsrc(rt
);
3450 spin_unlock_bh(&rt6_exception_lock
);
3455 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
3457 struct net
*net
= dev_net(ifp
->idev
->dev
);
3458 struct arg_dev_net_ip adni
= {
3459 .dev
= ifp
->idev
->dev
,
3463 fib6_clean_all(net
, fib6_remove_prefsrc
, &adni
);
3466 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
3468 /* Remove routers and update dst entries when gateway turn into host. */
3469 static int fib6_clean_tohost(struct rt6_info
*rt
, void *arg
)
3471 struct in6_addr
*gateway
= (struct in6_addr
*)arg
;
3473 if (((rt
->rt6i_flags
& RTF_RA_ROUTER
) == RTF_RA_ROUTER
) &&
3474 ipv6_addr_equal(gateway
, &rt
->rt6i_gateway
)) {
3478 /* Further clean up cached routes in exception table.
3479 * This is needed because cached route may have a different
3480 * gateway than its 'parent' in the case of an ip redirect.
3482 rt6_exceptions_clean_tohost(rt
, gateway
);
3487 void rt6_clean_tohost(struct net
*net
, struct in6_addr
*gateway
)
3489 fib6_clean_all(net
, fib6_clean_tohost
, gateway
);
3492 struct arg_dev_net
{
3493 struct net_device
*dev
;
3497 /* called with write lock held for table with rt */
3498 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
3500 const struct arg_dev_net
*adn
= arg
;
3501 const struct net_device
*dev
= adn
->dev
;
3503 if ((rt
->dst
.dev
== dev
|| !dev
) &&
3504 rt
!= adn
->net
->ipv6
.ip6_null_entry
&&
3505 (rt
->rt6i_nsiblings
== 0 ||
3506 (dev
&& netdev_unregistering(dev
)) ||
3507 !rt
->rt6i_idev
->cnf
.ignore_routes_with_linkdown
))
3513 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
3515 struct arg_dev_net adn
= {
3520 fib6_clean_all(net
, fib6_ifdown
, &adn
);
3522 rt6_uncached_list_flush_dev(net
, dev
);
3525 struct rt6_mtu_change_arg
{
3526 struct net_device
*dev
;
3530 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
3532 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
3533 struct inet6_dev
*idev
;
3535 /* In IPv6 pmtu discovery is not optional,
3536 so that RTAX_MTU lock cannot disable it.
3537 We still use this lock to block changes
3538 caused by addrconf/ndisc.
3541 idev
= __in6_dev_get(arg
->dev
);
3545 /* For administrative MTU increase, there is no way to discover
3546 IPv6 PMTU increase, so PMTU increase should be updated here.
3547 Since RFC 1981 doesn't include administrative MTU increase
3548 update PMTU increase is a MUST. (i.e. jumbo frame)
3550 if (rt
->dst
.dev
== arg
->dev
&&
3551 !dst_metric_locked(&rt
->dst
, RTAX_MTU
)) {
3552 spin_lock_bh(&rt6_exception_lock
);
3553 if (dst_metric_raw(&rt
->dst
, RTAX_MTU
) &&
3554 rt6_mtu_change_route_allowed(idev
, rt
, arg
->mtu
))
3555 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
3556 rt6_exceptions_update_pmtu(idev
, rt
, arg
->mtu
);
3557 spin_unlock_bh(&rt6_exception_lock
);
3562 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
3564 struct rt6_mtu_change_arg arg
= {
3569 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, &arg
);
3572 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
3573 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
3574 [RTA_OIF
] = { .type
= NLA_U32
},
3575 [RTA_IIF
] = { .type
= NLA_U32
},
3576 [RTA_PRIORITY
] = { .type
= NLA_U32
},
3577 [RTA_METRICS
] = { .type
= NLA_NESTED
},
3578 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
3579 [RTA_PREF
] = { .type
= NLA_U8
},
3580 [RTA_ENCAP_TYPE
] = { .type
= NLA_U16
},
3581 [RTA_ENCAP
] = { .type
= NLA_NESTED
},
3582 [RTA_EXPIRES
] = { .type
= NLA_U32
},
3583 [RTA_UID
] = { .type
= NLA_U32
},
3584 [RTA_MARK
] = { .type
= NLA_U32
},
3587 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3588 struct fib6_config
*cfg
,
3589 struct netlink_ext_ack
*extack
)
3592 struct nlattr
*tb
[RTA_MAX
+1];
3596 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
,
3602 rtm
= nlmsg_data(nlh
);
3603 memset(cfg
, 0, sizeof(*cfg
));
3605 cfg
->fc_table
= rtm
->rtm_table
;
3606 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
3607 cfg
->fc_src_len
= rtm
->rtm_src_len
;
3608 cfg
->fc_flags
= RTF_UP
;
3609 cfg
->fc_protocol
= rtm
->rtm_protocol
;
3610 cfg
->fc_type
= rtm
->rtm_type
;
3612 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
3613 rtm
->rtm_type
== RTN_BLACKHOLE
||
3614 rtm
->rtm_type
== RTN_PROHIBIT
||
3615 rtm
->rtm_type
== RTN_THROW
)
3616 cfg
->fc_flags
|= RTF_REJECT
;
3618 if (rtm
->rtm_type
== RTN_LOCAL
)
3619 cfg
->fc_flags
|= RTF_LOCAL
;
3621 if (rtm
->rtm_flags
& RTM_F_CLONED
)
3622 cfg
->fc_flags
|= RTF_CACHE
;
3624 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
3625 cfg
->fc_nlinfo
.nlh
= nlh
;
3626 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
3628 if (tb
[RTA_GATEWAY
]) {
3629 cfg
->fc_gateway
= nla_get_in6_addr(tb
[RTA_GATEWAY
]);
3630 cfg
->fc_flags
|= RTF_GATEWAY
;
3634 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
3636 if (nla_len(tb
[RTA_DST
]) < plen
)
3639 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
3643 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
3645 if (nla_len(tb
[RTA_SRC
]) < plen
)
3648 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
3651 if (tb
[RTA_PREFSRC
])
3652 cfg
->fc_prefsrc
= nla_get_in6_addr(tb
[RTA_PREFSRC
]);
3655 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
3657 if (tb
[RTA_PRIORITY
])
3658 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
3660 if (tb
[RTA_METRICS
]) {
3661 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
3662 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
3666 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
3668 if (tb
[RTA_MULTIPATH
]) {
3669 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
3670 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
3672 err
= lwtunnel_valid_encap_type_attr(cfg
->fc_mp
,
3673 cfg
->fc_mp_len
, extack
);
3679 pref
= nla_get_u8(tb
[RTA_PREF
]);
3680 if (pref
!= ICMPV6_ROUTER_PREF_LOW
&&
3681 pref
!= ICMPV6_ROUTER_PREF_HIGH
)
3682 pref
= ICMPV6_ROUTER_PREF_MEDIUM
;
3683 cfg
->fc_flags
|= RTF_PREF(pref
);
3687 cfg
->fc_encap
= tb
[RTA_ENCAP
];
3689 if (tb
[RTA_ENCAP_TYPE
]) {
3690 cfg
->fc_encap_type
= nla_get_u16(tb
[RTA_ENCAP_TYPE
]);
3692 err
= lwtunnel_valid_encap_type(cfg
->fc_encap_type
, extack
);
3697 if (tb
[RTA_EXPIRES
]) {
3698 unsigned long timeout
= addrconf_timeout_fixup(nla_get_u32(tb
[RTA_EXPIRES
]), HZ
);
3700 if (addrconf_finite_timeout(timeout
)) {
3701 cfg
->fc_expires
= jiffies_to_clock_t(timeout
* HZ
);
3702 cfg
->fc_flags
|= RTF_EXPIRES
;
3712 struct rt6_info
*rt6_info
;
3713 struct fib6_config r_cfg
;
3714 struct mx6_config mxc
;
3715 struct list_head next
;
3718 static void ip6_print_replace_route_err(struct list_head
*rt6_nh_list
)
3722 list_for_each_entry(nh
, rt6_nh_list
, next
) {
3723 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
3724 &nh
->r_cfg
.fc_dst
, &nh
->r_cfg
.fc_gateway
,
3725 nh
->r_cfg
.fc_ifindex
);
3729 static int ip6_route_info_append(struct list_head
*rt6_nh_list
,
3730 struct rt6_info
*rt
, struct fib6_config
*r_cfg
)
3735 list_for_each_entry(nh
, rt6_nh_list
, next
) {
3736 /* check if rt6_info already exists */
3737 if (rt6_duplicate_nexthop(nh
->rt6_info
, rt
))
3741 nh
= kzalloc(sizeof(*nh
), GFP_KERNEL
);
3745 err
= ip6_convert_metrics(&nh
->mxc
, r_cfg
);
3750 memcpy(&nh
->r_cfg
, r_cfg
, sizeof(*r_cfg
));
3751 list_add_tail(&nh
->next
, rt6_nh_list
);
3756 static void ip6_route_mpath_notify(struct rt6_info
*rt
,
3757 struct rt6_info
*rt_last
,
3758 struct nl_info
*info
,
3761 /* if this is an APPEND route, then rt points to the first route
3762 * inserted and rt_last points to last route inserted. Userspace
3763 * wants a consistent dump of the route which starts at the first
3764 * nexthop. Since sibling routes are always added at the end of
3765 * the list, find the first sibling of the last route appended
3767 if ((nlflags
& NLM_F_APPEND
) && rt_last
&& rt_last
->rt6i_nsiblings
) {
3768 rt
= list_first_entry(&rt_last
->rt6i_siblings
,
3774 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
, nlflags
);
3777 static int ip6_route_multipath_add(struct fib6_config
*cfg
,
3778 struct netlink_ext_ack
*extack
)
3780 struct rt6_info
*rt_notif
= NULL
, *rt_last
= NULL
;
3781 struct nl_info
*info
= &cfg
->fc_nlinfo
;
3782 struct fib6_config r_cfg
;
3783 struct rtnexthop
*rtnh
;
3784 struct rt6_info
*rt
;
3785 struct rt6_nh
*err_nh
;
3786 struct rt6_nh
*nh
, *nh_safe
;
3792 int replace
= (cfg
->fc_nlinfo
.nlh
&&
3793 (cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_REPLACE
));
3794 LIST_HEAD(rt6_nh_list
);
3796 nlflags
= replace
? NLM_F_REPLACE
: NLM_F_CREATE
;
3797 if (info
->nlh
&& info
->nlh
->nlmsg_flags
& NLM_F_APPEND
)
3798 nlflags
|= NLM_F_APPEND
;
3800 remaining
= cfg
->fc_mp_len
;
3801 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
3803 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3804 * rt6_info structs per nexthop
3806 while (rtnh_ok(rtnh
, remaining
)) {
3807 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
3808 if (rtnh
->rtnh_ifindex
)
3809 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
3811 attrlen
= rtnh_attrlen(rtnh
);
3813 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
3815 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
3817 r_cfg
.fc_gateway
= nla_get_in6_addr(nla
);
3818 r_cfg
.fc_flags
|= RTF_GATEWAY
;
3820 r_cfg
.fc_encap
= nla_find(attrs
, attrlen
, RTA_ENCAP
);
3821 nla
= nla_find(attrs
, attrlen
, RTA_ENCAP_TYPE
);
3823 r_cfg
.fc_encap_type
= nla_get_u16(nla
);
3826 rt
= ip6_route_info_create(&r_cfg
, extack
);
3833 err
= ip6_route_info_append(&rt6_nh_list
, rt
, &r_cfg
);
3835 dst_release_immediate(&rt
->dst
);
3839 rtnh
= rtnh_next(rtnh
, &remaining
);
3842 /* for add and replace send one notification with all nexthops.
3843 * Skip the notification in fib6_add_rt2node and send one with
3844 * the full route when done
3846 info
->skip_notify
= 1;
3849 list_for_each_entry(nh
, &rt6_nh_list
, next
) {
3850 rt_last
= nh
->rt6_info
;
3851 err
= __ip6_ins_rt(nh
->rt6_info
, info
, &nh
->mxc
, extack
);
3852 /* save reference to first route for notification */
3853 if (!rt_notif
&& !err
)
3854 rt_notif
= nh
->rt6_info
;
3856 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3857 nh
->rt6_info
= NULL
;
3860 ip6_print_replace_route_err(&rt6_nh_list
);
3865 /* Because each route is added like a single route we remove
3866 * these flags after the first nexthop: if there is a collision,
3867 * we have already failed to add the first nexthop:
3868 * fib6_add_rt2node() has rejected it; when replacing, old
3869 * nexthops have been replaced by first new, the rest should
3872 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~(NLM_F_EXCL
|
3877 /* success ... tell user about new route */
3878 ip6_route_mpath_notify(rt_notif
, rt_last
, info
, nlflags
);
3882 /* send notification for routes that were added so that
3883 * the delete notifications sent by ip6_route_del are
3887 ip6_route_mpath_notify(rt_notif
, rt_last
, info
, nlflags
);
3889 /* Delete routes that were already added */
3890 list_for_each_entry(nh
, &rt6_nh_list
, next
) {
3893 ip6_route_del(&nh
->r_cfg
, extack
);
3897 list_for_each_entry_safe(nh
, nh_safe
, &rt6_nh_list
, next
) {
3899 dst_release_immediate(&nh
->rt6_info
->dst
);
3901 list_del(&nh
->next
);
3908 static int ip6_route_multipath_del(struct fib6_config
*cfg
,
3909 struct netlink_ext_ack
*extack
)
3911 struct fib6_config r_cfg
;
3912 struct rtnexthop
*rtnh
;
3915 int err
= 1, last_err
= 0;
3917 remaining
= cfg
->fc_mp_len
;
3918 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
3920 /* Parse a Multipath Entry */
3921 while (rtnh_ok(rtnh
, remaining
)) {
3922 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
3923 if (rtnh
->rtnh_ifindex
)
3924 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
3926 attrlen
= rtnh_attrlen(rtnh
);
3928 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
3930 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
3932 nla_memcpy(&r_cfg
.fc_gateway
, nla
, 16);
3933 r_cfg
.fc_flags
|= RTF_GATEWAY
;
3936 err
= ip6_route_del(&r_cfg
, extack
);
3940 rtnh
= rtnh_next(rtnh
, &remaining
);
3946 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3947 struct netlink_ext_ack
*extack
)
3949 struct fib6_config cfg
;
3952 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
, extack
);
3957 return ip6_route_multipath_del(&cfg
, extack
);
3959 cfg
.fc_delete_all_nh
= 1;
3960 return ip6_route_del(&cfg
, extack
);
3964 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3965 struct netlink_ext_ack
*extack
)
3967 struct fib6_config cfg
;
3970 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
, extack
);
3975 return ip6_route_multipath_add(&cfg
, extack
);
3977 return ip6_route_add(&cfg
, extack
);
3980 static size_t rt6_nlmsg_size(struct rt6_info
*rt
)
3982 int nexthop_len
= 0;
3984 if (rt
->rt6i_nsiblings
) {
3985 nexthop_len
= nla_total_size(0) /* RTA_MULTIPATH */
3986 + NLA_ALIGN(sizeof(struct rtnexthop
))
3987 + nla_total_size(16) /* RTA_GATEWAY */
3988 + lwtunnel_get_encap_size(rt
->dst
.lwtstate
);
3990 nexthop_len
*= rt
->rt6i_nsiblings
;
3993 return NLMSG_ALIGN(sizeof(struct rtmsg
))
3994 + nla_total_size(16) /* RTA_SRC */
3995 + nla_total_size(16) /* RTA_DST */
3996 + nla_total_size(16) /* RTA_GATEWAY */
3997 + nla_total_size(16) /* RTA_PREFSRC */
3998 + nla_total_size(4) /* RTA_TABLE */
3999 + nla_total_size(4) /* RTA_IIF */
4000 + nla_total_size(4) /* RTA_OIF */
4001 + nla_total_size(4) /* RTA_PRIORITY */
4002 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
4003 + nla_total_size(sizeof(struct rta_cacheinfo
))
4004 + nla_total_size(TCP_CA_NAME_MAX
) /* RTAX_CC_ALGO */
4005 + nla_total_size(1) /* RTA_PREF */
4006 + lwtunnel_get_encap_size(rt
->dst
.lwtstate
)
4010 static int rt6_nexthop_info(struct sk_buff
*skb
, struct rt6_info
*rt
,
4011 unsigned int *flags
, bool skip_oif
)
4013 if (!netif_running(rt
->dst
.dev
) || !netif_carrier_ok(rt
->dst
.dev
)) {
4014 *flags
|= RTNH_F_LINKDOWN
;
4015 if (rt
->rt6i_idev
->cnf
.ignore_routes_with_linkdown
)
4016 *flags
|= RTNH_F_DEAD
;
4019 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
4020 if (nla_put_in6_addr(skb
, RTA_GATEWAY
, &rt
->rt6i_gateway
) < 0)
4021 goto nla_put_failure
;
4024 if (rt
->rt6i_nh_flags
& RTNH_F_OFFLOAD
)
4025 *flags
|= RTNH_F_OFFLOAD
;
4027 /* not needed for multipath encoding b/c it has a rtnexthop struct */
4028 if (!skip_oif
&& rt
->dst
.dev
&&
4029 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
4030 goto nla_put_failure
;
4032 if (rt
->dst
.lwtstate
&&
4033 lwtunnel_fill_encap(skb
, rt
->dst
.lwtstate
) < 0)
4034 goto nla_put_failure
;
4042 /* add multipath next hop */
4043 static int rt6_add_nexthop(struct sk_buff
*skb
, struct rt6_info
*rt
)
4045 struct rtnexthop
*rtnh
;
4046 unsigned int flags
= 0;
4048 rtnh
= nla_reserve_nohdr(skb
, sizeof(*rtnh
));
4050 goto nla_put_failure
;
4052 rtnh
->rtnh_hops
= 0;
4053 rtnh
->rtnh_ifindex
= rt
->dst
.dev
? rt
->dst
.dev
->ifindex
: 0;
4055 if (rt6_nexthop_info(skb
, rt
, &flags
, true) < 0)
4056 goto nla_put_failure
;
4058 rtnh
->rtnh_flags
= flags
;
4060 /* length of rtnetlink header + attributes */
4061 rtnh
->rtnh_len
= nlmsg_get_pos(skb
) - (void *)rtnh
;
4069 static int rt6_fill_node(struct net
*net
,
4070 struct sk_buff
*skb
, struct rt6_info
*rt
,
4071 struct in6_addr
*dst
, struct in6_addr
*src
,
4072 int iif
, int type
, u32 portid
, u32 seq
,
4075 u32 metrics
[RTAX_MAX
];
4077 struct nlmsghdr
*nlh
;
4081 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
4085 rtm
= nlmsg_data(nlh
);
4086 rtm
->rtm_family
= AF_INET6
;
4087 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
4088 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
4091 table
= rt
->rt6i_table
->tb6_id
;
4093 table
= RT6_TABLE_UNSPEC
;
4094 rtm
->rtm_table
= table
;
4095 if (nla_put_u32(skb
, RTA_TABLE
, table
))
4096 goto nla_put_failure
;
4097 if (rt
->rt6i_flags
& RTF_REJECT
) {
4098 switch (rt
->dst
.error
) {
4100 rtm
->rtm_type
= RTN_BLACKHOLE
;
4103 rtm
->rtm_type
= RTN_PROHIBIT
;
4106 rtm
->rtm_type
= RTN_THROW
;
4109 rtm
->rtm_type
= RTN_UNREACHABLE
;
4113 else if (rt
->rt6i_flags
& RTF_LOCAL
)
4114 rtm
->rtm_type
= RTN_LOCAL
;
4115 else if (rt
->rt6i_flags
& RTF_ANYCAST
)
4116 rtm
->rtm_type
= RTN_ANYCAST
;
4117 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
4118 rtm
->rtm_type
= RTN_LOCAL
;
4120 rtm
->rtm_type
= RTN_UNICAST
;
4122 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
4123 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
4125 if (rt
->rt6i_flags
& RTF_CACHE
)
4126 rtm
->rtm_flags
|= RTM_F_CLONED
;
4129 if (nla_put_in6_addr(skb
, RTA_DST
, dst
))
4130 goto nla_put_failure
;
4131 rtm
->rtm_dst_len
= 128;
4132 } else if (rtm
->rtm_dst_len
)
4133 if (nla_put_in6_addr(skb
, RTA_DST
, &rt
->rt6i_dst
.addr
))
4134 goto nla_put_failure
;
4135 #ifdef CONFIG_IPV6_SUBTREES
4137 if (nla_put_in6_addr(skb
, RTA_SRC
, src
))
4138 goto nla_put_failure
;
4139 rtm
->rtm_src_len
= 128;
4140 } else if (rtm
->rtm_src_len
&&
4141 nla_put_in6_addr(skb
, RTA_SRC
, &rt
->rt6i_src
.addr
))
4142 goto nla_put_failure
;
4145 #ifdef CONFIG_IPV6_MROUTE
4146 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
4147 int err
= ip6mr_get_route(net
, skb
, rtm
, portid
);
4152 goto nla_put_failure
;
4155 if (nla_put_u32(skb
, RTA_IIF
, iif
))
4156 goto nla_put_failure
;
4158 struct in6_addr saddr_buf
;
4159 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
4160 nla_put_in6_addr(skb
, RTA_PREFSRC
, &saddr_buf
))
4161 goto nla_put_failure
;
4164 if (rt
->rt6i_prefsrc
.plen
) {
4165 struct in6_addr saddr_buf
;
4166 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
4167 if (nla_put_in6_addr(skb
, RTA_PREFSRC
, &saddr_buf
))
4168 goto nla_put_failure
;
4171 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
4173 metrics
[RTAX_MTU
- 1] = rt
->rt6i_pmtu
;
4174 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
4175 goto nla_put_failure
;
4177 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
4178 goto nla_put_failure
;
4180 /* For multipath routes, walk the siblings list and add
4181 * each as a nexthop within RTA_MULTIPATH.
4183 if (rt
->rt6i_nsiblings
) {
4184 struct rt6_info
*sibling
, *next_sibling
;
4187 mp
= nla_nest_start(skb
, RTA_MULTIPATH
);
4189 goto nla_put_failure
;
4191 if (rt6_add_nexthop(skb
, rt
) < 0)
4192 goto nla_put_failure
;
4194 list_for_each_entry_safe(sibling
, next_sibling
,
4195 &rt
->rt6i_siblings
, rt6i_siblings
) {
4196 if (rt6_add_nexthop(skb
, sibling
) < 0)
4197 goto nla_put_failure
;
4200 nla_nest_end(skb
, mp
);
4202 if (rt6_nexthop_info(skb
, rt
, &rtm
->rtm_flags
, false) < 0)
4203 goto nla_put_failure
;
4206 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
4208 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
4209 goto nla_put_failure
;
4211 if (nla_put_u8(skb
, RTA_PREF
, IPV6_EXTRACT_PREF(rt
->rt6i_flags
)))
4212 goto nla_put_failure
;
4215 nlmsg_end(skb
, nlh
);
4219 nlmsg_cancel(skb
, nlh
);
4223 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
4225 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
4226 struct net
*net
= arg
->net
;
4228 if (rt
== net
->ipv6
.ip6_null_entry
)
4231 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
4232 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
4234 /* user wants prefix routes only */
4235 if (rtm
->rtm_flags
& RTM_F_PREFIX
&&
4236 !(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
4237 /* success since this is not a prefix route */
4242 return rt6_fill_node(net
,
4243 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
4244 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
4248 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
,
4249 struct netlink_ext_ack
*extack
)
4251 struct net
*net
= sock_net(in_skb
->sk
);
4252 struct nlattr
*tb
[RTA_MAX
+1];
4253 int err
, iif
= 0, oif
= 0;
4254 struct dst_entry
*dst
;
4255 struct rt6_info
*rt
;
4256 struct sk_buff
*skb
;
4261 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
,
4267 memset(&fl6
, 0, sizeof(fl6
));
4268 rtm
= nlmsg_data(nlh
);
4269 fl6
.flowlabel
= ip6_make_flowinfo(rtm
->rtm_tos
, 0);
4270 fibmatch
= !!(rtm
->rtm_flags
& RTM_F_FIB_MATCH
);
4273 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
4276 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
4280 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
4283 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
4287 iif
= nla_get_u32(tb
[RTA_IIF
]);
4290 oif
= nla_get_u32(tb
[RTA_OIF
]);
4293 fl6
.flowi6_mark
= nla_get_u32(tb
[RTA_MARK
]);
4296 fl6
.flowi6_uid
= make_kuid(current_user_ns(),
4297 nla_get_u32(tb
[RTA_UID
]));
4299 fl6
.flowi6_uid
= iif
? INVALID_UID
: current_uid();
4302 struct net_device
*dev
;
4307 dev
= dev_get_by_index_rcu(net
, iif
);
4314 fl6
.flowi6_iif
= iif
;
4316 if (!ipv6_addr_any(&fl6
.saddr
))
4317 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
4319 dst
= ip6_route_input_lookup(net
, dev
, &fl6
, flags
);
4323 fl6
.flowi6_oif
= oif
;
4325 dst
= ip6_route_output(net
, NULL
, &fl6
);
4329 rt
= container_of(dst
, struct rt6_info
, dst
);
4330 if (rt
->dst
.error
) {
4331 err
= rt
->dst
.error
;
4336 if (rt
== net
->ipv6
.ip6_null_entry
) {
4337 err
= rt
->dst
.error
;
4342 if (fibmatch
&& rt
->dst
.from
) {
4343 struct rt6_info
*ort
= container_of(rt
->dst
.from
,
4344 struct rt6_info
, dst
);
4346 dst_hold(&ort
->dst
);
4351 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
4358 skb_dst_set(skb
, &rt
->dst
);
4360 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, iif
,
4361 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
4364 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
4365 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
4372 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
4377 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
,
4378 unsigned int nlm_flags
)
4380 struct sk_buff
*skb
;
4381 struct net
*net
= info
->nl_net
;
4386 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
4388 skb
= nlmsg_new(rt6_nlmsg_size(rt
), gfp_any());
4392 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
4393 event
, info
->portid
, seq
, nlm_flags
);
4395 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
4396 WARN_ON(err
== -EMSGSIZE
);
4400 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
4401 info
->nlh
, gfp_any());
4405 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
4408 static int ip6_route_dev_notify(struct notifier_block
*this,
4409 unsigned long event
, void *ptr
)
4411 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
4412 struct net
*net
= dev_net(dev
);
4414 if (!(dev
->flags
& IFF_LOOPBACK
))
4417 if (event
== NETDEV_REGISTER
) {
4418 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
4419 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
4420 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4421 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
4422 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
4423 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
4424 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
4426 } else if (event
== NETDEV_UNREGISTER
&&
4427 dev
->reg_state
!= NETREG_UNREGISTERED
) {
4428 /* NETDEV_UNREGISTER could be fired for multiple times by
4429 * netdev_wait_allrefs(). Make sure we only call this once.
4431 in6_dev_put_clear(&net
->ipv6
.ip6_null_entry
->rt6i_idev
);
4432 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4433 in6_dev_put_clear(&net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
);
4434 in6_dev_put_clear(&net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
);
4445 #ifdef CONFIG_PROC_FS
4447 static const struct file_operations ipv6_route_proc_fops
= {
4448 .owner
= THIS_MODULE
,
4449 .open
= ipv6_route_open
,
4451 .llseek
= seq_lseek
,
4452 .release
= seq_release_net
,
4455 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
4457 struct net
*net
= (struct net
*)seq
->private;
4458 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
4459 net
->ipv6
.rt6_stats
->fib_nodes
,
4460 net
->ipv6
.rt6_stats
->fib_route_nodes
,
4461 atomic_read(&net
->ipv6
.rt6_stats
->fib_rt_alloc
),
4462 net
->ipv6
.rt6_stats
->fib_rt_entries
,
4463 net
->ipv6
.rt6_stats
->fib_rt_cache
,
4464 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
4465 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
4470 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
4472 return single_open_net(inode
, file
, rt6_stats_seq_show
);
4475 static const struct file_operations rt6_stats_seq_fops
= {
4476 .owner
= THIS_MODULE
,
4477 .open
= rt6_stats_seq_open
,
4479 .llseek
= seq_lseek
,
4480 .release
= single_release_net
,
4482 #endif /* CONFIG_PROC_FS */
4484 #ifdef CONFIG_SYSCTL
4487 int ipv6_sysctl_rtcache_flush(struct ctl_table
*ctl
, int write
,
4488 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
4495 net
= (struct net
*)ctl
->extra1
;
4496 delay
= net
->ipv6
.sysctl
.flush_delay
;
4497 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
4498 fib6_run_gc(delay
<= 0 ? 0 : (unsigned long)delay
, net
, delay
> 0);
4502 struct ctl_table ipv6_route_table_template
[] = {
4504 .procname
= "flush",
4505 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
4506 .maxlen
= sizeof(int),
4508 .proc_handler
= ipv6_sysctl_rtcache_flush
4511 .procname
= "gc_thresh",
4512 .data
= &ip6_dst_ops_template
.gc_thresh
,
4513 .maxlen
= sizeof(int),
4515 .proc_handler
= proc_dointvec
,
4518 .procname
= "max_size",
4519 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
4520 .maxlen
= sizeof(int),
4522 .proc_handler
= proc_dointvec
,
4525 .procname
= "gc_min_interval",
4526 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
4527 .maxlen
= sizeof(int),
4529 .proc_handler
= proc_dointvec_jiffies
,
4532 .procname
= "gc_timeout",
4533 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
4534 .maxlen
= sizeof(int),
4536 .proc_handler
= proc_dointvec_jiffies
,
4539 .procname
= "gc_interval",
4540 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
4541 .maxlen
= sizeof(int),
4543 .proc_handler
= proc_dointvec_jiffies
,
4546 .procname
= "gc_elasticity",
4547 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
4548 .maxlen
= sizeof(int),
4550 .proc_handler
= proc_dointvec
,
4553 .procname
= "mtu_expires",
4554 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
4555 .maxlen
= sizeof(int),
4557 .proc_handler
= proc_dointvec_jiffies
,
4560 .procname
= "min_adv_mss",
4561 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
4562 .maxlen
= sizeof(int),
4564 .proc_handler
= proc_dointvec
,
4567 .procname
= "gc_min_interval_ms",
4568 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
4569 .maxlen
= sizeof(int),
4571 .proc_handler
= proc_dointvec_ms_jiffies
,
4576 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
4578 struct ctl_table
*table
;
4580 table
= kmemdup(ipv6_route_table_template
,
4581 sizeof(ipv6_route_table_template
),
4585 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
4586 table
[0].extra1
= net
;
4587 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
4588 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
4589 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
4590 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
4591 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
4592 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
4593 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
4594 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
4595 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
4597 /* Don't export sysctls to unprivileged users */
4598 if (net
->user_ns
!= &init_user_ns
)
4599 table
[0].procname
= NULL
;
4606 static int __net_init
ip6_route_net_init(struct net
*net
)
4610 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
4611 sizeof(net
->ipv6
.ip6_dst_ops
));
4613 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
4614 goto out_ip6_dst_ops
;
4616 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
4617 sizeof(*net
->ipv6
.ip6_null_entry
),
4619 if (!net
->ipv6
.ip6_null_entry
)
4620 goto out_ip6_dst_entries
;
4621 net
->ipv6
.ip6_null_entry
->dst
.path
=
4622 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
4623 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
4624 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
4625 ip6_template_metrics
, true);
4627 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4628 net
->ipv6
.fib6_has_custom_rules
= false;
4629 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
4630 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
4632 if (!net
->ipv6
.ip6_prohibit_entry
)
4633 goto out_ip6_null_entry
;
4634 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
4635 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
4636 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
4637 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
4638 ip6_template_metrics
, true);
4640 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
4641 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
4643 if (!net
->ipv6
.ip6_blk_hole_entry
)
4644 goto out_ip6_prohibit_entry
;
4645 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
4646 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
4647 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
4648 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
4649 ip6_template_metrics
, true);
4652 net
->ipv6
.sysctl
.flush_delay
= 0;
4653 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
4654 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
4655 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
4656 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
4657 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
4658 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
4659 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
4661 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
4667 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4668 out_ip6_prohibit_entry
:
4669 kfree(net
->ipv6
.ip6_prohibit_entry
);
4671 kfree(net
->ipv6
.ip6_null_entry
);
4673 out_ip6_dst_entries
:
4674 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
4679 static void __net_exit
ip6_route_net_exit(struct net
*net
)
4681 kfree(net
->ipv6
.ip6_null_entry
);
4682 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4683 kfree(net
->ipv6
.ip6_prohibit_entry
);
4684 kfree(net
->ipv6
.ip6_blk_hole_entry
);
4686 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
4689 static int __net_init
ip6_route_net_init_late(struct net
*net
)
4691 #ifdef CONFIG_PROC_FS
4692 proc_create("ipv6_route", 0, net
->proc_net
, &ipv6_route_proc_fops
);
4693 proc_create("rt6_stats", S_IRUGO
, net
->proc_net
, &rt6_stats_seq_fops
);
4698 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
4700 #ifdef CONFIG_PROC_FS
4701 remove_proc_entry("ipv6_route", net
->proc_net
);
4702 remove_proc_entry("rt6_stats", net
->proc_net
);
4706 static struct pernet_operations ip6_route_net_ops
= {
4707 .init
= ip6_route_net_init
,
4708 .exit
= ip6_route_net_exit
,
4711 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
4713 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
4717 inet_peer_base_init(bp
);
4718 net
->ipv6
.peers
= bp
;
4722 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
4724 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
4726 net
->ipv6
.peers
= NULL
;
4727 inetpeer_invalidate_tree(bp
);
4731 static struct pernet_operations ipv6_inetpeer_ops
= {
4732 .init
= ipv6_inetpeer_init
,
4733 .exit
= ipv6_inetpeer_exit
,
4736 static struct pernet_operations ip6_route_net_late_ops
= {
4737 .init
= ip6_route_net_init_late
,
4738 .exit
= ip6_route_net_exit_late
,
4741 static struct notifier_block ip6_route_dev_notifier
= {
4742 .notifier_call
= ip6_route_dev_notify
,
4743 .priority
= ADDRCONF_NOTIFY_PRIORITY
- 10,
4746 void __init
ip6_route_init_special_entries(void)
4748 /* Registering of the loopback is done before this portion of code,
4749 * the loopback reference in rt6_info will not be taken, do it
4750 * manually for init_net */
4751 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
4752 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
4753 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4754 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
4755 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
4756 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
4757 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
4761 int __init
ip6_route_init(void)
4767 ip6_dst_ops_template
.kmem_cachep
=
4768 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
4769 SLAB_HWCACHE_ALIGN
, NULL
);
4770 if (!ip6_dst_ops_template
.kmem_cachep
)
4773 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
4775 goto out_kmem_cache
;
4777 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
4779 goto out_dst_entries
;
4781 ret
= register_pernet_subsys(&ip6_route_net_ops
);
4783 goto out_register_inetpeer
;
4785 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
4789 goto out_register_subsys
;
4795 ret
= fib6_rules_init();
4799 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
4801 goto fib6_rules_init
;
4804 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, 0) ||
4805 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, 0) ||
4806 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
,
4807 RTNL_FLAG_DOIT_UNLOCKED
))
4808 goto out_register_late_subsys
;
4810 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
4812 goto out_register_late_subsys
;
4814 for_each_possible_cpu(cpu
) {
4815 struct uncached_list
*ul
= per_cpu_ptr(&rt6_uncached_list
, cpu
);
4817 INIT_LIST_HEAD(&ul
->head
);
4818 spin_lock_init(&ul
->lock
);
4824 out_register_late_subsys
:
4825 unregister_pernet_subsys(&ip6_route_net_late_ops
);
4827 fib6_rules_cleanup();
4832 out_register_subsys
:
4833 unregister_pernet_subsys(&ip6_route_net_ops
);
4834 out_register_inetpeer
:
4835 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
4837 dst_entries_destroy(&ip6_dst_blackhole_ops
);
4839 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
4843 void ip6_route_cleanup(void)
4845 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
4846 unregister_pernet_subsys(&ip6_route_net_late_ops
);
4847 fib6_rules_cleanup();
4850 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
4851 unregister_pernet_subsys(&ip6_route_net_ops
);
4852 dst_entries_destroy(&ip6_dst_blackhole_ops
);
4853 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);