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 <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
69 RT6_NUD_FAIL_HARD
= -3,
70 RT6_NUD_FAIL_PROBE
= -2,
71 RT6_NUD_FAIL_DO_RR
= -1,
75 static void ip6_rt_copy_init(struct rt6_info
*rt
, struct rt6_info
*ort
);
76 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
77 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
78 static unsigned int ip6_mtu(const struct dst_entry
*dst
);
79 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
80 static void ip6_dst_destroy(struct dst_entry
*);
81 static void ip6_dst_ifdown(struct dst_entry
*,
82 struct net_device
*dev
, int how
);
83 static int ip6_dst_gc(struct dst_ops
*ops
);
85 static int ip6_pkt_discard(struct sk_buff
*skb
);
86 static int ip6_pkt_discard_out(struct sock
*sk
, struct sk_buff
*skb
);
87 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
88 static int ip6_pkt_prohibit_out(struct sock
*sk
, struct sk_buff
*skb
);
89 static void ip6_link_failure(struct sk_buff
*skb
);
90 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
91 struct sk_buff
*skb
, u32 mtu
);
92 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
94 static void rt6_dst_from_metrics_check(struct rt6_info
*rt
);
95 static int rt6_score_route(struct rt6_info
*rt
, int oif
, int strict
);
97 #ifdef CONFIG_IPV6_ROUTE_INFO
98 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
99 const struct in6_addr
*prefix
, int prefixlen
,
100 const struct in6_addr
*gwaddr
, int ifindex
,
102 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
103 const struct in6_addr
*prefix
, int prefixlen
,
104 const struct in6_addr
*gwaddr
, int ifindex
);
107 struct uncached_list
{
109 struct list_head head
;
112 static DEFINE_PER_CPU_ALIGNED(struct uncached_list
, rt6_uncached_list
);
114 static void rt6_uncached_list_add(struct rt6_info
*rt
)
116 struct uncached_list
*ul
= raw_cpu_ptr(&rt6_uncached_list
);
118 rt
->dst
.flags
|= DST_NOCACHE
;
119 rt
->rt6i_uncached_list
= ul
;
121 spin_lock_bh(&ul
->lock
);
122 list_add_tail(&rt
->rt6i_uncached
, &ul
->head
);
123 spin_unlock_bh(&ul
->lock
);
126 static void rt6_uncached_list_del(struct rt6_info
*rt
)
128 if (!list_empty(&rt
->rt6i_uncached
)) {
129 struct uncached_list
*ul
= rt
->rt6i_uncached_list
;
131 spin_lock_bh(&ul
->lock
);
132 list_del(&rt
->rt6i_uncached
);
133 spin_unlock_bh(&ul
->lock
);
137 static void rt6_uncached_list_flush_dev(struct net
*net
, struct net_device
*dev
)
139 struct net_device
*loopback_dev
= net
->loopback_dev
;
142 for_each_possible_cpu(cpu
) {
143 struct uncached_list
*ul
= per_cpu_ptr(&rt6_uncached_list
, cpu
);
146 spin_lock_bh(&ul
->lock
);
147 list_for_each_entry(rt
, &ul
->head
, rt6i_uncached
) {
148 struct inet6_dev
*rt_idev
= rt
->rt6i_idev
;
149 struct net_device
*rt_dev
= rt
->dst
.dev
;
151 if (rt_idev
&& (rt_idev
->dev
== dev
|| !dev
) &&
152 rt_idev
->dev
!= loopback_dev
) {
153 rt
->rt6i_idev
= in6_dev_get(loopback_dev
);
154 in6_dev_put(rt_idev
);
157 if (rt_dev
&& (rt_dev
== dev
|| !dev
) &&
158 rt_dev
!= loopback_dev
) {
159 rt
->dst
.dev
= loopback_dev
;
160 dev_hold(rt
->dst
.dev
);
164 spin_unlock_bh(&ul
->lock
);
168 static u32
*rt6_pcpu_cow_metrics(struct rt6_info
*rt
)
170 return dst_metrics_write_ptr(rt
->dst
.from
);
173 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
175 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
177 if (rt
->rt6i_flags
& RTF_PCPU
)
178 return rt6_pcpu_cow_metrics(rt
);
179 else if (rt
->rt6i_flags
& RTF_CACHE
)
182 return dst_cow_metrics_generic(dst
, old
);
185 static inline const void *choose_neigh_daddr(struct rt6_info
*rt
,
189 struct in6_addr
*p
= &rt
->rt6i_gateway
;
191 if (!ipv6_addr_any(p
))
192 return (const void *) p
;
194 return &ipv6_hdr(skb
)->daddr
;
198 static struct neighbour
*ip6_neigh_lookup(const struct dst_entry
*dst
,
202 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
205 daddr
= choose_neigh_daddr(rt
, skb
, daddr
);
206 n
= __ipv6_neigh_lookup(dst
->dev
, daddr
);
209 return neigh_create(&nd_tbl
, daddr
, dst
->dev
);
212 static struct dst_ops ip6_dst_ops_template
= {
216 .check
= ip6_dst_check
,
217 .default_advmss
= ip6_default_advmss
,
219 .cow_metrics
= ipv6_cow_metrics
,
220 .destroy
= ip6_dst_destroy
,
221 .ifdown
= ip6_dst_ifdown
,
222 .negative_advice
= ip6_negative_advice
,
223 .link_failure
= ip6_link_failure
,
224 .update_pmtu
= ip6_rt_update_pmtu
,
225 .redirect
= rt6_do_redirect
,
226 .local_out
= __ip6_local_out
,
227 .neigh_lookup
= ip6_neigh_lookup
,
230 static unsigned int ip6_blackhole_mtu(const struct dst_entry
*dst
)
232 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
234 return mtu
? : dst
->dev
->mtu
;
237 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
238 struct sk_buff
*skb
, u32 mtu
)
242 static void ip6_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
247 static u32
*ip6_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
253 static struct dst_ops ip6_dst_blackhole_ops
= {
255 .destroy
= ip6_dst_destroy
,
256 .check
= ip6_dst_check
,
257 .mtu
= ip6_blackhole_mtu
,
258 .default_advmss
= ip6_default_advmss
,
259 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
260 .redirect
= ip6_rt_blackhole_redirect
,
261 .cow_metrics
= ip6_rt_blackhole_cow_metrics
,
262 .neigh_lookup
= ip6_neigh_lookup
,
265 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
266 [RTAX_HOPLIMIT
- 1] = 0,
269 static const struct rt6_info ip6_null_entry_template
= {
271 .__refcnt
= ATOMIC_INIT(1),
273 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
274 .error
= -ENETUNREACH
,
275 .input
= ip6_pkt_discard
,
276 .output
= ip6_pkt_discard_out
,
278 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
279 .rt6i_protocol
= RTPROT_KERNEL
,
280 .rt6i_metric
= ~(u32
) 0,
281 .rt6i_ref
= ATOMIC_INIT(1),
284 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
286 static const struct rt6_info ip6_prohibit_entry_template
= {
288 .__refcnt
= ATOMIC_INIT(1),
290 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
292 .input
= ip6_pkt_prohibit
,
293 .output
= ip6_pkt_prohibit_out
,
295 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
296 .rt6i_protocol
= RTPROT_KERNEL
,
297 .rt6i_metric
= ~(u32
) 0,
298 .rt6i_ref
= ATOMIC_INIT(1),
301 static const struct rt6_info ip6_blk_hole_entry_template
= {
303 .__refcnt
= ATOMIC_INIT(1),
305 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
307 .input
= dst_discard
,
308 .output
= dst_discard_sk
,
310 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
311 .rt6i_protocol
= RTPROT_KERNEL
,
312 .rt6i_metric
= ~(u32
) 0,
313 .rt6i_ref
= ATOMIC_INIT(1),
318 /* allocate dst with ip6_dst_ops */
319 static struct rt6_info
*__ip6_dst_alloc(struct net
*net
,
320 struct net_device
*dev
,
322 struct fib6_table
*table
)
324 struct rt6_info
*rt
= dst_alloc(&net
->ipv6
.ip6_dst_ops
, dev
,
325 0, DST_OBSOLETE_FORCE_CHK
, flags
);
328 struct dst_entry
*dst
= &rt
->dst
;
330 memset(dst
+ 1, 0, sizeof(*rt
) - sizeof(*dst
));
331 INIT_LIST_HEAD(&rt
->rt6i_siblings
);
332 INIT_LIST_HEAD(&rt
->rt6i_uncached
);
337 static struct rt6_info
*ip6_dst_alloc(struct net
*net
,
338 struct net_device
*dev
,
340 struct fib6_table
*table
)
342 struct rt6_info
*rt
= __ip6_dst_alloc(net
, dev
, flags
, table
);
345 rt
->rt6i_pcpu
= alloc_percpu_gfp(struct rt6_info
*, GFP_ATOMIC
);
349 for_each_possible_cpu(cpu
) {
352 p
= per_cpu_ptr(rt
->rt6i_pcpu
, cpu
);
353 /* no one shares rt */
357 dst_destroy((struct dst_entry
*)rt
);
365 static void ip6_dst_destroy(struct dst_entry
*dst
)
367 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
368 struct dst_entry
*from
= dst
->from
;
369 struct inet6_dev
*idev
;
371 dst_destroy_metrics_generic(dst
);
374 free_percpu(rt
->rt6i_pcpu
);
376 rt6_uncached_list_del(rt
);
378 idev
= rt
->rt6i_idev
;
380 rt
->rt6i_idev
= NULL
;
388 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
391 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
392 struct inet6_dev
*idev
= rt
->rt6i_idev
;
393 struct net_device
*loopback_dev
=
394 dev_net(dev
)->loopback_dev
;
396 if (dev
!= loopback_dev
) {
397 if (idev
&& idev
->dev
== dev
) {
398 struct inet6_dev
*loopback_idev
=
399 in6_dev_get(loopback_dev
);
401 rt
->rt6i_idev
= loopback_idev
;
408 static bool rt6_check_expired(const struct rt6_info
*rt
)
410 if (rt
->rt6i_flags
& RTF_EXPIRES
) {
411 if (time_after(jiffies
, rt
->dst
.expires
))
413 } else if (rt
->dst
.from
) {
414 return rt6_check_expired((struct rt6_info
*) rt
->dst
.from
);
419 /* Multipath route selection:
420 * Hash based function using packet header and flowlabel.
421 * Adapted from fib_info_hashfn()
423 static int rt6_info_hash_nhsfn(unsigned int candidate_count
,
424 const struct flowi6
*fl6
)
426 unsigned int val
= fl6
->flowi6_proto
;
428 val
^= ipv6_addr_hash(&fl6
->daddr
);
429 val
^= ipv6_addr_hash(&fl6
->saddr
);
431 /* Work only if this not encapsulated */
432 switch (fl6
->flowi6_proto
) {
436 val
^= (__force u16
)fl6
->fl6_sport
;
437 val
^= (__force u16
)fl6
->fl6_dport
;
441 val
^= (__force u16
)fl6
->fl6_icmp_type
;
442 val
^= (__force u16
)fl6
->fl6_icmp_code
;
445 /* RFC6438 recommands to use flowlabel */
446 val
^= (__force u32
)fl6
->flowlabel
;
448 /* Perhaps, we need to tune, this function? */
449 val
= val
^ (val
>> 7) ^ (val
>> 12);
450 return val
% candidate_count
;
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 route_choosen
= rt6_info_hash_nhsfn(match
->rt6i_nsiblings
+ 1, fl6
);
461 /* Don't change the route, if route_choosen == 0
462 * (siblings does not include ourself)
465 list_for_each_entry_safe(sibling
, next_sibling
,
466 &match
->rt6i_siblings
, rt6i_siblings
) {
468 if (route_choosen
== 0) {
469 if (rt6_score_route(sibling
, oif
, strict
) < 0)
479 * Route lookup. Any table->tb6_lock is implied.
482 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
484 const struct in6_addr
*saddr
,
488 struct rt6_info
*local
= NULL
;
489 struct rt6_info
*sprt
;
491 if (!oif
&& ipv6_addr_any(saddr
))
494 for (sprt
= rt
; sprt
; sprt
= sprt
->dst
.rt6_next
) {
495 struct net_device
*dev
= sprt
->dst
.dev
;
498 if (dev
->ifindex
== oif
)
500 if (dev
->flags
& IFF_LOOPBACK
) {
501 if (!sprt
->rt6i_idev
||
502 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
503 if (flags
& RT6_LOOKUP_F_IFACE
&& oif
)
505 if (local
&& (!oif
||
506 local
->rt6i_idev
->dev
->ifindex
== oif
))
512 if (ipv6_chk_addr(net
, saddr
, dev
,
513 flags
& RT6_LOOKUP_F_IFACE
))
522 if (flags
& RT6_LOOKUP_F_IFACE
)
523 return net
->ipv6
.ip6_null_entry
;
529 #ifdef CONFIG_IPV6_ROUTER_PREF
530 struct __rt6_probe_work
{
531 struct work_struct work
;
532 struct in6_addr target
;
533 struct net_device
*dev
;
536 static void rt6_probe_deferred(struct work_struct
*w
)
538 struct in6_addr mcaddr
;
539 struct __rt6_probe_work
*work
=
540 container_of(w
, struct __rt6_probe_work
, work
);
542 addrconf_addr_solict_mult(&work
->target
, &mcaddr
);
543 ndisc_send_ns(work
->dev
, NULL
, &work
->target
, &mcaddr
, NULL
);
548 static void rt6_probe(struct rt6_info
*rt
)
550 struct neighbour
*neigh
;
552 * Okay, this does not seem to be appropriate
553 * for now, however, we need to check if it
554 * is really so; aka Router Reachability Probing.
556 * Router Reachability Probe MUST be rate-limited
557 * to no more than one per minute.
559 if (!rt
|| !(rt
->rt6i_flags
& RTF_GATEWAY
))
562 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
564 write_lock(&neigh
->lock
);
565 if (neigh
->nud_state
& NUD_VALID
)
570 time_after(jiffies
, neigh
->updated
+ rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
571 struct __rt6_probe_work
*work
;
573 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
576 __neigh_set_probe_once(neigh
);
579 write_unlock(&neigh
->lock
);
582 INIT_WORK(&work
->work
, rt6_probe_deferred
);
583 work
->target
= rt
->rt6i_gateway
;
584 dev_hold(rt
->dst
.dev
);
585 work
->dev
= rt
->dst
.dev
;
586 schedule_work(&work
->work
);
590 write_unlock(&neigh
->lock
);
592 rcu_read_unlock_bh();
595 static inline void rt6_probe(struct rt6_info
*rt
)
601 * Default Router Selection (RFC 2461 6.3.6)
603 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
605 struct net_device
*dev
= rt
->dst
.dev
;
606 if (!oif
|| dev
->ifindex
== oif
)
608 if ((dev
->flags
& IFF_LOOPBACK
) &&
609 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
614 static inline enum rt6_nud_state
rt6_check_neigh(struct rt6_info
*rt
)
616 struct neighbour
*neigh
;
617 enum rt6_nud_state ret
= RT6_NUD_FAIL_HARD
;
619 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
620 !(rt
->rt6i_flags
& RTF_GATEWAY
))
621 return RT6_NUD_SUCCEED
;
624 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
626 read_lock(&neigh
->lock
);
627 if (neigh
->nud_state
& NUD_VALID
)
628 ret
= RT6_NUD_SUCCEED
;
629 #ifdef CONFIG_IPV6_ROUTER_PREF
630 else if (!(neigh
->nud_state
& NUD_FAILED
))
631 ret
= RT6_NUD_SUCCEED
;
633 ret
= RT6_NUD_FAIL_PROBE
;
635 read_unlock(&neigh
->lock
);
637 ret
= IS_ENABLED(CONFIG_IPV6_ROUTER_PREF
) ?
638 RT6_NUD_SUCCEED
: RT6_NUD_FAIL_DO_RR
;
640 rcu_read_unlock_bh();
645 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
650 m
= rt6_check_dev(rt
, oif
);
651 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
652 return RT6_NUD_FAIL_HARD
;
653 #ifdef CONFIG_IPV6_ROUTER_PREF
654 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
656 if (strict
& RT6_LOOKUP_F_REACHABLE
) {
657 int n
= rt6_check_neigh(rt
);
664 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
665 int *mpri
, struct rt6_info
*match
,
669 bool match_do_rr
= false;
671 if (rt6_check_expired(rt
))
674 m
= rt6_score_route(rt
, oif
, strict
);
675 if (m
== RT6_NUD_FAIL_DO_RR
) {
677 m
= 0; /* lowest valid score */
678 } else if (m
== RT6_NUD_FAIL_HARD
) {
682 if (strict
& RT6_LOOKUP_F_REACHABLE
)
685 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
687 *do_rr
= match_do_rr
;
695 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
696 struct rt6_info
*rr_head
,
697 u32 metric
, int oif
, int strict
,
700 struct rt6_info
*rt
, *match
, *cont
;
705 for (rt
= rr_head
; rt
; rt
= rt
->dst
.rt6_next
) {
706 if (rt
->rt6i_metric
!= metric
) {
711 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
714 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
; rt
= rt
->dst
.rt6_next
) {
715 if (rt
->rt6i_metric
!= metric
) {
720 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
726 for (rt
= cont
; rt
; rt
= rt
->dst
.rt6_next
)
727 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
732 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
734 struct rt6_info
*match
, *rt0
;
740 fn
->rr_ptr
= rt0
= fn
->leaf
;
742 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
,
746 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
748 /* no entries matched; do round-robin */
749 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
756 net
= dev_net(rt0
->dst
.dev
);
757 return match
? match
: net
->ipv6
.ip6_null_entry
;
760 static bool rt6_is_gw_or_nonexthop(const struct rt6_info
*rt
)
762 return (rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
));
765 #ifdef CONFIG_IPV6_ROUTE_INFO
766 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
767 const struct in6_addr
*gwaddr
)
769 struct net
*net
= dev_net(dev
);
770 struct route_info
*rinfo
= (struct route_info
*) opt
;
771 struct in6_addr prefix_buf
, *prefix
;
773 unsigned long lifetime
;
776 if (len
< sizeof(struct route_info
)) {
780 /* Sanity check for prefix_len and length */
781 if (rinfo
->length
> 3) {
783 } else if (rinfo
->prefix_len
> 128) {
785 } else if (rinfo
->prefix_len
> 64) {
786 if (rinfo
->length
< 2) {
789 } else if (rinfo
->prefix_len
> 0) {
790 if (rinfo
->length
< 1) {
795 pref
= rinfo
->route_pref
;
796 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
799 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
801 if (rinfo
->length
== 3)
802 prefix
= (struct in6_addr
*)rinfo
->prefix
;
804 /* this function is safe */
805 ipv6_addr_prefix(&prefix_buf
,
806 (struct in6_addr
*)rinfo
->prefix
,
808 prefix
= &prefix_buf
;
811 if (rinfo
->prefix_len
== 0)
812 rt
= rt6_get_dflt_router(gwaddr
, dev
);
814 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
,
815 gwaddr
, dev
->ifindex
);
817 if (rt
&& !lifetime
) {
823 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
826 rt
->rt6i_flags
= RTF_ROUTEINFO
|
827 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
830 if (!addrconf_finite_timeout(lifetime
))
831 rt6_clean_expires(rt
);
833 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
841 static struct fib6_node
* fib6_backtrack(struct fib6_node
*fn
,
842 struct in6_addr
*saddr
)
844 struct fib6_node
*pn
;
846 if (fn
->fn_flags
& RTN_TL_ROOT
)
849 if (FIB6_SUBTREE(pn
) && FIB6_SUBTREE(pn
) != fn
)
850 fn
= fib6_lookup(FIB6_SUBTREE(pn
), NULL
, saddr
);
853 if (fn
->fn_flags
& RTN_RTINFO
)
858 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
859 struct fib6_table
*table
,
860 struct flowi6
*fl6
, int flags
)
862 struct fib6_node
*fn
;
865 read_lock_bh(&table
->tb6_lock
);
866 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
869 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
, fl6
->flowi6_oif
, flags
);
870 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
871 rt
= rt6_multipath_select(rt
, fl6
, fl6
->flowi6_oif
, flags
);
872 if (rt
== net
->ipv6
.ip6_null_entry
) {
873 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
877 dst_use(&rt
->dst
, jiffies
);
878 read_unlock_bh(&table
->tb6_lock
);
883 struct dst_entry
*ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
886 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
888 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
890 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
891 const struct in6_addr
*saddr
, int oif
, int strict
)
893 struct flowi6 fl6
= {
897 struct dst_entry
*dst
;
898 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
901 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
902 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
905 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
907 return (struct rt6_info
*) dst
;
913 EXPORT_SYMBOL(rt6_lookup
);
915 /* ip6_ins_rt is called with FREE table->tb6_lock.
916 It takes new route entry, the addition fails by any reason the
917 route is freed. In any case, if caller does not hold it, it may
921 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
,
922 struct mx6_config
*mxc
)
925 struct fib6_table
*table
;
927 table
= rt
->rt6i_table
;
928 write_lock_bh(&table
->tb6_lock
);
929 err
= fib6_add(&table
->tb6_root
, rt
, info
, mxc
);
930 write_unlock_bh(&table
->tb6_lock
);
935 int ip6_ins_rt(struct rt6_info
*rt
)
937 struct nl_info info
= { .nl_net
= dev_net(rt
->dst
.dev
), };
938 struct mx6_config mxc
= { .mx
= NULL
, };
940 return __ip6_ins_rt(rt
, &info
, &mxc
);
943 static struct rt6_info
*ip6_rt_cache_alloc(struct rt6_info
*ort
,
944 const struct in6_addr
*daddr
,
945 const struct in6_addr
*saddr
)
953 if (ort
->rt6i_flags
& (RTF_CACHE
| RTF_PCPU
))
954 ort
= (struct rt6_info
*)ort
->dst
.from
;
956 rt
= __ip6_dst_alloc(dev_net(ort
->dst
.dev
), ort
->dst
.dev
,
962 ip6_rt_copy_init(rt
, ort
);
963 rt
->rt6i_flags
|= RTF_CACHE
;
965 rt
->dst
.flags
|= DST_HOST
;
966 rt
->rt6i_dst
.addr
= *daddr
;
967 rt
->rt6i_dst
.plen
= 128;
969 if (!rt6_is_gw_or_nonexthop(ort
)) {
970 if (ort
->rt6i_dst
.plen
!= 128 &&
971 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
972 rt
->rt6i_flags
|= RTF_ANYCAST
;
973 #ifdef CONFIG_IPV6_SUBTREES
974 if (rt
->rt6i_src
.plen
&& saddr
) {
975 rt
->rt6i_src
.addr
= *saddr
;
976 rt
->rt6i_src
.plen
= 128;
984 static struct rt6_info
*ip6_rt_pcpu_alloc(struct rt6_info
*rt
)
986 struct rt6_info
*pcpu_rt
;
988 pcpu_rt
= __ip6_dst_alloc(dev_net(rt
->dst
.dev
),
989 rt
->dst
.dev
, rt
->dst
.flags
,
994 ip6_rt_copy_init(pcpu_rt
, rt
);
995 pcpu_rt
->rt6i_protocol
= rt
->rt6i_protocol
;
996 pcpu_rt
->rt6i_flags
|= RTF_PCPU
;
1000 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1001 static struct rt6_info
*rt6_get_pcpu_route(struct rt6_info
*rt
)
1003 struct rt6_info
*pcpu_rt
, *prev
, **p
;
1005 p
= this_cpu_ptr(rt
->rt6i_pcpu
);
1011 pcpu_rt
= ip6_rt_pcpu_alloc(rt
);
1013 struct net
*net
= dev_net(rt
->dst
.dev
);
1015 pcpu_rt
= net
->ipv6
.ip6_null_entry
;
1019 prev
= cmpxchg(p
, NULL
, pcpu_rt
);
1021 /* If someone did it before us, return prev instead */
1022 dst_destroy(&pcpu_rt
->dst
);
1027 dst_hold(&pcpu_rt
->dst
);
1028 rt6_dst_from_metrics_check(pcpu_rt
);
1032 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
1033 struct flowi6
*fl6
, int flags
)
1035 struct fib6_node
*fn
, *saved_fn
;
1036 struct rt6_info
*rt
;
1039 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
1040 if (net
->ipv6
.devconf_all
->forwarding
== 0)
1041 strict
|= RT6_LOOKUP_F_REACHABLE
;
1043 read_lock_bh(&table
->tb6_lock
);
1045 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1049 rt
= rt6_select(fn
, oif
, strict
);
1050 if (rt
->rt6i_nsiblings
)
1051 rt
= rt6_multipath_select(rt
, fl6
, oif
, strict
);
1052 if (rt
== net
->ipv6
.ip6_null_entry
) {
1053 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
1055 goto redo_rt6_select
;
1056 else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
1057 /* also consider unreachable route */
1058 strict
&= ~RT6_LOOKUP_F_REACHABLE
;
1060 goto redo_rt6_select
;
1065 if (rt
== net
->ipv6
.ip6_null_entry
|| (rt
->rt6i_flags
& RTF_CACHE
)) {
1066 dst_use(&rt
->dst
, jiffies
);
1067 read_unlock_bh(&table
->tb6_lock
);
1069 rt6_dst_from_metrics_check(rt
);
1071 } else if (unlikely((fl6
->flowi6_flags
& FLOWI_FLAG_KNOWN_NH
) &&
1072 !(rt
->rt6i_flags
& RTF_GATEWAY
))) {
1073 /* Create a RTF_CACHE clone which will not be
1074 * owned by the fib6 tree. It is for the special case where
1075 * the daddr in the skb during the neighbor look-up is different
1076 * from the fl6->daddr used to look-up route here.
1079 struct rt6_info
*uncached_rt
;
1081 dst_use(&rt
->dst
, jiffies
);
1082 read_unlock_bh(&table
->tb6_lock
);
1084 uncached_rt
= ip6_rt_cache_alloc(rt
, &fl6
->daddr
, NULL
);
1085 dst_release(&rt
->dst
);
1088 rt6_uncached_list_add(uncached_rt
);
1090 uncached_rt
= net
->ipv6
.ip6_null_entry
;
1092 dst_hold(&uncached_rt
->dst
);
1096 /* Get a percpu copy */
1098 struct rt6_info
*pcpu_rt
;
1100 rt
->dst
.lastuse
= jiffies
;
1102 pcpu_rt
= rt6_get_pcpu_route(rt
);
1103 read_unlock_bh(&table
->tb6_lock
);
1109 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
1110 struct flowi6
*fl6
, int flags
)
1112 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
1115 static struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
1116 struct net_device
*dev
,
1117 struct flowi6
*fl6
, int flags
)
1119 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
1120 flags
|= RT6_LOOKUP_F_IFACE
;
1122 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
1125 void ip6_route_input(struct sk_buff
*skb
)
1127 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1128 struct net
*net
= dev_net(skb
->dev
);
1129 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1130 struct flowi6 fl6
= {
1131 .flowi6_iif
= skb
->dev
->ifindex
,
1132 .daddr
= iph
->daddr
,
1133 .saddr
= iph
->saddr
,
1134 .flowlabel
= ip6_flowinfo(iph
),
1135 .flowi6_mark
= skb
->mark
,
1136 .flowi6_proto
= iph
->nexthdr
,
1139 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
1142 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
1143 struct flowi6
*fl6
, int flags
)
1145 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1148 struct dst_entry
*ip6_route_output(struct net
*net
, const struct sock
*sk
,
1153 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1155 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
))
1156 flags
|= RT6_LOOKUP_F_IFACE
;
1158 if (!ipv6_addr_any(&fl6
->saddr
))
1159 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1161 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1163 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1165 EXPORT_SYMBOL(ip6_route_output
);
1167 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1169 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1170 struct dst_entry
*new = NULL
;
1172 rt
= dst_alloc(&ip6_dst_blackhole_ops
, ort
->dst
.dev
, 1, DST_OBSOLETE_NONE
, 0);
1176 memset(new + 1, 0, sizeof(*rt
) - sizeof(*new));
1179 new->input
= dst_discard
;
1180 new->output
= dst_discard_sk
;
1182 if (dst_metrics_read_only(&ort
->dst
))
1183 new->_metrics
= ort
->dst
._metrics
;
1185 dst_copy_metrics(new, &ort
->dst
);
1186 rt
->rt6i_idev
= ort
->rt6i_idev
;
1188 in6_dev_hold(rt
->rt6i_idev
);
1190 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1191 rt
->rt6i_flags
= ort
->rt6i_flags
;
1192 rt
->rt6i_metric
= 0;
1194 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1195 #ifdef CONFIG_IPV6_SUBTREES
1196 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1202 dst_release(dst_orig
);
1203 return new ? new : ERR_PTR(-ENOMEM
);
1207 * Destination cache support functions
1210 static void rt6_dst_from_metrics_check(struct rt6_info
*rt
)
1213 dst_metrics_ptr(&rt
->dst
) != dst_metrics_ptr(rt
->dst
.from
))
1214 dst_init_metrics(&rt
->dst
, dst_metrics_ptr(rt
->dst
.from
), true);
1217 static struct dst_entry
*rt6_check(struct rt6_info
*rt
, u32 cookie
)
1219 if (!rt
->rt6i_node
|| (rt
->rt6i_node
->fn_sernum
!= cookie
))
1222 if (rt6_check_expired(rt
))
1228 static struct dst_entry
*rt6_dst_from_check(struct rt6_info
*rt
, u32 cookie
)
1230 if (rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1231 rt6_check((struct rt6_info
*)(rt
->dst
.from
), cookie
))
1237 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1239 struct rt6_info
*rt
;
1241 rt
= (struct rt6_info
*) dst
;
1243 /* All IPV6 dsts are created with ->obsolete set to the value
1244 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1245 * into this function always.
1248 rt6_dst_from_metrics_check(rt
);
1250 if ((rt
->rt6i_flags
& RTF_PCPU
) || unlikely(dst
->flags
& DST_NOCACHE
))
1251 return rt6_dst_from_check(rt
, cookie
);
1253 return rt6_check(rt
, cookie
);
1256 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1258 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1261 if (rt
->rt6i_flags
& RTF_CACHE
) {
1262 if (rt6_check_expired(rt
)) {
1274 static void ip6_link_failure(struct sk_buff
*skb
)
1276 struct rt6_info
*rt
;
1278 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1280 rt
= (struct rt6_info
*) skb_dst(skb
);
1282 if (rt
->rt6i_flags
& RTF_CACHE
) {
1286 } else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
)) {
1287 rt
->rt6i_node
->fn_sernum
= -1;
1292 static void rt6_do_update_pmtu(struct rt6_info
*rt
, u32 mtu
)
1294 struct net
*net
= dev_net(rt
->dst
.dev
);
1296 rt
->rt6i_flags
|= RTF_MODIFIED
;
1297 rt
->rt6i_pmtu
= mtu
;
1298 rt6_update_expires(rt
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1301 static void __ip6_rt_update_pmtu(struct dst_entry
*dst
, const struct sock
*sk
,
1302 const struct ipv6hdr
*iph
, u32 mtu
)
1304 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1306 if (rt6
->rt6i_flags
& RTF_LOCAL
)
1310 mtu
= max_t(u32
, mtu
, IPV6_MIN_MTU
);
1311 if (mtu
>= dst_mtu(dst
))
1314 if (rt6
->rt6i_flags
& RTF_CACHE
) {
1315 rt6_do_update_pmtu(rt6
, mtu
);
1317 const struct in6_addr
*daddr
, *saddr
;
1318 struct rt6_info
*nrt6
;
1321 daddr
= &iph
->daddr
;
1322 saddr
= &iph
->saddr
;
1324 daddr
= &sk
->sk_v6_daddr
;
1325 saddr
= &inet6_sk(sk
)->saddr
;
1329 nrt6
= ip6_rt_cache_alloc(rt6
, daddr
, saddr
);
1331 rt6_do_update_pmtu(nrt6
, mtu
);
1333 /* ip6_ins_rt(nrt6) will bump the
1334 * rt6->rt6i_node->fn_sernum
1335 * which will fail the next rt6_check() and
1336 * invalidate the sk->sk_dst_cache.
1343 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1344 struct sk_buff
*skb
, u32 mtu
)
1346 __ip6_rt_update_pmtu(dst
, sk
, skb
? ipv6_hdr(skb
) : NULL
, mtu
);
1349 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1352 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1353 struct dst_entry
*dst
;
1356 memset(&fl6
, 0, sizeof(fl6
));
1357 fl6
.flowi6_oif
= oif
;
1358 fl6
.flowi6_mark
= mark
? mark
: IP6_REPLY_MARK(net
, skb
->mark
);
1359 fl6
.daddr
= iph
->daddr
;
1360 fl6
.saddr
= iph
->saddr
;
1361 fl6
.flowlabel
= ip6_flowinfo(iph
);
1363 dst
= ip6_route_output(net
, NULL
, &fl6
);
1365 __ip6_rt_update_pmtu(dst
, NULL
, iph
, ntohl(mtu
));
1368 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1370 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1372 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1373 sk
->sk_bound_dev_if
, sk
->sk_mark
);
1375 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1377 /* Handle redirects */
1378 struct ip6rd_flowi
{
1380 struct in6_addr gateway
;
1383 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
1384 struct fib6_table
*table
,
1388 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl6
;
1389 struct rt6_info
*rt
;
1390 struct fib6_node
*fn
;
1392 /* Get the "current" route for this destination and
1393 * check if the redirect has come from approriate router.
1395 * RFC 4861 specifies that redirects should only be
1396 * accepted if they come from the nexthop to the target.
1397 * Due to the way the routes are chosen, this notion
1398 * is a bit fuzzy and one might need to check all possible
1402 read_lock_bh(&table
->tb6_lock
);
1403 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1405 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1406 if (rt6_check_expired(rt
))
1410 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
1412 if (fl6
->flowi6_oif
!= rt
->dst
.dev
->ifindex
)
1414 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
))
1420 rt
= net
->ipv6
.ip6_null_entry
;
1421 else if (rt
->dst
.error
) {
1422 rt
= net
->ipv6
.ip6_null_entry
;
1426 if (rt
== net
->ipv6
.ip6_null_entry
) {
1427 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
1435 read_unlock_bh(&table
->tb6_lock
);
1440 static struct dst_entry
*ip6_route_redirect(struct net
*net
,
1441 const struct flowi6
*fl6
,
1442 const struct in6_addr
*gateway
)
1444 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1445 struct ip6rd_flowi rdfl
;
1448 rdfl
.gateway
= *gateway
;
1450 return fib6_rule_lookup(net
, &rdfl
.fl6
,
1451 flags
, __ip6_route_redirect
);
1454 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
)
1456 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1457 struct dst_entry
*dst
;
1460 memset(&fl6
, 0, sizeof(fl6
));
1461 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1462 fl6
.flowi6_oif
= oif
;
1463 fl6
.flowi6_mark
= mark
;
1464 fl6
.daddr
= iph
->daddr
;
1465 fl6
.saddr
= iph
->saddr
;
1466 fl6
.flowlabel
= ip6_flowinfo(iph
);
1468 dst
= ip6_route_redirect(net
, &fl6
, &ipv6_hdr(skb
)->saddr
);
1469 rt6_do_redirect(dst
, NULL
, skb
);
1472 EXPORT_SYMBOL_GPL(ip6_redirect
);
1474 void ip6_redirect_no_header(struct sk_buff
*skb
, struct net
*net
, int oif
,
1477 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1478 const struct rd_msg
*msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1479 struct dst_entry
*dst
;
1482 memset(&fl6
, 0, sizeof(fl6
));
1483 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1484 fl6
.flowi6_oif
= oif
;
1485 fl6
.flowi6_mark
= mark
;
1486 fl6
.daddr
= msg
->dest
;
1487 fl6
.saddr
= iph
->daddr
;
1489 dst
= ip6_route_redirect(net
, &fl6
, &iph
->saddr
);
1490 rt6_do_redirect(dst
, NULL
, skb
);
1494 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1496 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
);
1498 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1500 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1502 struct net_device
*dev
= dst
->dev
;
1503 unsigned int mtu
= dst_mtu(dst
);
1504 struct net
*net
= dev_net(dev
);
1506 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1508 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1509 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1512 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1513 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1514 * IPV6_MAXPLEN is also valid and means: "any MSS,
1515 * rely only on pmtu discovery"
1517 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1522 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1524 const struct rt6_info
*rt
= (const struct rt6_info
*)dst
;
1525 unsigned int mtu
= rt
->rt6i_pmtu
;
1526 struct inet6_dev
*idev
;
1531 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1538 idev
= __in6_dev_get(dst
->dev
);
1540 mtu
= idev
->cnf
.mtu6
;
1544 return min_t(unsigned int, mtu
, IP6_MAX_MTU
);
1547 static struct dst_entry
*icmp6_dst_gc_list
;
1548 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1550 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1553 struct dst_entry
*dst
;
1554 struct rt6_info
*rt
;
1555 struct inet6_dev
*idev
= in6_dev_get(dev
);
1556 struct net
*net
= dev_net(dev
);
1558 if (unlikely(!idev
))
1559 return ERR_PTR(-ENODEV
);
1561 rt
= ip6_dst_alloc(net
, dev
, 0, NULL
);
1562 if (unlikely(!rt
)) {
1564 dst
= ERR_PTR(-ENOMEM
);
1568 rt
->dst
.flags
|= DST_HOST
;
1569 rt
->dst
.output
= ip6_output
;
1570 atomic_set(&rt
->dst
.__refcnt
, 1);
1571 rt
->rt6i_gateway
= fl6
->daddr
;
1572 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1573 rt
->rt6i_dst
.plen
= 128;
1574 rt
->rt6i_idev
= idev
;
1575 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1577 spin_lock_bh(&icmp6_dst_lock
);
1578 rt
->dst
.next
= icmp6_dst_gc_list
;
1579 icmp6_dst_gc_list
= &rt
->dst
;
1580 spin_unlock_bh(&icmp6_dst_lock
);
1582 fib6_force_start_gc(net
);
1584 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1590 int icmp6_dst_gc(void)
1592 struct dst_entry
*dst
, **pprev
;
1595 spin_lock_bh(&icmp6_dst_lock
);
1596 pprev
= &icmp6_dst_gc_list
;
1598 while ((dst
= *pprev
) != NULL
) {
1599 if (!atomic_read(&dst
->__refcnt
)) {
1608 spin_unlock_bh(&icmp6_dst_lock
);
1613 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1616 struct dst_entry
*dst
, **pprev
;
1618 spin_lock_bh(&icmp6_dst_lock
);
1619 pprev
= &icmp6_dst_gc_list
;
1620 while ((dst
= *pprev
) != NULL
) {
1621 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1622 if (func(rt
, arg
)) {
1629 spin_unlock_bh(&icmp6_dst_lock
);
1632 static int ip6_dst_gc(struct dst_ops
*ops
)
1634 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1635 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1636 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1637 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1638 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1639 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1642 entries
= dst_entries_get_fast(ops
);
1643 if (time_after(rt_last_gc
+ rt_min_interval
, jiffies
) &&
1644 entries
<= rt_max_size
)
1647 net
->ipv6
.ip6_rt_gc_expire
++;
1648 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
, true);
1649 entries
= dst_entries_get_slow(ops
);
1650 if (entries
< ops
->gc_thresh
)
1651 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1653 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1654 return entries
> rt_max_size
;
1657 static int ip6_convert_metrics(struct mx6_config
*mxc
,
1658 const struct fib6_config
*cfg
)
1667 mp
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_KERNEL
);
1671 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1672 int type
= nla_type(nla
);
1677 if (unlikely(type
> RTAX_MAX
))
1679 if (type
== RTAX_CC_ALGO
) {
1680 char tmp
[TCP_CA_NAME_MAX
];
1682 nla_strlcpy(tmp
, nla
, sizeof(tmp
));
1683 val
= tcp_ca_get_key_by_name(tmp
);
1684 if (val
== TCP_CA_UNSPEC
)
1687 val
= nla_get_u32(nla
);
1691 __set_bit(type
- 1, mxc
->mx_valid
);
1703 int ip6_route_add(struct fib6_config
*cfg
)
1706 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1707 struct rt6_info
*rt
= NULL
;
1708 struct net_device
*dev
= NULL
;
1709 struct inet6_dev
*idev
= NULL
;
1710 struct fib6_table
*table
;
1711 struct mx6_config mxc
= { .mx
= NULL
, };
1714 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1716 #ifndef CONFIG_IPV6_SUBTREES
1717 if (cfg
->fc_src_len
)
1720 if (cfg
->fc_ifindex
) {
1722 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1725 idev
= in6_dev_get(dev
);
1730 if (cfg
->fc_metric
== 0)
1731 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1734 if (cfg
->fc_nlinfo
.nlh
&&
1735 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1736 table
= fib6_get_table(net
, cfg
->fc_table
);
1738 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1739 table
= fib6_new_table(net
, cfg
->fc_table
);
1742 table
= fib6_new_table(net
, cfg
->fc_table
);
1748 rt
= ip6_dst_alloc(net
, NULL
, (cfg
->fc_flags
& RTF_ADDRCONF
) ? 0 : DST_NOCOUNT
, table
);
1755 if (cfg
->fc_flags
& RTF_EXPIRES
)
1756 rt6_set_expires(rt
, jiffies
+
1757 clock_t_to_jiffies(cfg
->fc_expires
));
1759 rt6_clean_expires(rt
);
1761 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1762 cfg
->fc_protocol
= RTPROT_BOOT
;
1763 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1765 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1767 if (addr_type
& IPV6_ADDR_MULTICAST
)
1768 rt
->dst
.input
= ip6_mc_input
;
1769 else if (cfg
->fc_flags
& RTF_LOCAL
)
1770 rt
->dst
.input
= ip6_input
;
1772 rt
->dst
.input
= ip6_forward
;
1774 rt
->dst
.output
= ip6_output
;
1776 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1777 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1778 if (rt
->rt6i_dst
.plen
== 128)
1779 rt
->dst
.flags
|= DST_HOST
;
1781 #ifdef CONFIG_IPV6_SUBTREES
1782 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1783 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1786 rt
->rt6i_metric
= cfg
->fc_metric
;
1788 /* We cannot add true routes via loopback here,
1789 they would result in kernel looping; promote them to reject routes
1791 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1792 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1793 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1794 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1795 /* hold loopback dev/idev if we haven't done so. */
1796 if (dev
!= net
->loopback_dev
) {
1801 dev
= net
->loopback_dev
;
1803 idev
= in6_dev_get(dev
);
1809 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1810 switch (cfg
->fc_type
) {
1812 rt
->dst
.error
= -EINVAL
;
1813 rt
->dst
.output
= dst_discard_sk
;
1814 rt
->dst
.input
= dst_discard
;
1817 rt
->dst
.error
= -EACCES
;
1818 rt
->dst
.output
= ip6_pkt_prohibit_out
;
1819 rt
->dst
.input
= ip6_pkt_prohibit
;
1823 rt
->dst
.error
= (cfg
->fc_type
== RTN_THROW
) ? -EAGAIN
1825 rt
->dst
.output
= ip6_pkt_discard_out
;
1826 rt
->dst
.input
= ip6_pkt_discard
;
1832 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1833 const struct in6_addr
*gw_addr
;
1836 gw_addr
= &cfg
->fc_gateway
;
1838 /* if gw_addr is local we will fail to detect this in case
1839 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1840 * will return already-added prefix route via interface that
1841 * prefix route was assigned to, which might be non-loopback.
1844 if (ipv6_chk_addr_and_flags(net
, gw_addr
, NULL
, 0, 0))
1847 rt
->rt6i_gateway
= *gw_addr
;
1848 gwa_type
= ipv6_addr_type(gw_addr
);
1850 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1851 struct rt6_info
*grt
;
1853 /* IPv6 strictly inhibits using not link-local
1854 addresses as nexthop address.
1855 Otherwise, router will not able to send redirects.
1856 It is very good, but in some (rare!) circumstances
1857 (SIT, PtP, NBMA NOARP links) it is handy to allow
1858 some exceptions. --ANK
1860 if (!(gwa_type
& IPV6_ADDR_UNICAST
))
1863 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1865 err
= -EHOSTUNREACH
;
1869 if (dev
!= grt
->dst
.dev
) {
1875 idev
= grt
->rt6i_idev
;
1877 in6_dev_hold(grt
->rt6i_idev
);
1879 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
1887 if (!dev
|| (dev
->flags
& IFF_LOOPBACK
))
1895 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
1896 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
1900 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
1901 rt
->rt6i_prefsrc
.plen
= 128;
1903 rt
->rt6i_prefsrc
.plen
= 0;
1905 rt
->rt6i_flags
= cfg
->fc_flags
;
1909 rt
->rt6i_idev
= idev
;
1910 rt
->rt6i_table
= table
;
1912 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1914 err
= ip6_convert_metrics(&mxc
, cfg
);
1918 err
= __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
, &mxc
);
1932 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1935 struct fib6_table
*table
;
1936 struct net
*net
= dev_net(rt
->dst
.dev
);
1938 if (rt
== net
->ipv6
.ip6_null_entry
) {
1943 table
= rt
->rt6i_table
;
1944 write_lock_bh(&table
->tb6_lock
);
1945 err
= fib6_del(rt
, info
);
1946 write_unlock_bh(&table
->tb6_lock
);
1953 int ip6_del_rt(struct rt6_info
*rt
)
1955 struct nl_info info
= {
1956 .nl_net
= dev_net(rt
->dst
.dev
),
1958 return __ip6_del_rt(rt
, &info
);
1961 static int ip6_route_del(struct fib6_config
*cfg
)
1963 struct fib6_table
*table
;
1964 struct fib6_node
*fn
;
1965 struct rt6_info
*rt
;
1968 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1972 read_lock_bh(&table
->tb6_lock
);
1974 fn
= fib6_locate(&table
->tb6_root
,
1975 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1976 &cfg
->fc_src
, cfg
->fc_src_len
);
1979 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1980 if ((rt
->rt6i_flags
& RTF_CACHE
) &&
1981 !(cfg
->fc_flags
& RTF_CACHE
))
1983 if (cfg
->fc_ifindex
&&
1985 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
1987 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1988 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1990 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1993 read_unlock_bh(&table
->tb6_lock
);
1995 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1998 read_unlock_bh(&table
->tb6_lock
);
2003 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
2005 struct net
*net
= dev_net(skb
->dev
);
2006 struct netevent_redirect netevent
;
2007 struct rt6_info
*rt
, *nrt
= NULL
;
2008 struct ndisc_options ndopts
;
2009 struct inet6_dev
*in6_dev
;
2010 struct neighbour
*neigh
;
2012 int optlen
, on_link
;
2015 optlen
= skb_tail_pointer(skb
) - skb_transport_header(skb
);
2016 optlen
-= sizeof(*msg
);
2019 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2023 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
2025 if (ipv6_addr_is_multicast(&msg
->dest
)) {
2026 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2031 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
2033 } else if (ipv6_addr_type(&msg
->target
) !=
2034 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
2035 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2039 in6_dev
= __in6_dev_get(skb
->dev
);
2042 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
2046 * The IP source address of the Redirect MUST be the same as the current
2047 * first-hop router for the specified ICMP Destination Address.
2050 if (!ndisc_parse_options(msg
->opt
, optlen
, &ndopts
)) {
2051 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2056 if (ndopts
.nd_opts_tgt_lladdr
) {
2057 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
2060 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2065 rt
= (struct rt6_info
*) dst
;
2066 if (rt
== net
->ipv6
.ip6_null_entry
) {
2067 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2071 /* Redirect received -> path was valid.
2072 * Look, redirects are sent only in response to data packets,
2073 * so that this nexthop apparently is reachable. --ANK
2075 dst_confirm(&rt
->dst
);
2077 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
2082 * We have finally decided to accept it.
2085 neigh_update(neigh
, lladdr
, NUD_STALE
,
2086 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
2087 NEIGH_UPDATE_F_OVERRIDE
|
2088 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
2089 NEIGH_UPDATE_F_ISROUTER
))
2092 nrt
= ip6_rt_cache_alloc(rt
, &msg
->dest
, NULL
);
2096 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
2098 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
2100 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
2102 if (ip6_ins_rt(nrt
))
2105 netevent
.old
= &rt
->dst
;
2106 netevent
.new = &nrt
->dst
;
2107 netevent
.daddr
= &msg
->dest
;
2108 netevent
.neigh
= neigh
;
2109 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
2111 if (rt
->rt6i_flags
& RTF_CACHE
) {
2112 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
2117 neigh_release(neigh
);
2121 * Misc support functions
2124 static void rt6_set_from(struct rt6_info
*rt
, struct rt6_info
*from
)
2126 BUG_ON(from
->dst
.from
);
2128 rt
->rt6i_flags
&= ~RTF_EXPIRES
;
2129 dst_hold(&from
->dst
);
2130 rt
->dst
.from
= &from
->dst
;
2131 dst_init_metrics(&rt
->dst
, dst_metrics_ptr(&from
->dst
), true);
2134 static void ip6_rt_copy_init(struct rt6_info
*rt
, struct rt6_info
*ort
)
2136 rt
->dst
.input
= ort
->dst
.input
;
2137 rt
->dst
.output
= ort
->dst
.output
;
2138 rt
->rt6i_dst
= ort
->rt6i_dst
;
2139 rt
->dst
.error
= ort
->dst
.error
;
2140 rt
->rt6i_idev
= ort
->rt6i_idev
;
2142 in6_dev_hold(rt
->rt6i_idev
);
2143 rt
->dst
.lastuse
= jiffies
;
2144 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
2145 rt
->rt6i_flags
= ort
->rt6i_flags
;
2146 rt6_set_from(rt
, ort
);
2147 rt
->rt6i_metric
= ort
->rt6i_metric
;
2148 #ifdef CONFIG_IPV6_SUBTREES
2149 rt
->rt6i_src
= ort
->rt6i_src
;
2151 rt
->rt6i_prefsrc
= ort
->rt6i_prefsrc
;
2152 rt
->rt6i_table
= ort
->rt6i_table
;
2155 #ifdef CONFIG_IPV6_ROUTE_INFO
2156 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
2157 const struct in6_addr
*prefix
, int prefixlen
,
2158 const struct in6_addr
*gwaddr
, int ifindex
)
2160 struct fib6_node
*fn
;
2161 struct rt6_info
*rt
= NULL
;
2162 struct fib6_table
*table
;
2164 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
2168 read_lock_bh(&table
->tb6_lock
);
2169 fn
= fib6_locate(&table
->tb6_root
, prefix
, prefixlen
, NULL
, 0);
2173 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2174 if (rt
->dst
.dev
->ifindex
!= ifindex
)
2176 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
2178 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
2184 read_unlock_bh(&table
->tb6_lock
);
2188 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
2189 const struct in6_addr
*prefix
, int prefixlen
,
2190 const struct in6_addr
*gwaddr
, int ifindex
,
2193 struct fib6_config cfg
= {
2194 .fc_table
= RT6_TABLE_INFO
,
2195 .fc_metric
= IP6_RT_PRIO_USER
,
2196 .fc_ifindex
= ifindex
,
2197 .fc_dst_len
= prefixlen
,
2198 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
2199 RTF_UP
| RTF_PREF(pref
),
2200 .fc_nlinfo
.portid
= 0,
2201 .fc_nlinfo
.nlh
= NULL
,
2202 .fc_nlinfo
.nl_net
= net
,
2205 cfg
.fc_dst
= *prefix
;
2206 cfg
.fc_gateway
= *gwaddr
;
2208 /* We should treat it as a default route if prefix length is 0. */
2210 cfg
.fc_flags
|= RTF_DEFAULT
;
2212 ip6_route_add(&cfg
);
2214 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
2218 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
2220 struct rt6_info
*rt
;
2221 struct fib6_table
*table
;
2223 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
2227 read_lock_bh(&table
->tb6_lock
);
2228 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2229 if (dev
== rt
->dst
.dev
&&
2230 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
2231 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
2236 read_unlock_bh(&table
->tb6_lock
);
2240 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
2241 struct net_device
*dev
,
2244 struct fib6_config cfg
= {
2245 .fc_table
= RT6_TABLE_DFLT
,
2246 .fc_metric
= IP6_RT_PRIO_USER
,
2247 .fc_ifindex
= dev
->ifindex
,
2248 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
2249 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
2250 .fc_nlinfo
.portid
= 0,
2251 .fc_nlinfo
.nlh
= NULL
,
2252 .fc_nlinfo
.nl_net
= dev_net(dev
),
2255 cfg
.fc_gateway
= *gwaddr
;
2257 ip6_route_add(&cfg
);
2259 return rt6_get_dflt_router(gwaddr
, dev
);
2262 void rt6_purge_dflt_routers(struct net
*net
)
2264 struct rt6_info
*rt
;
2265 struct fib6_table
*table
;
2267 /* NOTE: Keep consistent with rt6_get_dflt_router */
2268 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
2273 read_lock_bh(&table
->tb6_lock
);
2274 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2275 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
) &&
2276 (!rt
->rt6i_idev
|| rt
->rt6i_idev
->cnf
.accept_ra
!= 2)) {
2278 read_unlock_bh(&table
->tb6_lock
);
2283 read_unlock_bh(&table
->tb6_lock
);
2286 static void rtmsg_to_fib6_config(struct net
*net
,
2287 struct in6_rtmsg
*rtmsg
,
2288 struct fib6_config
*cfg
)
2290 memset(cfg
, 0, sizeof(*cfg
));
2292 cfg
->fc_table
= RT6_TABLE_MAIN
;
2293 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
2294 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
2295 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
2296 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
2297 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
2298 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
2300 cfg
->fc_nlinfo
.nl_net
= net
;
2302 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
2303 cfg
->fc_src
= rtmsg
->rtmsg_src
;
2304 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
2307 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
2309 struct fib6_config cfg
;
2310 struct in6_rtmsg rtmsg
;
2314 case SIOCADDRT
: /* Add a route */
2315 case SIOCDELRT
: /* Delete a route */
2316 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2318 err
= copy_from_user(&rtmsg
, arg
,
2319 sizeof(struct in6_rtmsg
));
2323 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
2328 err
= ip6_route_add(&cfg
);
2331 err
= ip6_route_del(&cfg
);
2345 * Drop the packet on the floor
2348 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2351 struct dst_entry
*dst
= skb_dst(skb
);
2352 switch (ipstats_mib_noroutes
) {
2353 case IPSTATS_MIB_INNOROUTES
:
2354 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2355 if (type
== IPV6_ADDR_ANY
) {
2356 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2357 IPSTATS_MIB_INADDRERRORS
);
2361 case IPSTATS_MIB_OUTNOROUTES
:
2362 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2363 ipstats_mib_noroutes
);
2366 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2371 static int ip6_pkt_discard(struct sk_buff
*skb
)
2373 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2376 static int ip6_pkt_discard_out(struct sock
*sk
, struct sk_buff
*skb
)
2378 skb
->dev
= skb_dst(skb
)->dev
;
2379 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2382 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2384 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2387 static int ip6_pkt_prohibit_out(struct sock
*sk
, struct sk_buff
*skb
)
2389 skb
->dev
= skb_dst(skb
)->dev
;
2390 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2394 * Allocate a dst for local (unicast / anycast) address.
2397 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2398 const struct in6_addr
*addr
,
2401 struct net
*net
= dev_net(idev
->dev
);
2402 struct rt6_info
*rt
= ip6_dst_alloc(net
, net
->loopback_dev
,
2405 return ERR_PTR(-ENOMEM
);
2409 rt
->dst
.flags
|= DST_HOST
;
2410 rt
->dst
.input
= ip6_input
;
2411 rt
->dst
.output
= ip6_output
;
2412 rt
->rt6i_idev
= idev
;
2414 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2416 rt
->rt6i_flags
|= RTF_ANYCAST
;
2418 rt
->rt6i_flags
|= RTF_LOCAL
;
2420 rt
->rt6i_gateway
= *addr
;
2421 rt
->rt6i_dst
.addr
= *addr
;
2422 rt
->rt6i_dst
.plen
= 128;
2423 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2425 atomic_set(&rt
->dst
.__refcnt
, 1);
2430 int ip6_route_get_saddr(struct net
*net
,
2431 struct rt6_info
*rt
,
2432 const struct in6_addr
*daddr
,
2434 struct in6_addr
*saddr
)
2436 struct inet6_dev
*idev
=
2437 rt
? ip6_dst_idev((struct dst_entry
*)rt
) : NULL
;
2439 if (rt
&& rt
->rt6i_prefsrc
.plen
)
2440 *saddr
= rt
->rt6i_prefsrc
.addr
;
2442 err
= ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2443 daddr
, prefs
, saddr
);
2447 /* remove deleted ip from prefsrc entries */
2448 struct arg_dev_net_ip
{
2449 struct net_device
*dev
;
2451 struct in6_addr
*addr
;
2454 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2456 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2457 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2458 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2460 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2461 rt
!= net
->ipv6
.ip6_null_entry
&&
2462 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2463 /* remove prefsrc entry */
2464 rt
->rt6i_prefsrc
.plen
= 0;
2469 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2471 struct net
*net
= dev_net(ifp
->idev
->dev
);
2472 struct arg_dev_net_ip adni
= {
2473 .dev
= ifp
->idev
->dev
,
2477 fib6_clean_all(net
, fib6_remove_prefsrc
, &adni
);
2480 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2481 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2483 /* Remove routers and update dst entries when gateway turn into host. */
2484 static int fib6_clean_tohost(struct rt6_info
*rt
, void *arg
)
2486 struct in6_addr
*gateway
= (struct in6_addr
*)arg
;
2488 if ((((rt
->rt6i_flags
& RTF_RA_ROUTER
) == RTF_RA_ROUTER
) ||
2489 ((rt
->rt6i_flags
& RTF_CACHE_GATEWAY
) == RTF_CACHE_GATEWAY
)) &&
2490 ipv6_addr_equal(gateway
, &rt
->rt6i_gateway
)) {
2496 void rt6_clean_tohost(struct net
*net
, struct in6_addr
*gateway
)
2498 fib6_clean_all(net
, fib6_clean_tohost
, gateway
);
2501 struct arg_dev_net
{
2502 struct net_device
*dev
;
2506 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2508 const struct arg_dev_net
*adn
= arg
;
2509 const struct net_device
*dev
= adn
->dev
;
2511 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2512 rt
!= adn
->net
->ipv6
.ip6_null_entry
)
2518 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2520 struct arg_dev_net adn
= {
2525 fib6_clean_all(net
, fib6_ifdown
, &adn
);
2526 icmp6_clean_all(fib6_ifdown
, &adn
);
2527 rt6_uncached_list_flush_dev(net
, dev
);
2530 struct rt6_mtu_change_arg
{
2531 struct net_device
*dev
;
2535 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2537 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2538 struct inet6_dev
*idev
;
2540 /* In IPv6 pmtu discovery is not optional,
2541 so that RTAX_MTU lock cannot disable it.
2542 We still use this lock to block changes
2543 caused by addrconf/ndisc.
2546 idev
= __in6_dev_get(arg
->dev
);
2550 /* For administrative MTU increase, there is no way to discover
2551 IPv6 PMTU increase, so PMTU increase should be updated here.
2552 Since RFC 1981 doesn't include administrative MTU increase
2553 update PMTU increase is a MUST. (i.e. jumbo frame)
2556 If new MTU is less than route PMTU, this new MTU will be the
2557 lowest MTU in the path, update the route PMTU to reflect PMTU
2558 decreases; if new MTU is greater than route PMTU, and the
2559 old MTU is the lowest MTU in the path, update the route PMTU
2560 to reflect the increase. In this case if the other nodes' MTU
2561 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2564 if (rt
->dst
.dev
== arg
->dev
&&
2565 !dst_metric_locked(&rt
->dst
, RTAX_MTU
)) {
2566 if (rt
->rt6i_flags
& RTF_CACHE
) {
2567 /* For RTF_CACHE with rt6i_pmtu == 0
2568 * (i.e. a redirected route),
2569 * the metrics of its rt->dst.from has already
2572 if (rt
->rt6i_pmtu
&& rt
->rt6i_pmtu
> arg
->mtu
)
2573 rt
->rt6i_pmtu
= arg
->mtu
;
2574 } else if (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2575 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2576 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
)) {
2577 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2583 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2585 struct rt6_mtu_change_arg arg
= {
2590 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, &arg
);
2593 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2594 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2595 [RTA_OIF
] = { .type
= NLA_U32
},
2596 [RTA_IIF
] = { .type
= NLA_U32
},
2597 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2598 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2599 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2600 [RTA_PREF
] = { .type
= NLA_U8
},
2603 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2604 struct fib6_config
*cfg
)
2607 struct nlattr
*tb
[RTA_MAX
+1];
2611 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2616 rtm
= nlmsg_data(nlh
);
2617 memset(cfg
, 0, sizeof(*cfg
));
2619 cfg
->fc_table
= rtm
->rtm_table
;
2620 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2621 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2622 cfg
->fc_flags
= RTF_UP
;
2623 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2624 cfg
->fc_type
= rtm
->rtm_type
;
2626 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2627 rtm
->rtm_type
== RTN_BLACKHOLE
||
2628 rtm
->rtm_type
== RTN_PROHIBIT
||
2629 rtm
->rtm_type
== RTN_THROW
)
2630 cfg
->fc_flags
|= RTF_REJECT
;
2632 if (rtm
->rtm_type
== RTN_LOCAL
)
2633 cfg
->fc_flags
|= RTF_LOCAL
;
2635 if (rtm
->rtm_flags
& RTM_F_CLONED
)
2636 cfg
->fc_flags
|= RTF_CACHE
;
2638 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2639 cfg
->fc_nlinfo
.nlh
= nlh
;
2640 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2642 if (tb
[RTA_GATEWAY
]) {
2643 cfg
->fc_gateway
= nla_get_in6_addr(tb
[RTA_GATEWAY
]);
2644 cfg
->fc_flags
|= RTF_GATEWAY
;
2648 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2650 if (nla_len(tb
[RTA_DST
]) < plen
)
2653 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2657 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2659 if (nla_len(tb
[RTA_SRC
]) < plen
)
2662 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2665 if (tb
[RTA_PREFSRC
])
2666 cfg
->fc_prefsrc
= nla_get_in6_addr(tb
[RTA_PREFSRC
]);
2669 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2671 if (tb
[RTA_PRIORITY
])
2672 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2674 if (tb
[RTA_METRICS
]) {
2675 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2676 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2680 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2682 if (tb
[RTA_MULTIPATH
]) {
2683 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2684 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2688 pref
= nla_get_u8(tb
[RTA_PREF
]);
2689 if (pref
!= ICMPV6_ROUTER_PREF_LOW
&&
2690 pref
!= ICMPV6_ROUTER_PREF_HIGH
)
2691 pref
= ICMPV6_ROUTER_PREF_MEDIUM
;
2692 cfg
->fc_flags
|= RTF_PREF(pref
);
2700 static int ip6_route_multipath(struct fib6_config
*cfg
, int add
)
2702 struct fib6_config r_cfg
;
2703 struct rtnexthop
*rtnh
;
2706 int err
= 0, last_err
= 0;
2708 remaining
= cfg
->fc_mp_len
;
2710 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
2712 /* Parse a Multipath Entry */
2713 while (rtnh_ok(rtnh
, remaining
)) {
2714 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
2715 if (rtnh
->rtnh_ifindex
)
2716 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
2718 attrlen
= rtnh_attrlen(rtnh
);
2720 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
2722 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
2724 r_cfg
.fc_gateway
= nla_get_in6_addr(nla
);
2725 r_cfg
.fc_flags
|= RTF_GATEWAY
;
2728 err
= add
? ip6_route_add(&r_cfg
) : ip6_route_del(&r_cfg
);
2731 /* If we are trying to remove a route, do not stop the
2732 * loop when ip6_route_del() fails (because next hop is
2733 * already gone), we should try to remove all next hops.
2736 /* If add fails, we should try to delete all
2737 * next hops that have been already added.
2740 remaining
= cfg
->fc_mp_len
- remaining
;
2744 /* Because each route is added like a single route we remove
2745 * these flags after the first nexthop: if there is a collision,
2746 * we have already failed to add the first nexthop:
2747 * fib6_add_rt2node() has rejected it; when replacing, old
2748 * nexthops have been replaced by first new, the rest should
2751 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~(NLM_F_EXCL
|
2753 rtnh
= rtnh_next(rtnh
, &remaining
);
2759 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2761 struct fib6_config cfg
;
2764 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2769 return ip6_route_multipath(&cfg
, 0);
2771 return ip6_route_del(&cfg
);
2774 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2776 struct fib6_config cfg
;
2779 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2784 return ip6_route_multipath(&cfg
, 1);
2786 return ip6_route_add(&cfg
);
2789 static inline size_t rt6_nlmsg_size(void)
2791 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2792 + nla_total_size(16) /* RTA_SRC */
2793 + nla_total_size(16) /* RTA_DST */
2794 + nla_total_size(16) /* RTA_GATEWAY */
2795 + nla_total_size(16) /* RTA_PREFSRC */
2796 + nla_total_size(4) /* RTA_TABLE */
2797 + nla_total_size(4) /* RTA_IIF */
2798 + nla_total_size(4) /* RTA_OIF */
2799 + nla_total_size(4) /* RTA_PRIORITY */
2800 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2801 + nla_total_size(sizeof(struct rta_cacheinfo
))
2802 + nla_total_size(TCP_CA_NAME_MAX
) /* RTAX_CC_ALGO */
2803 + nla_total_size(1); /* RTA_PREF */
2806 static int rt6_fill_node(struct net
*net
,
2807 struct sk_buff
*skb
, struct rt6_info
*rt
,
2808 struct in6_addr
*dst
, struct in6_addr
*src
,
2809 int iif
, int type
, u32 portid
, u32 seq
,
2810 int prefix
, int nowait
, unsigned int flags
)
2812 u32 metrics
[RTAX_MAX
];
2814 struct nlmsghdr
*nlh
;
2818 if (prefix
) { /* user wants prefix routes only */
2819 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2820 /* success since this is not a prefix route */
2825 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
2829 rtm
= nlmsg_data(nlh
);
2830 rtm
->rtm_family
= AF_INET6
;
2831 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2832 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2835 table
= rt
->rt6i_table
->tb6_id
;
2837 table
= RT6_TABLE_UNSPEC
;
2838 rtm
->rtm_table
= table
;
2839 if (nla_put_u32(skb
, RTA_TABLE
, table
))
2840 goto nla_put_failure
;
2841 if (rt
->rt6i_flags
& RTF_REJECT
) {
2842 switch (rt
->dst
.error
) {
2844 rtm
->rtm_type
= RTN_BLACKHOLE
;
2847 rtm
->rtm_type
= RTN_PROHIBIT
;
2850 rtm
->rtm_type
= RTN_THROW
;
2853 rtm
->rtm_type
= RTN_UNREACHABLE
;
2857 else if (rt
->rt6i_flags
& RTF_LOCAL
)
2858 rtm
->rtm_type
= RTN_LOCAL
;
2859 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
2860 rtm
->rtm_type
= RTN_LOCAL
;
2862 rtm
->rtm_type
= RTN_UNICAST
;
2864 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2865 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2866 if (rt
->rt6i_flags
& RTF_DYNAMIC
)
2867 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2868 else if (rt
->rt6i_flags
& RTF_ADDRCONF
) {
2869 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ROUTEINFO
))
2870 rtm
->rtm_protocol
= RTPROT_RA
;
2872 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2875 if (rt
->rt6i_flags
& RTF_CACHE
)
2876 rtm
->rtm_flags
|= RTM_F_CLONED
;
2879 if (nla_put_in6_addr(skb
, RTA_DST
, dst
))
2880 goto nla_put_failure
;
2881 rtm
->rtm_dst_len
= 128;
2882 } else if (rtm
->rtm_dst_len
)
2883 if (nla_put_in6_addr(skb
, RTA_DST
, &rt
->rt6i_dst
.addr
))
2884 goto nla_put_failure
;
2885 #ifdef CONFIG_IPV6_SUBTREES
2887 if (nla_put_in6_addr(skb
, RTA_SRC
, src
))
2888 goto nla_put_failure
;
2889 rtm
->rtm_src_len
= 128;
2890 } else if (rtm
->rtm_src_len
&&
2891 nla_put_in6_addr(skb
, RTA_SRC
, &rt
->rt6i_src
.addr
))
2892 goto nla_put_failure
;
2895 #ifdef CONFIG_IPV6_MROUTE
2896 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2897 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2902 goto nla_put_failure
;
2904 if (err
== -EMSGSIZE
)
2905 goto nla_put_failure
;
2910 if (nla_put_u32(skb
, RTA_IIF
, iif
))
2911 goto nla_put_failure
;
2913 struct in6_addr saddr_buf
;
2914 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
2915 nla_put_in6_addr(skb
, RTA_PREFSRC
, &saddr_buf
))
2916 goto nla_put_failure
;
2919 if (rt
->rt6i_prefsrc
.plen
) {
2920 struct in6_addr saddr_buf
;
2921 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
2922 if (nla_put_in6_addr(skb
, RTA_PREFSRC
, &saddr_buf
))
2923 goto nla_put_failure
;
2926 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
2928 metrics
[RTAX_MTU
- 1] = rt
->rt6i_pmtu
;
2929 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
2930 goto nla_put_failure
;
2932 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2933 if (nla_put_in6_addr(skb
, RTA_GATEWAY
, &rt
->rt6i_gateway
) < 0)
2934 goto nla_put_failure
;
2938 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2939 goto nla_put_failure
;
2940 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
2941 goto nla_put_failure
;
2943 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
2945 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
2946 goto nla_put_failure
;
2948 if (nla_put_u8(skb
, RTA_PREF
, IPV6_EXTRACT_PREF(rt
->rt6i_flags
)))
2949 goto nla_put_failure
;
2951 nlmsg_end(skb
, nlh
);
2955 nlmsg_cancel(skb
, nlh
);
2959 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2961 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2964 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2965 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2966 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2970 return rt6_fill_node(arg
->net
,
2971 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2972 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
2973 prefix
, 0, NLM_F_MULTI
);
2976 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
)
2978 struct net
*net
= sock_net(in_skb
->sk
);
2979 struct nlattr
*tb
[RTA_MAX
+1];
2980 struct rt6_info
*rt
;
2981 struct sk_buff
*skb
;
2984 int err
, iif
= 0, oif
= 0;
2986 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2991 memset(&fl6
, 0, sizeof(fl6
));
2994 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2997 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
3001 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
3004 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
3008 iif
= nla_get_u32(tb
[RTA_IIF
]);
3011 oif
= nla_get_u32(tb
[RTA_OIF
]);
3014 fl6
.flowi6_mark
= nla_get_u32(tb
[RTA_MARK
]);
3017 struct net_device
*dev
;
3020 dev
= __dev_get_by_index(net
, iif
);
3026 fl6
.flowi6_iif
= iif
;
3028 if (!ipv6_addr_any(&fl6
.saddr
))
3029 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
3031 rt
= (struct rt6_info
*)ip6_route_input_lookup(net
, dev
, &fl6
,
3034 fl6
.flowi6_oif
= oif
;
3036 rt
= (struct rt6_info
*)ip6_route_output(net
, NULL
, &fl6
);
3039 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
3046 /* Reserve room for dummy headers, this skb can pass
3047 through good chunk of routing engine.
3049 skb_reset_mac_header(skb
);
3050 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
3052 skb_dst_set(skb
, &rt
->dst
);
3054 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
3055 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
3056 nlh
->nlmsg_seq
, 0, 0, 0);
3062 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
3067 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
3069 struct sk_buff
*skb
;
3070 struct net
*net
= info
->nl_net
;
3075 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
3077 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
3081 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
3082 event
, info
->portid
, seq
, 0, 0, 0);
3084 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3085 WARN_ON(err
== -EMSGSIZE
);
3089 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
3090 info
->nlh
, gfp_any());
3094 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
3097 static int ip6_route_dev_notify(struct notifier_block
*this,
3098 unsigned long event
, void *ptr
)
3100 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3101 struct net
*net
= dev_net(dev
);
3103 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
3104 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
3105 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
3106 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3107 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
3108 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
3109 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
3110 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
3121 #ifdef CONFIG_PROC_FS
3123 static const struct file_operations ipv6_route_proc_fops
= {
3124 .owner
= THIS_MODULE
,
3125 .open
= ipv6_route_open
,
3127 .llseek
= seq_lseek
,
3128 .release
= seq_release_net
,
3131 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
3133 struct net
*net
= (struct net
*)seq
->private;
3134 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
3135 net
->ipv6
.rt6_stats
->fib_nodes
,
3136 net
->ipv6
.rt6_stats
->fib_route_nodes
,
3137 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
3138 net
->ipv6
.rt6_stats
->fib_rt_entries
,
3139 net
->ipv6
.rt6_stats
->fib_rt_cache
,
3140 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
3141 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
3146 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
3148 return single_open_net(inode
, file
, rt6_stats_seq_show
);
3151 static const struct file_operations rt6_stats_seq_fops
= {
3152 .owner
= THIS_MODULE
,
3153 .open
= rt6_stats_seq_open
,
3155 .llseek
= seq_lseek
,
3156 .release
= single_release_net
,
3158 #endif /* CONFIG_PROC_FS */
3160 #ifdef CONFIG_SYSCTL
3163 int ipv6_sysctl_rtcache_flush(struct ctl_table
*ctl
, int write
,
3164 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
3171 net
= (struct net
*)ctl
->extra1
;
3172 delay
= net
->ipv6
.sysctl
.flush_delay
;
3173 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
3174 fib6_run_gc(delay
<= 0 ? 0 : (unsigned long)delay
, net
, delay
> 0);
3178 struct ctl_table ipv6_route_table_template
[] = {
3180 .procname
= "flush",
3181 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
3182 .maxlen
= sizeof(int),
3184 .proc_handler
= ipv6_sysctl_rtcache_flush
3187 .procname
= "gc_thresh",
3188 .data
= &ip6_dst_ops_template
.gc_thresh
,
3189 .maxlen
= sizeof(int),
3191 .proc_handler
= proc_dointvec
,
3194 .procname
= "max_size",
3195 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
3196 .maxlen
= sizeof(int),
3198 .proc_handler
= proc_dointvec
,
3201 .procname
= "gc_min_interval",
3202 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
3203 .maxlen
= sizeof(int),
3205 .proc_handler
= proc_dointvec_jiffies
,
3208 .procname
= "gc_timeout",
3209 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
3210 .maxlen
= sizeof(int),
3212 .proc_handler
= proc_dointvec_jiffies
,
3215 .procname
= "gc_interval",
3216 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
3217 .maxlen
= sizeof(int),
3219 .proc_handler
= proc_dointvec_jiffies
,
3222 .procname
= "gc_elasticity",
3223 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
3224 .maxlen
= sizeof(int),
3226 .proc_handler
= proc_dointvec
,
3229 .procname
= "mtu_expires",
3230 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
3231 .maxlen
= sizeof(int),
3233 .proc_handler
= proc_dointvec_jiffies
,
3236 .procname
= "min_adv_mss",
3237 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
3238 .maxlen
= sizeof(int),
3240 .proc_handler
= proc_dointvec
,
3243 .procname
= "gc_min_interval_ms",
3244 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
3245 .maxlen
= sizeof(int),
3247 .proc_handler
= proc_dointvec_ms_jiffies
,
3252 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
3254 struct ctl_table
*table
;
3256 table
= kmemdup(ipv6_route_table_template
,
3257 sizeof(ipv6_route_table_template
),
3261 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
3262 table
[0].extra1
= net
;
3263 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
3264 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
3265 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
3266 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
3267 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
3268 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
3269 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
3270 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
3271 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
3273 /* Don't export sysctls to unprivileged users */
3274 if (net
->user_ns
!= &init_user_ns
)
3275 table
[0].procname
= NULL
;
3282 static int __net_init
ip6_route_net_init(struct net
*net
)
3286 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
3287 sizeof(net
->ipv6
.ip6_dst_ops
));
3289 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
3290 goto out_ip6_dst_ops
;
3292 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
3293 sizeof(*net
->ipv6
.ip6_null_entry
),
3295 if (!net
->ipv6
.ip6_null_entry
)
3296 goto out_ip6_dst_entries
;
3297 net
->ipv6
.ip6_null_entry
->dst
.path
=
3298 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
3299 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3300 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
3301 ip6_template_metrics
, true);
3303 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3304 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
3305 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
3307 if (!net
->ipv6
.ip6_prohibit_entry
)
3308 goto out_ip6_null_entry
;
3309 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
3310 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
3311 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3312 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
3313 ip6_template_metrics
, true);
3315 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
3316 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
3318 if (!net
->ipv6
.ip6_blk_hole_entry
)
3319 goto out_ip6_prohibit_entry
;
3320 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
3321 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
3322 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3323 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
3324 ip6_template_metrics
, true);
3327 net
->ipv6
.sysctl
.flush_delay
= 0;
3328 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
3329 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
3330 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
3331 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
3332 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
3333 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
3334 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
3336 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
3342 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3343 out_ip6_prohibit_entry
:
3344 kfree(net
->ipv6
.ip6_prohibit_entry
);
3346 kfree(net
->ipv6
.ip6_null_entry
);
3348 out_ip6_dst_entries
:
3349 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3354 static void __net_exit
ip6_route_net_exit(struct net
*net
)
3356 kfree(net
->ipv6
.ip6_null_entry
);
3357 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3358 kfree(net
->ipv6
.ip6_prohibit_entry
);
3359 kfree(net
->ipv6
.ip6_blk_hole_entry
);
3361 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3364 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3366 #ifdef CONFIG_PROC_FS
3367 proc_create("ipv6_route", 0, net
->proc_net
, &ipv6_route_proc_fops
);
3368 proc_create("rt6_stats", S_IRUGO
, net
->proc_net
, &rt6_stats_seq_fops
);
3373 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
3375 #ifdef CONFIG_PROC_FS
3376 remove_proc_entry("ipv6_route", net
->proc_net
);
3377 remove_proc_entry("rt6_stats", net
->proc_net
);
3381 static struct pernet_operations ip6_route_net_ops
= {
3382 .init
= ip6_route_net_init
,
3383 .exit
= ip6_route_net_exit
,
3386 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
3388 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
3392 inet_peer_base_init(bp
);
3393 net
->ipv6
.peers
= bp
;
3397 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
3399 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
3401 net
->ipv6
.peers
= NULL
;
3402 inetpeer_invalidate_tree(bp
);
3406 static struct pernet_operations ipv6_inetpeer_ops
= {
3407 .init
= ipv6_inetpeer_init
,
3408 .exit
= ipv6_inetpeer_exit
,
3411 static struct pernet_operations ip6_route_net_late_ops
= {
3412 .init
= ip6_route_net_init_late
,
3413 .exit
= ip6_route_net_exit_late
,
3416 static struct notifier_block ip6_route_dev_notifier
= {
3417 .notifier_call
= ip6_route_dev_notify
,
3421 int __init
ip6_route_init(void)
3427 ip6_dst_ops_template
.kmem_cachep
=
3428 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
3429 SLAB_HWCACHE_ALIGN
, NULL
);
3430 if (!ip6_dst_ops_template
.kmem_cachep
)
3433 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
3435 goto out_kmem_cache
;
3437 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
3439 goto out_dst_entries
;
3441 ret
= register_pernet_subsys(&ip6_route_net_ops
);
3443 goto out_register_inetpeer
;
3445 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
3447 /* Registering of the loopback is done before this portion of code,
3448 * the loopback reference in rt6_info will not be taken, do it
3449 * manually for init_net */
3450 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
3451 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3452 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3453 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
3454 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3455 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
3456 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3460 goto out_register_subsys
;
3466 ret
= fib6_rules_init();
3470 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
3472 goto fib6_rules_init
;
3475 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
3476 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
3477 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
3478 goto out_register_late_subsys
;
3480 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
3482 goto out_register_late_subsys
;
3484 for_each_possible_cpu(cpu
) {
3485 struct uncached_list
*ul
= per_cpu_ptr(&rt6_uncached_list
, cpu
);
3487 INIT_LIST_HEAD(&ul
->head
);
3488 spin_lock_init(&ul
->lock
);
3494 out_register_late_subsys
:
3495 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3497 fib6_rules_cleanup();
3502 out_register_subsys
:
3503 unregister_pernet_subsys(&ip6_route_net_ops
);
3504 out_register_inetpeer
:
3505 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3507 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3509 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
3513 void ip6_route_cleanup(void)
3515 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
3516 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3517 fib6_rules_cleanup();
3520 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3521 unregister_pernet_subsys(&ip6_route_net_ops
);
3522 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3523 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);