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>
61 #include <net/lwtunnel.h>
63 #include <asm/uaccess.h>
66 #include <linux/sysctl.h>
70 RT6_NUD_FAIL_HARD
= -3,
71 RT6_NUD_FAIL_PROBE
= -2,
72 RT6_NUD_FAIL_DO_RR
= -1,
76 static void ip6_rt_copy_init(struct rt6_info
*rt
, struct rt6_info
*ort
);
77 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
78 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
79 static unsigned int ip6_mtu(const struct dst_entry
*dst
);
80 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
81 static void ip6_dst_destroy(struct dst_entry
*);
82 static void ip6_dst_ifdown(struct dst_entry
*,
83 struct net_device
*dev
, int how
);
84 static int ip6_dst_gc(struct dst_ops
*ops
);
86 static int ip6_pkt_discard(struct sk_buff
*skb
);
87 static int ip6_pkt_discard_out(struct sock
*sk
, struct sk_buff
*skb
);
88 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
89 static int ip6_pkt_prohibit_out(struct sock
*sk
, struct sk_buff
*skb
);
90 static void ip6_link_failure(struct sk_buff
*skb
);
91 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
92 struct sk_buff
*skb
, u32 mtu
);
93 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
95 static void rt6_dst_from_metrics_check(struct rt6_info
*rt
);
96 static int rt6_score_route(struct rt6_info
*rt
, int oif
, int strict
);
98 #ifdef CONFIG_IPV6_ROUTE_INFO
99 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
100 const struct in6_addr
*prefix
, int prefixlen
,
101 const struct in6_addr
*gwaddr
, int ifindex
,
103 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
104 const struct in6_addr
*prefix
, int prefixlen
,
105 const struct in6_addr
*gwaddr
, int ifindex
);
108 struct uncached_list
{
110 struct list_head head
;
113 static DEFINE_PER_CPU_ALIGNED(struct uncached_list
, rt6_uncached_list
);
115 static void rt6_uncached_list_add(struct rt6_info
*rt
)
117 struct uncached_list
*ul
= raw_cpu_ptr(&rt6_uncached_list
);
119 rt
->dst
.flags
|= DST_NOCACHE
;
120 rt
->rt6i_uncached_list
= ul
;
122 spin_lock_bh(&ul
->lock
);
123 list_add_tail(&rt
->rt6i_uncached
, &ul
->head
);
124 spin_unlock_bh(&ul
->lock
);
127 static void rt6_uncached_list_del(struct rt6_info
*rt
)
129 if (!list_empty(&rt
->rt6i_uncached
)) {
130 struct uncached_list
*ul
= rt
->rt6i_uncached_list
;
132 spin_lock_bh(&ul
->lock
);
133 list_del(&rt
->rt6i_uncached
);
134 spin_unlock_bh(&ul
->lock
);
138 static void rt6_uncached_list_flush_dev(struct net
*net
, struct net_device
*dev
)
140 struct net_device
*loopback_dev
= net
->loopback_dev
;
143 for_each_possible_cpu(cpu
) {
144 struct uncached_list
*ul
= per_cpu_ptr(&rt6_uncached_list
, cpu
);
147 spin_lock_bh(&ul
->lock
);
148 list_for_each_entry(rt
, &ul
->head
, rt6i_uncached
) {
149 struct inet6_dev
*rt_idev
= rt
->rt6i_idev
;
150 struct net_device
*rt_dev
= rt
->dst
.dev
;
152 if (rt_idev
&& (rt_idev
->dev
== dev
|| !dev
) &&
153 rt_idev
->dev
!= loopback_dev
) {
154 rt
->rt6i_idev
= in6_dev_get(loopback_dev
);
155 in6_dev_put(rt_idev
);
158 if (rt_dev
&& (rt_dev
== dev
|| !dev
) &&
159 rt_dev
!= loopback_dev
) {
160 rt
->dst
.dev
= loopback_dev
;
161 dev_hold(rt
->dst
.dev
);
165 spin_unlock_bh(&ul
->lock
);
169 static u32
*rt6_pcpu_cow_metrics(struct rt6_info
*rt
)
171 return dst_metrics_write_ptr(rt
->dst
.from
);
174 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
176 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
178 if (rt
->rt6i_flags
& RTF_PCPU
)
179 return rt6_pcpu_cow_metrics(rt
);
180 else if (rt
->rt6i_flags
& RTF_CACHE
)
183 return dst_cow_metrics_generic(dst
, old
);
186 static inline const void *choose_neigh_daddr(struct rt6_info
*rt
,
190 struct in6_addr
*p
= &rt
->rt6i_gateway
;
192 if (!ipv6_addr_any(p
))
193 return (const void *) p
;
195 return &ipv6_hdr(skb
)->daddr
;
199 static struct neighbour
*ip6_neigh_lookup(const struct dst_entry
*dst
,
203 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
206 daddr
= choose_neigh_daddr(rt
, skb
, daddr
);
207 n
= __ipv6_neigh_lookup(dst
->dev
, daddr
);
210 return neigh_create(&nd_tbl
, daddr
, dst
->dev
);
213 static struct dst_ops ip6_dst_ops_template
= {
217 .check
= ip6_dst_check
,
218 .default_advmss
= ip6_default_advmss
,
220 .cow_metrics
= ipv6_cow_metrics
,
221 .destroy
= ip6_dst_destroy
,
222 .ifdown
= ip6_dst_ifdown
,
223 .negative_advice
= ip6_negative_advice
,
224 .link_failure
= ip6_link_failure
,
225 .update_pmtu
= ip6_rt_update_pmtu
,
226 .redirect
= rt6_do_redirect
,
227 .local_out
= __ip6_local_out
,
228 .neigh_lookup
= ip6_neigh_lookup
,
231 static unsigned int ip6_blackhole_mtu(const struct dst_entry
*dst
)
233 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
235 return mtu
? : dst
->dev
->mtu
;
238 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
239 struct sk_buff
*skb
, u32 mtu
)
243 static void ip6_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
248 static u32
*ip6_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
254 static struct dst_ops ip6_dst_blackhole_ops
= {
256 .destroy
= ip6_dst_destroy
,
257 .check
= ip6_dst_check
,
258 .mtu
= ip6_blackhole_mtu
,
259 .default_advmss
= ip6_default_advmss
,
260 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
261 .redirect
= ip6_rt_blackhole_redirect
,
262 .cow_metrics
= ip6_rt_blackhole_cow_metrics
,
263 .neigh_lookup
= ip6_neigh_lookup
,
266 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
267 [RTAX_HOPLIMIT
- 1] = 0,
270 static const struct rt6_info ip6_null_entry_template
= {
272 .__refcnt
= ATOMIC_INIT(1),
274 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
275 .error
= -ENETUNREACH
,
276 .input
= ip6_pkt_discard
,
277 .output
= ip6_pkt_discard_out
,
279 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
280 .rt6i_protocol
= RTPROT_KERNEL
,
281 .rt6i_metric
= ~(u32
) 0,
282 .rt6i_ref
= ATOMIC_INIT(1),
285 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
287 static const struct rt6_info ip6_prohibit_entry_template
= {
289 .__refcnt
= ATOMIC_INIT(1),
291 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
293 .input
= ip6_pkt_prohibit
,
294 .output
= ip6_pkt_prohibit_out
,
296 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
297 .rt6i_protocol
= RTPROT_KERNEL
,
298 .rt6i_metric
= ~(u32
) 0,
299 .rt6i_ref
= ATOMIC_INIT(1),
302 static const struct rt6_info ip6_blk_hole_entry_template
= {
304 .__refcnt
= ATOMIC_INIT(1),
306 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
308 .input
= dst_discard
,
309 .output
= dst_discard_sk
,
311 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
312 .rt6i_protocol
= RTPROT_KERNEL
,
313 .rt6i_metric
= ~(u32
) 0,
314 .rt6i_ref
= ATOMIC_INIT(1),
319 /* allocate dst with ip6_dst_ops */
320 static struct rt6_info
*__ip6_dst_alloc(struct net
*net
,
321 struct net_device
*dev
,
323 struct fib6_table
*table
)
325 struct rt6_info
*rt
= dst_alloc(&net
->ipv6
.ip6_dst_ops
, dev
,
326 0, DST_OBSOLETE_FORCE_CHK
, flags
);
329 struct dst_entry
*dst
= &rt
->dst
;
331 memset(dst
+ 1, 0, sizeof(*rt
) - sizeof(*dst
));
332 INIT_LIST_HEAD(&rt
->rt6i_siblings
);
333 INIT_LIST_HEAD(&rt
->rt6i_uncached
);
338 static struct rt6_info
*ip6_dst_alloc(struct net
*net
,
339 struct net_device
*dev
,
341 struct fib6_table
*table
)
343 struct rt6_info
*rt
= __ip6_dst_alloc(net
, dev
, flags
, table
);
346 rt
->rt6i_pcpu
= alloc_percpu_gfp(struct rt6_info
*, GFP_ATOMIC
);
350 for_each_possible_cpu(cpu
) {
353 p
= per_cpu_ptr(rt
->rt6i_pcpu
, cpu
);
354 /* no one shares rt */
358 dst_destroy((struct dst_entry
*)rt
);
366 static void ip6_dst_destroy(struct dst_entry
*dst
)
368 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
369 struct dst_entry
*from
= dst
->from
;
370 struct inet6_dev
*idev
;
372 dst_destroy_metrics_generic(dst
);
373 free_percpu(rt
->rt6i_pcpu
);
374 rt6_uncached_list_del(rt
);
376 idev
= rt
->rt6i_idev
;
378 rt
->rt6i_idev
= NULL
;
386 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
389 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
390 struct inet6_dev
*idev
= rt
->rt6i_idev
;
391 struct net_device
*loopback_dev
=
392 dev_net(dev
)->loopback_dev
;
394 if (dev
!= loopback_dev
) {
395 if (idev
&& idev
->dev
== dev
) {
396 struct inet6_dev
*loopback_idev
=
397 in6_dev_get(loopback_dev
);
399 rt
->rt6i_idev
= loopback_idev
;
406 static bool rt6_check_expired(const struct rt6_info
*rt
)
408 if (rt
->rt6i_flags
& RTF_EXPIRES
) {
409 if (time_after(jiffies
, rt
->dst
.expires
))
411 } else if (rt
->dst
.from
) {
412 return rt6_check_expired((struct rt6_info
*) rt
->dst
.from
);
417 /* Multipath route selection:
418 * Hash based function using packet header and flowlabel.
419 * Adapted from fib_info_hashfn()
421 static int rt6_info_hash_nhsfn(unsigned int candidate_count
,
422 const struct flowi6
*fl6
)
424 unsigned int val
= fl6
->flowi6_proto
;
426 val
^= ipv6_addr_hash(&fl6
->daddr
);
427 val
^= ipv6_addr_hash(&fl6
->saddr
);
429 /* Work only if this not encapsulated */
430 switch (fl6
->flowi6_proto
) {
434 val
^= (__force u16
)fl6
->fl6_sport
;
435 val
^= (__force u16
)fl6
->fl6_dport
;
439 val
^= (__force u16
)fl6
->fl6_icmp_type
;
440 val
^= (__force u16
)fl6
->fl6_icmp_code
;
443 /* RFC6438 recommands to use flowlabel */
444 val
^= (__force u32
)fl6
->flowlabel
;
446 /* Perhaps, we need to tune, this function? */
447 val
= val
^ (val
>> 7) ^ (val
>> 12);
448 return val
% candidate_count
;
451 static struct rt6_info
*rt6_multipath_select(struct rt6_info
*match
,
452 struct flowi6
*fl6
, int oif
,
455 struct rt6_info
*sibling
, *next_sibling
;
458 route_choosen
= rt6_info_hash_nhsfn(match
->rt6i_nsiblings
+ 1, fl6
);
459 /* Don't change the route, if route_choosen == 0
460 * (siblings does not include ourself)
463 list_for_each_entry_safe(sibling
, next_sibling
,
464 &match
->rt6i_siblings
, rt6i_siblings
) {
466 if (route_choosen
== 0) {
467 if (rt6_score_route(sibling
, oif
, strict
) < 0)
477 * Route lookup. Any table->tb6_lock is implied.
480 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
482 const struct in6_addr
*saddr
,
486 struct rt6_info
*local
= NULL
;
487 struct rt6_info
*sprt
;
489 if (!oif
&& ipv6_addr_any(saddr
))
492 for (sprt
= rt
; sprt
; sprt
= sprt
->dst
.rt6_next
) {
493 struct net_device
*dev
= sprt
->dst
.dev
;
496 if (dev
->ifindex
== oif
)
498 if (dev
->flags
& IFF_LOOPBACK
) {
499 if (!sprt
->rt6i_idev
||
500 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
501 if (flags
& RT6_LOOKUP_F_IFACE
&& oif
)
503 if (local
&& (!oif
||
504 local
->rt6i_idev
->dev
->ifindex
== oif
))
510 if (ipv6_chk_addr(net
, saddr
, dev
,
511 flags
& RT6_LOOKUP_F_IFACE
))
520 if (flags
& RT6_LOOKUP_F_IFACE
)
521 return net
->ipv6
.ip6_null_entry
;
527 #ifdef CONFIG_IPV6_ROUTER_PREF
528 struct __rt6_probe_work
{
529 struct work_struct work
;
530 struct in6_addr target
;
531 struct net_device
*dev
;
534 static void rt6_probe_deferred(struct work_struct
*w
)
536 struct in6_addr mcaddr
;
537 struct __rt6_probe_work
*work
=
538 container_of(w
, struct __rt6_probe_work
, work
);
540 addrconf_addr_solict_mult(&work
->target
, &mcaddr
);
541 ndisc_send_ns(work
->dev
, NULL
, &work
->target
, &mcaddr
, NULL
);
546 static void rt6_probe(struct rt6_info
*rt
)
548 struct __rt6_probe_work
*work
;
549 struct neighbour
*neigh
;
551 * Okay, this does not seem to be appropriate
552 * for now, however, we need to check if it
553 * is really so; aka Router Reachability Probing.
555 * Router Reachability Probe MUST be rate-limited
556 * to no more than one per minute.
558 if (!rt
|| !(rt
->rt6i_flags
& RTF_GATEWAY
))
561 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
563 if (neigh
->nud_state
& NUD_VALID
)
567 write_lock(&neigh
->lock
);
568 if (!(neigh
->nud_state
& NUD_VALID
) &&
571 rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
572 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
574 __neigh_set_probe_once(neigh
);
576 write_unlock(&neigh
->lock
);
578 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
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 rcu_read_unlock_bh();
593 static inline void rt6_probe(struct rt6_info
*rt
)
599 * Default Router Selection (RFC 2461 6.3.6)
601 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
603 struct net_device
*dev
= rt
->dst
.dev
;
604 if (!oif
|| dev
->ifindex
== oif
)
606 if ((dev
->flags
& IFF_LOOPBACK
) &&
607 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
612 static inline enum rt6_nud_state
rt6_check_neigh(struct rt6_info
*rt
)
614 struct neighbour
*neigh
;
615 enum rt6_nud_state ret
= RT6_NUD_FAIL_HARD
;
617 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
618 !(rt
->rt6i_flags
& RTF_GATEWAY
))
619 return RT6_NUD_SUCCEED
;
622 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
624 read_lock(&neigh
->lock
);
625 if (neigh
->nud_state
& NUD_VALID
)
626 ret
= RT6_NUD_SUCCEED
;
627 #ifdef CONFIG_IPV6_ROUTER_PREF
628 else if (!(neigh
->nud_state
& NUD_FAILED
))
629 ret
= RT6_NUD_SUCCEED
;
631 ret
= RT6_NUD_FAIL_PROBE
;
633 read_unlock(&neigh
->lock
);
635 ret
= IS_ENABLED(CONFIG_IPV6_ROUTER_PREF
) ?
636 RT6_NUD_SUCCEED
: RT6_NUD_FAIL_DO_RR
;
638 rcu_read_unlock_bh();
643 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
648 m
= rt6_check_dev(rt
, oif
);
649 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
650 return RT6_NUD_FAIL_HARD
;
651 #ifdef CONFIG_IPV6_ROUTER_PREF
652 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
654 if (strict
& RT6_LOOKUP_F_REACHABLE
) {
655 int n
= rt6_check_neigh(rt
);
662 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
663 int *mpri
, struct rt6_info
*match
,
667 bool match_do_rr
= false;
669 if (rt6_check_expired(rt
))
672 m
= rt6_score_route(rt
, oif
, strict
);
673 if (m
== RT6_NUD_FAIL_DO_RR
) {
675 m
= 0; /* lowest valid score */
676 } else if (m
== RT6_NUD_FAIL_HARD
) {
680 if (strict
& RT6_LOOKUP_F_REACHABLE
)
683 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
685 *do_rr
= match_do_rr
;
693 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
694 struct rt6_info
*rr_head
,
695 u32 metric
, int oif
, int strict
,
698 struct rt6_info
*rt
, *match
, *cont
;
703 for (rt
= rr_head
; rt
; rt
= rt
->dst
.rt6_next
) {
704 if (rt
->rt6i_metric
!= metric
) {
709 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
712 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
; rt
= rt
->dst
.rt6_next
) {
713 if (rt
->rt6i_metric
!= metric
) {
718 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
724 for (rt
= cont
; rt
; rt
= rt
->dst
.rt6_next
)
725 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
730 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
732 struct rt6_info
*match
, *rt0
;
738 fn
->rr_ptr
= rt0
= fn
->leaf
;
740 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
,
744 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
746 /* no entries matched; do round-robin */
747 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
754 net
= dev_net(rt0
->dst
.dev
);
755 return match
? match
: net
->ipv6
.ip6_null_entry
;
758 static bool rt6_is_gw_or_nonexthop(const struct rt6_info
*rt
)
760 return (rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
));
763 #ifdef CONFIG_IPV6_ROUTE_INFO
764 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
765 const struct in6_addr
*gwaddr
)
767 struct net
*net
= dev_net(dev
);
768 struct route_info
*rinfo
= (struct route_info
*) opt
;
769 struct in6_addr prefix_buf
, *prefix
;
771 unsigned long lifetime
;
774 if (len
< sizeof(struct route_info
)) {
778 /* Sanity check for prefix_len and length */
779 if (rinfo
->length
> 3) {
781 } else if (rinfo
->prefix_len
> 128) {
783 } else if (rinfo
->prefix_len
> 64) {
784 if (rinfo
->length
< 2) {
787 } else if (rinfo
->prefix_len
> 0) {
788 if (rinfo
->length
< 1) {
793 pref
= rinfo
->route_pref
;
794 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
797 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
799 if (rinfo
->length
== 3)
800 prefix
= (struct in6_addr
*)rinfo
->prefix
;
802 /* this function is safe */
803 ipv6_addr_prefix(&prefix_buf
,
804 (struct in6_addr
*)rinfo
->prefix
,
806 prefix
= &prefix_buf
;
809 if (rinfo
->prefix_len
== 0)
810 rt
= rt6_get_dflt_router(gwaddr
, dev
);
812 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
,
813 gwaddr
, dev
->ifindex
);
815 if (rt
&& !lifetime
) {
821 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
824 rt
->rt6i_flags
= RTF_ROUTEINFO
|
825 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
828 if (!addrconf_finite_timeout(lifetime
))
829 rt6_clean_expires(rt
);
831 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
839 static struct fib6_node
* fib6_backtrack(struct fib6_node
*fn
,
840 struct in6_addr
*saddr
)
842 struct fib6_node
*pn
;
844 if (fn
->fn_flags
& RTN_TL_ROOT
)
847 if (FIB6_SUBTREE(pn
) && FIB6_SUBTREE(pn
) != fn
)
848 fn
= fib6_lookup(FIB6_SUBTREE(pn
), NULL
, saddr
);
851 if (fn
->fn_flags
& RTN_RTINFO
)
856 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
857 struct fib6_table
*table
,
858 struct flowi6
*fl6
, int flags
)
860 struct fib6_node
*fn
;
863 read_lock_bh(&table
->tb6_lock
);
864 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
867 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
, fl6
->flowi6_oif
, flags
);
868 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
869 rt
= rt6_multipath_select(rt
, fl6
, fl6
->flowi6_oif
, flags
);
870 if (rt
== net
->ipv6
.ip6_null_entry
) {
871 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
875 dst_use(&rt
->dst
, jiffies
);
876 read_unlock_bh(&table
->tb6_lock
);
881 struct dst_entry
*ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
884 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
886 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
888 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
889 const struct in6_addr
*saddr
, int oif
, int strict
)
891 struct flowi6 fl6
= {
895 struct dst_entry
*dst
;
896 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
899 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
900 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
903 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
905 return (struct rt6_info
*) dst
;
911 EXPORT_SYMBOL(rt6_lookup
);
913 /* ip6_ins_rt is called with FREE table->tb6_lock.
914 It takes new route entry, the addition fails by any reason the
915 route is freed. In any case, if caller does not hold it, it may
919 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
,
920 struct mx6_config
*mxc
)
923 struct fib6_table
*table
;
925 table
= rt
->rt6i_table
;
926 write_lock_bh(&table
->tb6_lock
);
927 err
= fib6_add(&table
->tb6_root
, rt
, info
, mxc
);
928 write_unlock_bh(&table
->tb6_lock
);
933 int ip6_ins_rt(struct rt6_info
*rt
)
935 struct nl_info info
= { .nl_net
= dev_net(rt
->dst
.dev
), };
936 struct mx6_config mxc
= { .mx
= NULL
, };
938 return __ip6_ins_rt(rt
, &info
, &mxc
);
941 static struct rt6_info
*ip6_rt_cache_alloc(struct rt6_info
*ort
,
942 const struct in6_addr
*daddr
,
943 const struct in6_addr
*saddr
)
951 if (ort
->rt6i_flags
& (RTF_CACHE
| RTF_PCPU
))
952 ort
= (struct rt6_info
*)ort
->dst
.from
;
954 rt
= __ip6_dst_alloc(dev_net(ort
->dst
.dev
), ort
->dst
.dev
,
960 ip6_rt_copy_init(rt
, ort
);
961 rt
->rt6i_flags
|= RTF_CACHE
;
963 rt
->dst
.flags
|= DST_HOST
;
964 rt
->rt6i_dst
.addr
= *daddr
;
965 rt
->rt6i_dst
.plen
= 128;
967 if (!rt6_is_gw_or_nonexthop(ort
)) {
968 if (ort
->rt6i_dst
.plen
!= 128 &&
969 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
970 rt
->rt6i_flags
|= RTF_ANYCAST
;
971 #ifdef CONFIG_IPV6_SUBTREES
972 if (rt
->rt6i_src
.plen
&& saddr
) {
973 rt
->rt6i_src
.addr
= *saddr
;
974 rt
->rt6i_src
.plen
= 128;
982 static struct rt6_info
*ip6_rt_pcpu_alloc(struct rt6_info
*rt
)
984 struct rt6_info
*pcpu_rt
;
986 pcpu_rt
= __ip6_dst_alloc(dev_net(rt
->dst
.dev
),
987 rt
->dst
.dev
, rt
->dst
.flags
,
992 ip6_rt_copy_init(pcpu_rt
, rt
);
993 pcpu_rt
->rt6i_protocol
= rt
->rt6i_protocol
;
994 pcpu_rt
->rt6i_flags
|= RTF_PCPU
;
998 /* It should be called with read_lock_bh(&tb6_lock) acquired */
999 static struct rt6_info
*rt6_get_pcpu_route(struct rt6_info
*rt
)
1001 struct rt6_info
*pcpu_rt
, *prev
, **p
;
1003 p
= this_cpu_ptr(rt
->rt6i_pcpu
);
1009 pcpu_rt
= ip6_rt_pcpu_alloc(rt
);
1011 struct net
*net
= dev_net(rt
->dst
.dev
);
1013 pcpu_rt
= net
->ipv6
.ip6_null_entry
;
1017 prev
= cmpxchg(p
, NULL
, pcpu_rt
);
1019 /* If someone did it before us, return prev instead */
1020 dst_destroy(&pcpu_rt
->dst
);
1025 dst_hold(&pcpu_rt
->dst
);
1026 rt6_dst_from_metrics_check(pcpu_rt
);
1030 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
1031 struct flowi6
*fl6
, int flags
)
1033 struct fib6_node
*fn
, *saved_fn
;
1034 struct rt6_info
*rt
;
1037 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
1038 if (net
->ipv6
.devconf_all
->forwarding
== 0)
1039 strict
|= RT6_LOOKUP_F_REACHABLE
;
1041 read_lock_bh(&table
->tb6_lock
);
1043 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1047 rt
= rt6_select(fn
, oif
, strict
);
1048 if (rt
->rt6i_nsiblings
)
1049 rt
= rt6_multipath_select(rt
, fl6
, oif
, strict
);
1050 if (rt
== net
->ipv6
.ip6_null_entry
) {
1051 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
1053 goto redo_rt6_select
;
1054 else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
1055 /* also consider unreachable route */
1056 strict
&= ~RT6_LOOKUP_F_REACHABLE
;
1058 goto redo_rt6_select
;
1063 if (rt
== net
->ipv6
.ip6_null_entry
|| (rt
->rt6i_flags
& RTF_CACHE
)) {
1064 dst_use(&rt
->dst
, jiffies
);
1065 read_unlock_bh(&table
->tb6_lock
);
1067 rt6_dst_from_metrics_check(rt
);
1069 } else if (unlikely((fl6
->flowi6_flags
& FLOWI_FLAG_KNOWN_NH
) &&
1070 !(rt
->rt6i_flags
& RTF_GATEWAY
))) {
1071 /* Create a RTF_CACHE clone which will not be
1072 * owned by the fib6 tree. It is for the special case where
1073 * the daddr in the skb during the neighbor look-up is different
1074 * from the fl6->daddr used to look-up route here.
1077 struct rt6_info
*uncached_rt
;
1079 dst_use(&rt
->dst
, jiffies
);
1080 read_unlock_bh(&table
->tb6_lock
);
1082 uncached_rt
= ip6_rt_cache_alloc(rt
, &fl6
->daddr
, NULL
);
1083 dst_release(&rt
->dst
);
1086 rt6_uncached_list_add(uncached_rt
);
1088 uncached_rt
= net
->ipv6
.ip6_null_entry
;
1090 dst_hold(&uncached_rt
->dst
);
1094 /* Get a percpu copy */
1096 struct rt6_info
*pcpu_rt
;
1098 rt
->dst
.lastuse
= jiffies
;
1100 pcpu_rt
= rt6_get_pcpu_route(rt
);
1101 read_unlock_bh(&table
->tb6_lock
);
1107 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
1108 struct flowi6
*fl6
, int flags
)
1110 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
1113 static struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
1114 struct net_device
*dev
,
1115 struct flowi6
*fl6
, int flags
)
1117 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
1118 flags
|= RT6_LOOKUP_F_IFACE
;
1120 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
1123 void ip6_route_input(struct sk_buff
*skb
)
1125 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1126 struct net
*net
= dev_net(skb
->dev
);
1127 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1128 struct flowi6 fl6
= {
1129 .flowi6_iif
= skb
->dev
->ifindex
,
1130 .daddr
= iph
->daddr
,
1131 .saddr
= iph
->saddr
,
1132 .flowlabel
= ip6_flowinfo(iph
),
1133 .flowi6_mark
= skb
->mark
,
1134 .flowi6_proto
= iph
->nexthdr
,
1137 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
1140 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
1141 struct flowi6
*fl6
, int flags
)
1143 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1146 struct dst_entry
*ip6_route_output(struct net
*net
, const struct sock
*sk
,
1151 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1153 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
))
1154 flags
|= RT6_LOOKUP_F_IFACE
;
1156 if (!ipv6_addr_any(&fl6
->saddr
))
1157 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1159 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1161 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1163 EXPORT_SYMBOL(ip6_route_output
);
1165 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1167 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1168 struct dst_entry
*new = NULL
;
1170 rt
= dst_alloc(&ip6_dst_blackhole_ops
, ort
->dst
.dev
, 1, DST_OBSOLETE_NONE
, 0);
1174 memset(new + 1, 0, sizeof(*rt
) - sizeof(*new));
1177 new->input
= dst_discard
;
1178 new->output
= dst_discard_sk
;
1180 if (dst_metrics_read_only(&ort
->dst
))
1181 new->_metrics
= ort
->dst
._metrics
;
1183 dst_copy_metrics(new, &ort
->dst
);
1184 rt
->rt6i_idev
= ort
->rt6i_idev
;
1186 in6_dev_hold(rt
->rt6i_idev
);
1188 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1189 rt
->rt6i_flags
= ort
->rt6i_flags
;
1190 rt
->rt6i_metric
= 0;
1192 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1193 #ifdef CONFIG_IPV6_SUBTREES
1194 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1200 dst_release(dst_orig
);
1201 return new ? new : ERR_PTR(-ENOMEM
);
1205 * Destination cache support functions
1208 static void rt6_dst_from_metrics_check(struct rt6_info
*rt
)
1211 dst_metrics_ptr(&rt
->dst
) != dst_metrics_ptr(rt
->dst
.from
))
1212 dst_init_metrics(&rt
->dst
, dst_metrics_ptr(rt
->dst
.from
), true);
1215 static struct dst_entry
*rt6_check(struct rt6_info
*rt
, u32 cookie
)
1217 if (!rt
->rt6i_node
|| (rt
->rt6i_node
->fn_sernum
!= cookie
))
1220 if (rt6_check_expired(rt
))
1226 static struct dst_entry
*rt6_dst_from_check(struct rt6_info
*rt
, u32 cookie
)
1228 if (rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1229 rt6_check((struct rt6_info
*)(rt
->dst
.from
), cookie
))
1235 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1237 struct rt6_info
*rt
;
1239 rt
= (struct rt6_info
*) dst
;
1241 /* All IPV6 dsts are created with ->obsolete set to the value
1242 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1243 * into this function always.
1246 rt6_dst_from_metrics_check(rt
);
1248 if ((rt
->rt6i_flags
& RTF_PCPU
) || unlikely(dst
->flags
& DST_NOCACHE
))
1249 return rt6_dst_from_check(rt
, cookie
);
1251 return rt6_check(rt
, cookie
);
1254 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1256 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1259 if (rt
->rt6i_flags
& RTF_CACHE
) {
1260 if (rt6_check_expired(rt
)) {
1272 static void ip6_link_failure(struct sk_buff
*skb
)
1274 struct rt6_info
*rt
;
1276 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1278 rt
= (struct rt6_info
*) skb_dst(skb
);
1280 if (rt
->rt6i_flags
& RTF_CACHE
) {
1284 } else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
)) {
1285 rt
->rt6i_node
->fn_sernum
= -1;
1290 static void rt6_do_update_pmtu(struct rt6_info
*rt
, u32 mtu
)
1292 struct net
*net
= dev_net(rt
->dst
.dev
);
1294 rt
->rt6i_flags
|= RTF_MODIFIED
;
1295 rt
->rt6i_pmtu
= mtu
;
1296 rt6_update_expires(rt
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1299 static void __ip6_rt_update_pmtu(struct dst_entry
*dst
, const struct sock
*sk
,
1300 const struct ipv6hdr
*iph
, u32 mtu
)
1302 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1304 if (rt6
->rt6i_flags
& RTF_LOCAL
)
1308 mtu
= max_t(u32
, mtu
, IPV6_MIN_MTU
);
1309 if (mtu
>= dst_mtu(dst
))
1312 if (rt6
->rt6i_flags
& RTF_CACHE
) {
1313 rt6_do_update_pmtu(rt6
, mtu
);
1315 const struct in6_addr
*daddr
, *saddr
;
1316 struct rt6_info
*nrt6
;
1319 daddr
= &iph
->daddr
;
1320 saddr
= &iph
->saddr
;
1322 daddr
= &sk
->sk_v6_daddr
;
1323 saddr
= &inet6_sk(sk
)->saddr
;
1327 nrt6
= ip6_rt_cache_alloc(rt6
, daddr
, saddr
);
1329 rt6_do_update_pmtu(nrt6
, mtu
);
1331 /* ip6_ins_rt(nrt6) will bump the
1332 * rt6->rt6i_node->fn_sernum
1333 * which will fail the next rt6_check() and
1334 * invalidate the sk->sk_dst_cache.
1341 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1342 struct sk_buff
*skb
, u32 mtu
)
1344 __ip6_rt_update_pmtu(dst
, sk
, skb
? ipv6_hdr(skb
) : NULL
, mtu
);
1347 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1350 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1351 struct dst_entry
*dst
;
1354 memset(&fl6
, 0, sizeof(fl6
));
1355 fl6
.flowi6_oif
= oif
;
1356 fl6
.flowi6_mark
= mark
? mark
: IP6_REPLY_MARK(net
, skb
->mark
);
1357 fl6
.daddr
= iph
->daddr
;
1358 fl6
.saddr
= iph
->saddr
;
1359 fl6
.flowlabel
= ip6_flowinfo(iph
);
1361 dst
= ip6_route_output(net
, NULL
, &fl6
);
1363 __ip6_rt_update_pmtu(dst
, NULL
, iph
, ntohl(mtu
));
1366 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1368 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1370 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1371 sk
->sk_bound_dev_if
, sk
->sk_mark
);
1373 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1375 /* Handle redirects */
1376 struct ip6rd_flowi
{
1378 struct in6_addr gateway
;
1381 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
1382 struct fib6_table
*table
,
1386 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl6
;
1387 struct rt6_info
*rt
;
1388 struct fib6_node
*fn
;
1390 /* Get the "current" route for this destination and
1391 * check if the redirect has come from approriate router.
1393 * RFC 4861 specifies that redirects should only be
1394 * accepted if they come from the nexthop to the target.
1395 * Due to the way the routes are chosen, this notion
1396 * is a bit fuzzy and one might need to check all possible
1400 read_lock_bh(&table
->tb6_lock
);
1401 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1403 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1404 if (rt6_check_expired(rt
))
1408 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
1410 if (fl6
->flowi6_oif
!= rt
->dst
.dev
->ifindex
)
1412 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
))
1418 rt
= net
->ipv6
.ip6_null_entry
;
1419 else if (rt
->dst
.error
) {
1420 rt
= net
->ipv6
.ip6_null_entry
;
1424 if (rt
== net
->ipv6
.ip6_null_entry
) {
1425 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
1433 read_unlock_bh(&table
->tb6_lock
);
1438 static struct dst_entry
*ip6_route_redirect(struct net
*net
,
1439 const struct flowi6
*fl6
,
1440 const struct in6_addr
*gateway
)
1442 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1443 struct ip6rd_flowi rdfl
;
1446 rdfl
.gateway
= *gateway
;
1448 return fib6_rule_lookup(net
, &rdfl
.fl6
,
1449 flags
, __ip6_route_redirect
);
1452 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
)
1454 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1455 struct dst_entry
*dst
;
1458 memset(&fl6
, 0, sizeof(fl6
));
1459 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1460 fl6
.flowi6_oif
= oif
;
1461 fl6
.flowi6_mark
= mark
;
1462 fl6
.daddr
= iph
->daddr
;
1463 fl6
.saddr
= iph
->saddr
;
1464 fl6
.flowlabel
= ip6_flowinfo(iph
);
1466 dst
= ip6_route_redirect(net
, &fl6
, &ipv6_hdr(skb
)->saddr
);
1467 rt6_do_redirect(dst
, NULL
, skb
);
1470 EXPORT_SYMBOL_GPL(ip6_redirect
);
1472 void ip6_redirect_no_header(struct sk_buff
*skb
, struct net
*net
, int oif
,
1475 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1476 const struct rd_msg
*msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1477 struct dst_entry
*dst
;
1480 memset(&fl6
, 0, sizeof(fl6
));
1481 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1482 fl6
.flowi6_oif
= oif
;
1483 fl6
.flowi6_mark
= mark
;
1484 fl6
.daddr
= msg
->dest
;
1485 fl6
.saddr
= iph
->daddr
;
1487 dst
= ip6_route_redirect(net
, &fl6
, &iph
->saddr
);
1488 rt6_do_redirect(dst
, NULL
, skb
);
1492 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1494 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
);
1496 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1498 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1500 struct net_device
*dev
= dst
->dev
;
1501 unsigned int mtu
= dst_mtu(dst
);
1502 struct net
*net
= dev_net(dev
);
1504 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1506 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1507 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1510 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1511 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1512 * IPV6_MAXPLEN is also valid and means: "any MSS,
1513 * rely only on pmtu discovery"
1515 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1520 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1522 const struct rt6_info
*rt
= (const struct rt6_info
*)dst
;
1523 unsigned int mtu
= rt
->rt6i_pmtu
;
1524 struct inet6_dev
*idev
;
1529 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1536 idev
= __in6_dev_get(dst
->dev
);
1538 mtu
= idev
->cnf
.mtu6
;
1542 return min_t(unsigned int, mtu
, IP6_MAX_MTU
);
1545 static struct dst_entry
*icmp6_dst_gc_list
;
1546 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1548 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1551 struct dst_entry
*dst
;
1552 struct rt6_info
*rt
;
1553 struct inet6_dev
*idev
= in6_dev_get(dev
);
1554 struct net
*net
= dev_net(dev
);
1556 if (unlikely(!idev
))
1557 return ERR_PTR(-ENODEV
);
1559 rt
= ip6_dst_alloc(net
, dev
, 0, NULL
);
1560 if (unlikely(!rt
)) {
1562 dst
= ERR_PTR(-ENOMEM
);
1566 rt
->dst
.flags
|= DST_HOST
;
1567 rt
->dst
.output
= ip6_output
;
1568 atomic_set(&rt
->dst
.__refcnt
, 1);
1569 rt
->rt6i_gateway
= fl6
->daddr
;
1570 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1571 rt
->rt6i_dst
.plen
= 128;
1572 rt
->rt6i_idev
= idev
;
1573 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1575 spin_lock_bh(&icmp6_dst_lock
);
1576 rt
->dst
.next
= icmp6_dst_gc_list
;
1577 icmp6_dst_gc_list
= &rt
->dst
;
1578 spin_unlock_bh(&icmp6_dst_lock
);
1580 fib6_force_start_gc(net
);
1582 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1588 int icmp6_dst_gc(void)
1590 struct dst_entry
*dst
, **pprev
;
1593 spin_lock_bh(&icmp6_dst_lock
);
1594 pprev
= &icmp6_dst_gc_list
;
1596 while ((dst
= *pprev
) != NULL
) {
1597 if (!atomic_read(&dst
->__refcnt
)) {
1606 spin_unlock_bh(&icmp6_dst_lock
);
1611 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1614 struct dst_entry
*dst
, **pprev
;
1616 spin_lock_bh(&icmp6_dst_lock
);
1617 pprev
= &icmp6_dst_gc_list
;
1618 while ((dst
= *pprev
) != NULL
) {
1619 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1620 if (func(rt
, arg
)) {
1627 spin_unlock_bh(&icmp6_dst_lock
);
1630 static int ip6_dst_gc(struct dst_ops
*ops
)
1632 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1633 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1634 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1635 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1636 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1637 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1640 entries
= dst_entries_get_fast(ops
);
1641 if (time_after(rt_last_gc
+ rt_min_interval
, jiffies
) &&
1642 entries
<= rt_max_size
)
1645 net
->ipv6
.ip6_rt_gc_expire
++;
1646 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
, true);
1647 entries
= dst_entries_get_slow(ops
);
1648 if (entries
< ops
->gc_thresh
)
1649 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1651 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1652 return entries
> rt_max_size
;
1655 static int ip6_convert_metrics(struct mx6_config
*mxc
,
1656 const struct fib6_config
*cfg
)
1665 mp
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_KERNEL
);
1669 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1670 int type
= nla_type(nla
);
1675 if (unlikely(type
> RTAX_MAX
))
1677 if (type
== RTAX_CC_ALGO
) {
1678 char tmp
[TCP_CA_NAME_MAX
];
1680 nla_strlcpy(tmp
, nla
, sizeof(tmp
));
1681 val
= tcp_ca_get_key_by_name(tmp
);
1682 if (val
== TCP_CA_UNSPEC
)
1685 val
= nla_get_u32(nla
);
1689 __set_bit(type
- 1, mxc
->mx_valid
);
1701 int ip6_route_add(struct fib6_config
*cfg
)
1704 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1705 struct rt6_info
*rt
= NULL
;
1706 struct net_device
*dev
= NULL
;
1707 struct inet6_dev
*idev
= NULL
;
1708 struct fib6_table
*table
;
1709 struct mx6_config mxc
= { .mx
= NULL
, };
1712 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1714 #ifndef CONFIG_IPV6_SUBTREES
1715 if (cfg
->fc_src_len
)
1718 if (cfg
->fc_ifindex
) {
1720 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1723 idev
= in6_dev_get(dev
);
1728 if (cfg
->fc_metric
== 0)
1729 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1732 if (cfg
->fc_nlinfo
.nlh
&&
1733 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1734 table
= fib6_get_table(net
, cfg
->fc_table
);
1736 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1737 table
= fib6_new_table(net
, cfg
->fc_table
);
1740 table
= fib6_new_table(net
, cfg
->fc_table
);
1746 rt
= ip6_dst_alloc(net
, NULL
, (cfg
->fc_flags
& RTF_ADDRCONF
) ? 0 : DST_NOCOUNT
, table
);
1753 if (cfg
->fc_flags
& RTF_EXPIRES
)
1754 rt6_set_expires(rt
, jiffies
+
1755 clock_t_to_jiffies(cfg
->fc_expires
));
1757 rt6_clean_expires(rt
);
1759 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1760 cfg
->fc_protocol
= RTPROT_BOOT
;
1761 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1763 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1765 if (addr_type
& IPV6_ADDR_MULTICAST
)
1766 rt
->dst
.input
= ip6_mc_input
;
1767 else if (cfg
->fc_flags
& RTF_LOCAL
)
1768 rt
->dst
.input
= ip6_input
;
1770 rt
->dst
.input
= ip6_forward
;
1772 rt
->dst
.output
= ip6_output
;
1774 if (cfg
->fc_encap
) {
1775 struct lwtunnel_state
*lwtstate
;
1777 err
= lwtunnel_build_state(dev
, cfg
->fc_encap_type
,
1778 cfg
->fc_encap
, &lwtstate
);
1781 rt
->rt6i_lwtstate
= lwtstate_get(lwtstate
);
1782 if (lwtunnel_output_redirect(rt
->rt6i_lwtstate
))
1783 rt
->dst
.output
= lwtunnel_output6
;
1786 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1787 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1788 if (rt
->rt6i_dst
.plen
== 128)
1789 rt
->dst
.flags
|= DST_HOST
;
1791 #ifdef CONFIG_IPV6_SUBTREES
1792 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1793 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1796 rt
->rt6i_metric
= cfg
->fc_metric
;
1798 /* We cannot add true routes via loopback here,
1799 they would result in kernel looping; promote them to reject routes
1801 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1802 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1803 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1804 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1805 /* hold loopback dev/idev if we haven't done so. */
1806 if (dev
!= net
->loopback_dev
) {
1811 dev
= net
->loopback_dev
;
1813 idev
= in6_dev_get(dev
);
1819 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1820 switch (cfg
->fc_type
) {
1822 rt
->dst
.error
= -EINVAL
;
1823 rt
->dst
.output
= dst_discard_sk
;
1824 rt
->dst
.input
= dst_discard
;
1827 rt
->dst
.error
= -EACCES
;
1828 rt
->dst
.output
= ip6_pkt_prohibit_out
;
1829 rt
->dst
.input
= ip6_pkt_prohibit
;
1833 rt
->dst
.error
= (cfg
->fc_type
== RTN_THROW
) ? -EAGAIN
1835 rt
->dst
.output
= ip6_pkt_discard_out
;
1836 rt
->dst
.input
= ip6_pkt_discard
;
1842 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1843 const struct in6_addr
*gw_addr
;
1846 gw_addr
= &cfg
->fc_gateway
;
1848 /* if gw_addr is local we will fail to detect this in case
1849 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1850 * will return already-added prefix route via interface that
1851 * prefix route was assigned to, which might be non-loopback.
1854 if (ipv6_chk_addr_and_flags(net
, gw_addr
, NULL
, 0, 0))
1857 rt
->rt6i_gateway
= *gw_addr
;
1858 gwa_type
= ipv6_addr_type(gw_addr
);
1860 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1861 struct rt6_info
*grt
;
1863 /* IPv6 strictly inhibits using not link-local
1864 addresses as nexthop address.
1865 Otherwise, router will not able to send redirects.
1866 It is very good, but in some (rare!) circumstances
1867 (SIT, PtP, NBMA NOARP links) it is handy to allow
1868 some exceptions. --ANK
1870 if (!(gwa_type
& IPV6_ADDR_UNICAST
))
1873 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1875 err
= -EHOSTUNREACH
;
1879 if (dev
!= grt
->dst
.dev
) {
1885 idev
= grt
->rt6i_idev
;
1887 in6_dev_hold(grt
->rt6i_idev
);
1889 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
1897 if (!dev
|| (dev
->flags
& IFF_LOOPBACK
))
1905 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
1906 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
1910 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
1911 rt
->rt6i_prefsrc
.plen
= 128;
1913 rt
->rt6i_prefsrc
.plen
= 0;
1915 rt
->rt6i_flags
= cfg
->fc_flags
;
1919 rt
->rt6i_idev
= idev
;
1920 rt
->rt6i_table
= table
;
1922 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1924 err
= ip6_convert_metrics(&mxc
, cfg
);
1928 err
= __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
, &mxc
);
1942 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1945 struct fib6_table
*table
;
1946 struct net
*net
= dev_net(rt
->dst
.dev
);
1948 if (rt
== net
->ipv6
.ip6_null_entry
) {
1953 table
= rt
->rt6i_table
;
1954 write_lock_bh(&table
->tb6_lock
);
1955 err
= fib6_del(rt
, info
);
1956 write_unlock_bh(&table
->tb6_lock
);
1963 int ip6_del_rt(struct rt6_info
*rt
)
1965 struct nl_info info
= {
1966 .nl_net
= dev_net(rt
->dst
.dev
),
1968 return __ip6_del_rt(rt
, &info
);
1971 static int ip6_route_del(struct fib6_config
*cfg
)
1973 struct fib6_table
*table
;
1974 struct fib6_node
*fn
;
1975 struct rt6_info
*rt
;
1978 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1982 read_lock_bh(&table
->tb6_lock
);
1984 fn
= fib6_locate(&table
->tb6_root
,
1985 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1986 &cfg
->fc_src
, cfg
->fc_src_len
);
1989 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1990 if ((rt
->rt6i_flags
& RTF_CACHE
) &&
1991 !(cfg
->fc_flags
& RTF_CACHE
))
1993 if (cfg
->fc_ifindex
&&
1995 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
1997 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1998 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
2000 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
2003 read_unlock_bh(&table
->tb6_lock
);
2005 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
2008 read_unlock_bh(&table
->tb6_lock
);
2013 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
2015 struct net
*net
= dev_net(skb
->dev
);
2016 struct netevent_redirect netevent
;
2017 struct rt6_info
*rt
, *nrt
= NULL
;
2018 struct ndisc_options ndopts
;
2019 struct inet6_dev
*in6_dev
;
2020 struct neighbour
*neigh
;
2022 int optlen
, on_link
;
2025 optlen
= skb_tail_pointer(skb
) - skb_transport_header(skb
);
2026 optlen
-= sizeof(*msg
);
2029 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2033 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
2035 if (ipv6_addr_is_multicast(&msg
->dest
)) {
2036 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2041 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
2043 } else if (ipv6_addr_type(&msg
->target
) !=
2044 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
2045 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2049 in6_dev
= __in6_dev_get(skb
->dev
);
2052 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
2056 * The IP source address of the Redirect MUST be the same as the current
2057 * first-hop router for the specified ICMP Destination Address.
2060 if (!ndisc_parse_options(msg
->opt
, optlen
, &ndopts
)) {
2061 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2066 if (ndopts
.nd_opts_tgt_lladdr
) {
2067 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
2070 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2075 rt
= (struct rt6_info
*) dst
;
2076 if (rt
== net
->ipv6
.ip6_null_entry
) {
2077 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2081 /* Redirect received -> path was valid.
2082 * Look, redirects are sent only in response to data packets,
2083 * so that this nexthop apparently is reachable. --ANK
2085 dst_confirm(&rt
->dst
);
2087 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
2092 * We have finally decided to accept it.
2095 neigh_update(neigh
, lladdr
, NUD_STALE
,
2096 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
2097 NEIGH_UPDATE_F_OVERRIDE
|
2098 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
2099 NEIGH_UPDATE_F_ISROUTER
))
2102 nrt
= ip6_rt_cache_alloc(rt
, &msg
->dest
, NULL
);
2106 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
2108 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
2110 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
2112 if (ip6_ins_rt(nrt
))
2115 netevent
.old
= &rt
->dst
;
2116 netevent
.new = &nrt
->dst
;
2117 netevent
.daddr
= &msg
->dest
;
2118 netevent
.neigh
= neigh
;
2119 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
2121 if (rt
->rt6i_flags
& RTF_CACHE
) {
2122 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
2127 neigh_release(neigh
);
2131 * Misc support functions
2134 static void rt6_set_from(struct rt6_info
*rt
, struct rt6_info
*from
)
2136 BUG_ON(from
->dst
.from
);
2138 rt
->rt6i_flags
&= ~RTF_EXPIRES
;
2139 dst_hold(&from
->dst
);
2140 rt
->dst
.from
= &from
->dst
;
2141 dst_init_metrics(&rt
->dst
, dst_metrics_ptr(&from
->dst
), true);
2144 static void ip6_rt_copy_init(struct rt6_info
*rt
, struct rt6_info
*ort
)
2146 rt
->dst
.input
= ort
->dst
.input
;
2147 rt
->dst
.output
= ort
->dst
.output
;
2148 rt
->rt6i_dst
= ort
->rt6i_dst
;
2149 rt
->dst
.error
= ort
->dst
.error
;
2150 rt
->rt6i_idev
= ort
->rt6i_idev
;
2152 in6_dev_hold(rt
->rt6i_idev
);
2153 rt
->dst
.lastuse
= jiffies
;
2154 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
2155 rt
->rt6i_flags
= ort
->rt6i_flags
;
2156 rt6_set_from(rt
, ort
);
2157 rt
->rt6i_metric
= ort
->rt6i_metric
;
2158 #ifdef CONFIG_IPV6_SUBTREES
2159 rt
->rt6i_src
= ort
->rt6i_src
;
2161 rt
->rt6i_prefsrc
= ort
->rt6i_prefsrc
;
2162 rt
->rt6i_table
= ort
->rt6i_table
;
2163 rt
->rt6i_lwtstate
= lwtstate_get(ort
->rt6i_lwtstate
);
2166 #ifdef CONFIG_IPV6_ROUTE_INFO
2167 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
2168 const struct in6_addr
*prefix
, int prefixlen
,
2169 const struct in6_addr
*gwaddr
, int ifindex
)
2171 struct fib6_node
*fn
;
2172 struct rt6_info
*rt
= NULL
;
2173 struct fib6_table
*table
;
2175 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
2179 read_lock_bh(&table
->tb6_lock
);
2180 fn
= fib6_locate(&table
->tb6_root
, prefix
, prefixlen
, NULL
, 0);
2184 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2185 if (rt
->dst
.dev
->ifindex
!= ifindex
)
2187 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
2189 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
2195 read_unlock_bh(&table
->tb6_lock
);
2199 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
2200 const struct in6_addr
*prefix
, int prefixlen
,
2201 const struct in6_addr
*gwaddr
, int ifindex
,
2204 struct fib6_config cfg
= {
2205 .fc_table
= RT6_TABLE_INFO
,
2206 .fc_metric
= IP6_RT_PRIO_USER
,
2207 .fc_ifindex
= ifindex
,
2208 .fc_dst_len
= prefixlen
,
2209 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
2210 RTF_UP
| RTF_PREF(pref
),
2211 .fc_nlinfo
.portid
= 0,
2212 .fc_nlinfo
.nlh
= NULL
,
2213 .fc_nlinfo
.nl_net
= net
,
2216 cfg
.fc_dst
= *prefix
;
2217 cfg
.fc_gateway
= *gwaddr
;
2219 /* We should treat it as a default route if prefix length is 0. */
2221 cfg
.fc_flags
|= RTF_DEFAULT
;
2223 ip6_route_add(&cfg
);
2225 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
2229 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
2231 struct rt6_info
*rt
;
2232 struct fib6_table
*table
;
2234 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
2238 read_lock_bh(&table
->tb6_lock
);
2239 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2240 if (dev
== rt
->dst
.dev
&&
2241 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
2242 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
2247 read_unlock_bh(&table
->tb6_lock
);
2251 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
2252 struct net_device
*dev
,
2255 struct fib6_config cfg
= {
2256 .fc_table
= RT6_TABLE_DFLT
,
2257 .fc_metric
= IP6_RT_PRIO_USER
,
2258 .fc_ifindex
= dev
->ifindex
,
2259 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
2260 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
2261 .fc_nlinfo
.portid
= 0,
2262 .fc_nlinfo
.nlh
= NULL
,
2263 .fc_nlinfo
.nl_net
= dev_net(dev
),
2266 cfg
.fc_gateway
= *gwaddr
;
2268 ip6_route_add(&cfg
);
2270 return rt6_get_dflt_router(gwaddr
, dev
);
2273 void rt6_purge_dflt_routers(struct net
*net
)
2275 struct rt6_info
*rt
;
2276 struct fib6_table
*table
;
2278 /* NOTE: Keep consistent with rt6_get_dflt_router */
2279 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
2284 read_lock_bh(&table
->tb6_lock
);
2285 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2286 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
) &&
2287 (!rt
->rt6i_idev
|| rt
->rt6i_idev
->cnf
.accept_ra
!= 2)) {
2289 read_unlock_bh(&table
->tb6_lock
);
2294 read_unlock_bh(&table
->tb6_lock
);
2297 static void rtmsg_to_fib6_config(struct net
*net
,
2298 struct in6_rtmsg
*rtmsg
,
2299 struct fib6_config
*cfg
)
2301 memset(cfg
, 0, sizeof(*cfg
));
2303 cfg
->fc_table
= RT6_TABLE_MAIN
;
2304 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
2305 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
2306 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
2307 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
2308 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
2309 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
2311 cfg
->fc_nlinfo
.nl_net
= net
;
2313 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
2314 cfg
->fc_src
= rtmsg
->rtmsg_src
;
2315 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
2318 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
2320 struct fib6_config cfg
;
2321 struct in6_rtmsg rtmsg
;
2325 case SIOCADDRT
: /* Add a route */
2326 case SIOCDELRT
: /* Delete a route */
2327 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2329 err
= copy_from_user(&rtmsg
, arg
,
2330 sizeof(struct in6_rtmsg
));
2334 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
2339 err
= ip6_route_add(&cfg
);
2342 err
= ip6_route_del(&cfg
);
2356 * Drop the packet on the floor
2359 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2362 struct dst_entry
*dst
= skb_dst(skb
);
2363 switch (ipstats_mib_noroutes
) {
2364 case IPSTATS_MIB_INNOROUTES
:
2365 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2366 if (type
== IPV6_ADDR_ANY
) {
2367 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2368 IPSTATS_MIB_INADDRERRORS
);
2372 case IPSTATS_MIB_OUTNOROUTES
:
2373 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2374 ipstats_mib_noroutes
);
2377 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2382 static int ip6_pkt_discard(struct sk_buff
*skb
)
2384 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2387 static int ip6_pkt_discard_out(struct sock
*sk
, struct sk_buff
*skb
)
2389 skb
->dev
= skb_dst(skb
)->dev
;
2390 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2393 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2395 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2398 static int ip6_pkt_prohibit_out(struct sock
*sk
, struct sk_buff
*skb
)
2400 skb
->dev
= skb_dst(skb
)->dev
;
2401 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2405 * Allocate a dst for local (unicast / anycast) address.
2408 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2409 const struct in6_addr
*addr
,
2412 struct net
*net
= dev_net(idev
->dev
);
2413 struct rt6_info
*rt
= ip6_dst_alloc(net
, net
->loopback_dev
,
2416 return ERR_PTR(-ENOMEM
);
2420 rt
->dst
.flags
|= DST_HOST
;
2421 rt
->dst
.input
= ip6_input
;
2422 rt
->dst
.output
= ip6_output
;
2423 rt
->rt6i_idev
= idev
;
2425 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2427 rt
->rt6i_flags
|= RTF_ANYCAST
;
2429 rt
->rt6i_flags
|= RTF_LOCAL
;
2431 rt
->rt6i_gateway
= *addr
;
2432 rt
->rt6i_dst
.addr
= *addr
;
2433 rt
->rt6i_dst
.plen
= 128;
2434 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2436 atomic_set(&rt
->dst
.__refcnt
, 1);
2441 int ip6_route_get_saddr(struct net
*net
,
2442 struct rt6_info
*rt
,
2443 const struct in6_addr
*daddr
,
2445 struct in6_addr
*saddr
)
2447 struct inet6_dev
*idev
=
2448 rt
? ip6_dst_idev((struct dst_entry
*)rt
) : NULL
;
2450 if (rt
&& rt
->rt6i_prefsrc
.plen
)
2451 *saddr
= rt
->rt6i_prefsrc
.addr
;
2453 err
= ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2454 daddr
, prefs
, saddr
);
2458 /* remove deleted ip from prefsrc entries */
2459 struct arg_dev_net_ip
{
2460 struct net_device
*dev
;
2462 struct in6_addr
*addr
;
2465 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2467 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2468 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2469 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2471 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2472 rt
!= net
->ipv6
.ip6_null_entry
&&
2473 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2474 /* remove prefsrc entry */
2475 rt
->rt6i_prefsrc
.plen
= 0;
2480 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2482 struct net
*net
= dev_net(ifp
->idev
->dev
);
2483 struct arg_dev_net_ip adni
= {
2484 .dev
= ifp
->idev
->dev
,
2488 fib6_clean_all(net
, fib6_remove_prefsrc
, &adni
);
2491 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2492 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2494 /* Remove routers and update dst entries when gateway turn into host. */
2495 static int fib6_clean_tohost(struct rt6_info
*rt
, void *arg
)
2497 struct in6_addr
*gateway
= (struct in6_addr
*)arg
;
2499 if ((((rt
->rt6i_flags
& RTF_RA_ROUTER
) == RTF_RA_ROUTER
) ||
2500 ((rt
->rt6i_flags
& RTF_CACHE_GATEWAY
) == RTF_CACHE_GATEWAY
)) &&
2501 ipv6_addr_equal(gateway
, &rt
->rt6i_gateway
)) {
2507 void rt6_clean_tohost(struct net
*net
, struct in6_addr
*gateway
)
2509 fib6_clean_all(net
, fib6_clean_tohost
, gateway
);
2512 struct arg_dev_net
{
2513 struct net_device
*dev
;
2517 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2519 const struct arg_dev_net
*adn
= arg
;
2520 const struct net_device
*dev
= adn
->dev
;
2522 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2523 rt
!= adn
->net
->ipv6
.ip6_null_entry
)
2529 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2531 struct arg_dev_net adn
= {
2536 fib6_clean_all(net
, fib6_ifdown
, &adn
);
2537 icmp6_clean_all(fib6_ifdown
, &adn
);
2538 rt6_uncached_list_flush_dev(net
, dev
);
2541 struct rt6_mtu_change_arg
{
2542 struct net_device
*dev
;
2546 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2548 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2549 struct inet6_dev
*idev
;
2551 /* In IPv6 pmtu discovery is not optional,
2552 so that RTAX_MTU lock cannot disable it.
2553 We still use this lock to block changes
2554 caused by addrconf/ndisc.
2557 idev
= __in6_dev_get(arg
->dev
);
2561 /* For administrative MTU increase, there is no way to discover
2562 IPv6 PMTU increase, so PMTU increase should be updated here.
2563 Since RFC 1981 doesn't include administrative MTU increase
2564 update PMTU increase is a MUST. (i.e. jumbo frame)
2567 If new MTU is less than route PMTU, this new MTU will be the
2568 lowest MTU in the path, update the route PMTU to reflect PMTU
2569 decreases; if new MTU is greater than route PMTU, and the
2570 old MTU is the lowest MTU in the path, update the route PMTU
2571 to reflect the increase. In this case if the other nodes' MTU
2572 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2575 if (rt
->dst
.dev
== arg
->dev
&&
2576 !dst_metric_locked(&rt
->dst
, RTAX_MTU
)) {
2577 if (rt
->rt6i_flags
& RTF_CACHE
) {
2578 /* For RTF_CACHE with rt6i_pmtu == 0
2579 * (i.e. a redirected route),
2580 * the metrics of its rt->dst.from has already
2583 if (rt
->rt6i_pmtu
&& rt
->rt6i_pmtu
> arg
->mtu
)
2584 rt
->rt6i_pmtu
= arg
->mtu
;
2585 } else if (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2586 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2587 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
)) {
2588 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2594 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2596 struct rt6_mtu_change_arg arg
= {
2601 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, &arg
);
2604 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2605 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2606 [RTA_OIF
] = { .type
= NLA_U32
},
2607 [RTA_IIF
] = { .type
= NLA_U32
},
2608 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2609 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2610 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2611 [RTA_PREF
] = { .type
= NLA_U8
},
2612 [RTA_ENCAP_TYPE
] = { .type
= NLA_U16
},
2613 [RTA_ENCAP
] = { .type
= NLA_NESTED
},
2616 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2617 struct fib6_config
*cfg
)
2620 struct nlattr
*tb
[RTA_MAX
+1];
2624 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2629 rtm
= nlmsg_data(nlh
);
2630 memset(cfg
, 0, sizeof(*cfg
));
2632 cfg
->fc_table
= rtm
->rtm_table
;
2633 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2634 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2635 cfg
->fc_flags
= RTF_UP
;
2636 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2637 cfg
->fc_type
= rtm
->rtm_type
;
2639 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2640 rtm
->rtm_type
== RTN_BLACKHOLE
||
2641 rtm
->rtm_type
== RTN_PROHIBIT
||
2642 rtm
->rtm_type
== RTN_THROW
)
2643 cfg
->fc_flags
|= RTF_REJECT
;
2645 if (rtm
->rtm_type
== RTN_LOCAL
)
2646 cfg
->fc_flags
|= RTF_LOCAL
;
2648 if (rtm
->rtm_flags
& RTM_F_CLONED
)
2649 cfg
->fc_flags
|= RTF_CACHE
;
2651 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2652 cfg
->fc_nlinfo
.nlh
= nlh
;
2653 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2655 if (tb
[RTA_GATEWAY
]) {
2656 cfg
->fc_gateway
= nla_get_in6_addr(tb
[RTA_GATEWAY
]);
2657 cfg
->fc_flags
|= RTF_GATEWAY
;
2661 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2663 if (nla_len(tb
[RTA_DST
]) < plen
)
2666 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2670 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2672 if (nla_len(tb
[RTA_SRC
]) < plen
)
2675 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2678 if (tb
[RTA_PREFSRC
])
2679 cfg
->fc_prefsrc
= nla_get_in6_addr(tb
[RTA_PREFSRC
]);
2682 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2684 if (tb
[RTA_PRIORITY
])
2685 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2687 if (tb
[RTA_METRICS
]) {
2688 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2689 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2693 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2695 if (tb
[RTA_MULTIPATH
]) {
2696 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2697 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2701 pref
= nla_get_u8(tb
[RTA_PREF
]);
2702 if (pref
!= ICMPV6_ROUTER_PREF_LOW
&&
2703 pref
!= ICMPV6_ROUTER_PREF_HIGH
)
2704 pref
= ICMPV6_ROUTER_PREF_MEDIUM
;
2705 cfg
->fc_flags
|= RTF_PREF(pref
);
2709 cfg
->fc_encap
= tb
[RTA_ENCAP
];
2711 if (tb
[RTA_ENCAP_TYPE
])
2712 cfg
->fc_encap_type
= nla_get_u16(tb
[RTA_ENCAP_TYPE
]);
2719 static int ip6_route_multipath(struct fib6_config
*cfg
, int add
)
2721 struct fib6_config r_cfg
;
2722 struct rtnexthop
*rtnh
;
2725 int err
= 0, last_err
= 0;
2727 remaining
= cfg
->fc_mp_len
;
2729 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
2731 /* Parse a Multipath Entry */
2732 while (rtnh_ok(rtnh
, remaining
)) {
2733 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
2734 if (rtnh
->rtnh_ifindex
)
2735 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
2737 attrlen
= rtnh_attrlen(rtnh
);
2739 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
2741 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
2743 r_cfg
.fc_gateway
= nla_get_in6_addr(nla
);
2744 r_cfg
.fc_flags
|= RTF_GATEWAY
;
2746 r_cfg
.fc_encap
= nla_find(attrs
, attrlen
, RTA_ENCAP
);
2747 nla
= nla_find(attrs
, attrlen
, RTA_ENCAP_TYPE
);
2749 r_cfg
.fc_encap_type
= nla_get_u16(nla
);
2751 err
= add
? ip6_route_add(&r_cfg
) : ip6_route_del(&r_cfg
);
2754 /* If we are trying to remove a route, do not stop the
2755 * loop when ip6_route_del() fails (because next hop is
2756 * already gone), we should try to remove all next hops.
2759 /* If add fails, we should try to delete all
2760 * next hops that have been already added.
2763 remaining
= cfg
->fc_mp_len
- remaining
;
2767 /* Because each route is added like a single route we remove
2768 * these flags after the first nexthop: if there is a collision,
2769 * we have already failed to add the first nexthop:
2770 * fib6_add_rt2node() has rejected it; when replacing, old
2771 * nexthops have been replaced by first new, the rest should
2774 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~(NLM_F_EXCL
|
2776 rtnh
= rtnh_next(rtnh
, &remaining
);
2782 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2784 struct fib6_config cfg
;
2787 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2792 return ip6_route_multipath(&cfg
, 0);
2794 return ip6_route_del(&cfg
);
2797 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
2799 struct fib6_config cfg
;
2802 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2807 return ip6_route_multipath(&cfg
, 1);
2809 return ip6_route_add(&cfg
);
2812 static inline size_t rt6_nlmsg_size(struct rt6_info
*rt
)
2814 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2815 + nla_total_size(16) /* RTA_SRC */
2816 + nla_total_size(16) /* RTA_DST */
2817 + nla_total_size(16) /* RTA_GATEWAY */
2818 + nla_total_size(16) /* RTA_PREFSRC */
2819 + nla_total_size(4) /* RTA_TABLE */
2820 + nla_total_size(4) /* RTA_IIF */
2821 + nla_total_size(4) /* RTA_OIF */
2822 + nla_total_size(4) /* RTA_PRIORITY */
2823 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2824 + nla_total_size(sizeof(struct rta_cacheinfo
))
2825 + nla_total_size(TCP_CA_NAME_MAX
) /* RTAX_CC_ALGO */
2826 + nla_total_size(1) /* RTA_PREF */
2827 + lwtunnel_get_encap_size(rt
->rt6i_lwtstate
);
2830 static int rt6_fill_node(struct net
*net
,
2831 struct sk_buff
*skb
, struct rt6_info
*rt
,
2832 struct in6_addr
*dst
, struct in6_addr
*src
,
2833 int iif
, int type
, u32 portid
, u32 seq
,
2834 int prefix
, int nowait
, unsigned int flags
)
2836 u32 metrics
[RTAX_MAX
];
2838 struct nlmsghdr
*nlh
;
2842 if (prefix
) { /* user wants prefix routes only */
2843 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2844 /* success since this is not a prefix route */
2849 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
2853 rtm
= nlmsg_data(nlh
);
2854 rtm
->rtm_family
= AF_INET6
;
2855 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2856 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2859 table
= rt
->rt6i_table
->tb6_id
;
2861 table
= RT6_TABLE_UNSPEC
;
2862 rtm
->rtm_table
= table
;
2863 if (nla_put_u32(skb
, RTA_TABLE
, table
))
2864 goto nla_put_failure
;
2865 if (rt
->rt6i_flags
& RTF_REJECT
) {
2866 switch (rt
->dst
.error
) {
2868 rtm
->rtm_type
= RTN_BLACKHOLE
;
2871 rtm
->rtm_type
= RTN_PROHIBIT
;
2874 rtm
->rtm_type
= RTN_THROW
;
2877 rtm
->rtm_type
= RTN_UNREACHABLE
;
2881 else if (rt
->rt6i_flags
& RTF_LOCAL
)
2882 rtm
->rtm_type
= RTN_LOCAL
;
2883 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
2884 rtm
->rtm_type
= RTN_LOCAL
;
2886 rtm
->rtm_type
= RTN_UNICAST
;
2888 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2889 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2890 if (rt
->rt6i_flags
& RTF_DYNAMIC
)
2891 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2892 else if (rt
->rt6i_flags
& RTF_ADDRCONF
) {
2893 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ROUTEINFO
))
2894 rtm
->rtm_protocol
= RTPROT_RA
;
2896 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2899 if (rt
->rt6i_flags
& RTF_CACHE
)
2900 rtm
->rtm_flags
|= RTM_F_CLONED
;
2903 if (nla_put_in6_addr(skb
, RTA_DST
, dst
))
2904 goto nla_put_failure
;
2905 rtm
->rtm_dst_len
= 128;
2906 } else if (rtm
->rtm_dst_len
)
2907 if (nla_put_in6_addr(skb
, RTA_DST
, &rt
->rt6i_dst
.addr
))
2908 goto nla_put_failure
;
2909 #ifdef CONFIG_IPV6_SUBTREES
2911 if (nla_put_in6_addr(skb
, RTA_SRC
, src
))
2912 goto nla_put_failure
;
2913 rtm
->rtm_src_len
= 128;
2914 } else if (rtm
->rtm_src_len
&&
2915 nla_put_in6_addr(skb
, RTA_SRC
, &rt
->rt6i_src
.addr
))
2916 goto nla_put_failure
;
2919 #ifdef CONFIG_IPV6_MROUTE
2920 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2921 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2926 goto nla_put_failure
;
2928 if (err
== -EMSGSIZE
)
2929 goto nla_put_failure
;
2934 if (nla_put_u32(skb
, RTA_IIF
, iif
))
2935 goto nla_put_failure
;
2937 struct in6_addr saddr_buf
;
2938 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
2939 nla_put_in6_addr(skb
, RTA_PREFSRC
, &saddr_buf
))
2940 goto nla_put_failure
;
2943 if (rt
->rt6i_prefsrc
.plen
) {
2944 struct in6_addr saddr_buf
;
2945 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
2946 if (nla_put_in6_addr(skb
, RTA_PREFSRC
, &saddr_buf
))
2947 goto nla_put_failure
;
2950 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
2952 metrics
[RTAX_MTU
- 1] = rt
->rt6i_pmtu
;
2953 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
2954 goto nla_put_failure
;
2956 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2957 if (nla_put_in6_addr(skb
, RTA_GATEWAY
, &rt
->rt6i_gateway
) < 0)
2958 goto nla_put_failure
;
2962 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2963 goto nla_put_failure
;
2964 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
2965 goto nla_put_failure
;
2967 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
2969 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
2970 goto nla_put_failure
;
2972 if (nla_put_u8(skb
, RTA_PREF
, IPV6_EXTRACT_PREF(rt
->rt6i_flags
)))
2973 goto nla_put_failure
;
2975 lwtunnel_fill_encap(skb
, rt
->rt6i_lwtstate
);
2977 nlmsg_end(skb
, nlh
);
2981 nlmsg_cancel(skb
, nlh
);
2985 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2987 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2990 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2991 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2992 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2996 return rt6_fill_node(arg
->net
,
2997 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2998 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
2999 prefix
, 0, NLM_F_MULTI
);
3002 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
)
3004 struct net
*net
= sock_net(in_skb
->sk
);
3005 struct nlattr
*tb
[RTA_MAX
+1];
3006 struct rt6_info
*rt
;
3007 struct sk_buff
*skb
;
3010 int err
, iif
= 0, oif
= 0;
3012 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
3017 memset(&fl6
, 0, sizeof(fl6
));
3020 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
3023 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
3027 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
3030 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
3034 iif
= nla_get_u32(tb
[RTA_IIF
]);
3037 oif
= nla_get_u32(tb
[RTA_OIF
]);
3040 fl6
.flowi6_mark
= nla_get_u32(tb
[RTA_MARK
]);
3043 struct net_device
*dev
;
3046 dev
= __dev_get_by_index(net
, iif
);
3052 fl6
.flowi6_iif
= iif
;
3054 if (!ipv6_addr_any(&fl6
.saddr
))
3055 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
3057 rt
= (struct rt6_info
*)ip6_route_input_lookup(net
, dev
, &fl6
,
3060 fl6
.flowi6_oif
= oif
;
3062 rt
= (struct rt6_info
*)ip6_route_output(net
, NULL
, &fl6
);
3065 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
3072 /* Reserve room for dummy headers, this skb can pass
3073 through good chunk of routing engine.
3075 skb_reset_mac_header(skb
);
3076 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
3078 skb_dst_set(skb
, &rt
->dst
);
3080 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
3081 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
3082 nlh
->nlmsg_seq
, 0, 0, 0);
3088 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
3093 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
3095 struct sk_buff
*skb
;
3096 struct net
*net
= info
->nl_net
;
3101 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
3103 skb
= nlmsg_new(rt6_nlmsg_size(rt
), gfp_any());
3107 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
3108 event
, info
->portid
, seq
, 0, 0, 0);
3110 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3111 WARN_ON(err
== -EMSGSIZE
);
3115 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
3116 info
->nlh
, gfp_any());
3120 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
3123 static int ip6_route_dev_notify(struct notifier_block
*this,
3124 unsigned long event
, void *ptr
)
3126 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3127 struct net
*net
= dev_net(dev
);
3129 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
3130 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
3131 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
3132 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3133 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
3134 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
3135 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
3136 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
3147 #ifdef CONFIG_PROC_FS
3149 static const struct file_operations ipv6_route_proc_fops
= {
3150 .owner
= THIS_MODULE
,
3151 .open
= ipv6_route_open
,
3153 .llseek
= seq_lseek
,
3154 .release
= seq_release_net
,
3157 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
3159 struct net
*net
= (struct net
*)seq
->private;
3160 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
3161 net
->ipv6
.rt6_stats
->fib_nodes
,
3162 net
->ipv6
.rt6_stats
->fib_route_nodes
,
3163 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
3164 net
->ipv6
.rt6_stats
->fib_rt_entries
,
3165 net
->ipv6
.rt6_stats
->fib_rt_cache
,
3166 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
3167 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
3172 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
3174 return single_open_net(inode
, file
, rt6_stats_seq_show
);
3177 static const struct file_operations rt6_stats_seq_fops
= {
3178 .owner
= THIS_MODULE
,
3179 .open
= rt6_stats_seq_open
,
3181 .llseek
= seq_lseek
,
3182 .release
= single_release_net
,
3184 #endif /* CONFIG_PROC_FS */
3186 #ifdef CONFIG_SYSCTL
3189 int ipv6_sysctl_rtcache_flush(struct ctl_table
*ctl
, int write
,
3190 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
3197 net
= (struct net
*)ctl
->extra1
;
3198 delay
= net
->ipv6
.sysctl
.flush_delay
;
3199 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
3200 fib6_run_gc(delay
<= 0 ? 0 : (unsigned long)delay
, net
, delay
> 0);
3204 struct ctl_table ipv6_route_table_template
[] = {
3206 .procname
= "flush",
3207 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
3208 .maxlen
= sizeof(int),
3210 .proc_handler
= ipv6_sysctl_rtcache_flush
3213 .procname
= "gc_thresh",
3214 .data
= &ip6_dst_ops_template
.gc_thresh
,
3215 .maxlen
= sizeof(int),
3217 .proc_handler
= proc_dointvec
,
3220 .procname
= "max_size",
3221 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
3222 .maxlen
= sizeof(int),
3224 .proc_handler
= proc_dointvec
,
3227 .procname
= "gc_min_interval",
3228 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
3229 .maxlen
= sizeof(int),
3231 .proc_handler
= proc_dointvec_jiffies
,
3234 .procname
= "gc_timeout",
3235 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
3236 .maxlen
= sizeof(int),
3238 .proc_handler
= proc_dointvec_jiffies
,
3241 .procname
= "gc_interval",
3242 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
3243 .maxlen
= sizeof(int),
3245 .proc_handler
= proc_dointvec_jiffies
,
3248 .procname
= "gc_elasticity",
3249 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
3250 .maxlen
= sizeof(int),
3252 .proc_handler
= proc_dointvec
,
3255 .procname
= "mtu_expires",
3256 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
3257 .maxlen
= sizeof(int),
3259 .proc_handler
= proc_dointvec_jiffies
,
3262 .procname
= "min_adv_mss",
3263 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
3264 .maxlen
= sizeof(int),
3266 .proc_handler
= proc_dointvec
,
3269 .procname
= "gc_min_interval_ms",
3270 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
3271 .maxlen
= sizeof(int),
3273 .proc_handler
= proc_dointvec_ms_jiffies
,
3278 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
3280 struct ctl_table
*table
;
3282 table
= kmemdup(ipv6_route_table_template
,
3283 sizeof(ipv6_route_table_template
),
3287 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
3288 table
[0].extra1
= net
;
3289 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
3290 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
3291 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
3292 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
3293 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
3294 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
3295 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
3296 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
3297 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
3299 /* Don't export sysctls to unprivileged users */
3300 if (net
->user_ns
!= &init_user_ns
)
3301 table
[0].procname
= NULL
;
3308 static int __net_init
ip6_route_net_init(struct net
*net
)
3312 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
3313 sizeof(net
->ipv6
.ip6_dst_ops
));
3315 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
3316 goto out_ip6_dst_ops
;
3318 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
3319 sizeof(*net
->ipv6
.ip6_null_entry
),
3321 if (!net
->ipv6
.ip6_null_entry
)
3322 goto out_ip6_dst_entries
;
3323 net
->ipv6
.ip6_null_entry
->dst
.path
=
3324 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
3325 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3326 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
3327 ip6_template_metrics
, true);
3329 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3330 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
3331 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
3333 if (!net
->ipv6
.ip6_prohibit_entry
)
3334 goto out_ip6_null_entry
;
3335 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
3336 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
3337 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3338 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
3339 ip6_template_metrics
, true);
3341 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
3342 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
3344 if (!net
->ipv6
.ip6_blk_hole_entry
)
3345 goto out_ip6_prohibit_entry
;
3346 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
3347 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
3348 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3349 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
3350 ip6_template_metrics
, true);
3353 net
->ipv6
.sysctl
.flush_delay
= 0;
3354 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
3355 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
3356 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
3357 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
3358 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
3359 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
3360 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
3362 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
3368 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3369 out_ip6_prohibit_entry
:
3370 kfree(net
->ipv6
.ip6_prohibit_entry
);
3372 kfree(net
->ipv6
.ip6_null_entry
);
3374 out_ip6_dst_entries
:
3375 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3380 static void __net_exit
ip6_route_net_exit(struct net
*net
)
3382 kfree(net
->ipv6
.ip6_null_entry
);
3383 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3384 kfree(net
->ipv6
.ip6_prohibit_entry
);
3385 kfree(net
->ipv6
.ip6_blk_hole_entry
);
3387 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3390 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3392 #ifdef CONFIG_PROC_FS
3393 proc_create("ipv6_route", 0, net
->proc_net
, &ipv6_route_proc_fops
);
3394 proc_create("rt6_stats", S_IRUGO
, net
->proc_net
, &rt6_stats_seq_fops
);
3399 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
3401 #ifdef CONFIG_PROC_FS
3402 remove_proc_entry("ipv6_route", net
->proc_net
);
3403 remove_proc_entry("rt6_stats", net
->proc_net
);
3407 static struct pernet_operations ip6_route_net_ops
= {
3408 .init
= ip6_route_net_init
,
3409 .exit
= ip6_route_net_exit
,
3412 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
3414 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
3418 inet_peer_base_init(bp
);
3419 net
->ipv6
.peers
= bp
;
3423 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
3425 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
3427 net
->ipv6
.peers
= NULL
;
3428 inetpeer_invalidate_tree(bp
);
3432 static struct pernet_operations ipv6_inetpeer_ops
= {
3433 .init
= ipv6_inetpeer_init
,
3434 .exit
= ipv6_inetpeer_exit
,
3437 static struct pernet_operations ip6_route_net_late_ops
= {
3438 .init
= ip6_route_net_init_late
,
3439 .exit
= ip6_route_net_exit_late
,
3442 static struct notifier_block ip6_route_dev_notifier
= {
3443 .notifier_call
= ip6_route_dev_notify
,
3447 int __init
ip6_route_init(void)
3453 ip6_dst_ops_template
.kmem_cachep
=
3454 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
3455 SLAB_HWCACHE_ALIGN
, NULL
);
3456 if (!ip6_dst_ops_template
.kmem_cachep
)
3459 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
3461 goto out_kmem_cache
;
3463 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
3465 goto out_dst_entries
;
3467 ret
= register_pernet_subsys(&ip6_route_net_ops
);
3469 goto out_register_inetpeer
;
3471 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
3473 /* Registering of the loopback is done before this portion of code,
3474 * the loopback reference in rt6_info will not be taken, do it
3475 * manually for init_net */
3476 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
3477 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3478 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3479 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
3480 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3481 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
3482 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3486 goto out_register_subsys
;
3492 ret
= fib6_rules_init();
3496 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
3498 goto fib6_rules_init
;
3501 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
3502 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
3503 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
3504 goto out_register_late_subsys
;
3506 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
3508 goto out_register_late_subsys
;
3510 for_each_possible_cpu(cpu
) {
3511 struct uncached_list
*ul
= per_cpu_ptr(&rt6_uncached_list
, cpu
);
3513 INIT_LIST_HEAD(&ul
->head
);
3514 spin_lock_init(&ul
->lock
);
3520 out_register_late_subsys
:
3521 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3523 fib6_rules_cleanup();
3528 out_register_subsys
:
3529 unregister_pernet_subsys(&ip6_route_net_ops
);
3530 out_register_inetpeer
:
3531 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3533 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3535 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
3539 void ip6_route_cleanup(void)
3541 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
3542 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3543 fib6_rules_cleanup();
3546 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3547 unregister_pernet_subsys(&ip6_route_net_ops
);
3548 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3549 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);