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>
57 #include <net/dst_metadata.h>
59 #include <net/netevent.h>
60 #include <net/netlink.h>
61 #include <net/nexthop.h>
62 #include <net/lwtunnel.h>
63 #include <net/ip_tunnels.h>
64 #include <net/l3mdev.h>
65 #include <trace/events/fib6.h>
67 #include <linux/uaccess.h>
70 #include <linux/sysctl.h>
74 RT6_NUD_FAIL_HARD
= -3,
75 RT6_NUD_FAIL_PROBE
= -2,
76 RT6_NUD_FAIL_DO_RR
= -1,
80 static void ip6_rt_copy_init(struct rt6_info
*rt
, struct rt6_info
*ort
);
81 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
82 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
83 static unsigned int ip6_mtu(const struct dst_entry
*dst
);
84 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
85 static void ip6_dst_destroy(struct dst_entry
*);
86 static void ip6_dst_ifdown(struct dst_entry
*,
87 struct net_device
*dev
, int how
);
88 static int ip6_dst_gc(struct dst_ops
*ops
);
90 static int ip6_pkt_discard(struct sk_buff
*skb
);
91 static int ip6_pkt_discard_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
92 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
93 static int ip6_pkt_prohibit_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
);
94 static void ip6_link_failure(struct sk_buff
*skb
);
95 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
96 struct sk_buff
*skb
, u32 mtu
);
97 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
99 static void rt6_dst_from_metrics_check(struct rt6_info
*rt
);
100 static int rt6_score_route(struct rt6_info
*rt
, int oif
, int strict
);
101 static size_t rt6_nlmsg_size(struct rt6_info
*rt
);
102 static int rt6_fill_node(struct net
*net
,
103 struct sk_buff
*skb
, struct rt6_info
*rt
,
104 struct in6_addr
*dst
, struct in6_addr
*src
,
105 int iif
, int type
, u32 portid
, u32 seq
,
108 #ifdef CONFIG_IPV6_ROUTE_INFO
109 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
110 const struct in6_addr
*prefix
, int prefixlen
,
111 const struct in6_addr
*gwaddr
,
112 struct net_device
*dev
,
114 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
115 const struct in6_addr
*prefix
, int prefixlen
,
116 const struct in6_addr
*gwaddr
,
117 struct net_device
*dev
);
120 struct uncached_list
{
122 struct list_head head
;
125 static DEFINE_PER_CPU_ALIGNED(struct uncached_list
, rt6_uncached_list
);
127 void rt6_uncached_list_add(struct rt6_info
*rt
)
129 struct uncached_list
*ul
= raw_cpu_ptr(&rt6_uncached_list
);
131 rt
->rt6i_uncached_list
= ul
;
133 spin_lock_bh(&ul
->lock
);
134 list_add_tail(&rt
->rt6i_uncached
, &ul
->head
);
135 spin_unlock_bh(&ul
->lock
);
138 void rt6_uncached_list_del(struct rt6_info
*rt
)
140 if (!list_empty(&rt
->rt6i_uncached
)) {
141 struct uncached_list
*ul
= rt
->rt6i_uncached_list
;
143 spin_lock_bh(&ul
->lock
);
144 list_del(&rt
->rt6i_uncached
);
145 spin_unlock_bh(&ul
->lock
);
149 static void rt6_uncached_list_flush_dev(struct net
*net
, struct net_device
*dev
)
151 struct net_device
*loopback_dev
= net
->loopback_dev
;
154 if (dev
== loopback_dev
)
157 for_each_possible_cpu(cpu
) {
158 struct uncached_list
*ul
= per_cpu_ptr(&rt6_uncached_list
, cpu
);
161 spin_lock_bh(&ul
->lock
);
162 list_for_each_entry(rt
, &ul
->head
, rt6i_uncached
) {
163 struct inet6_dev
*rt_idev
= rt
->rt6i_idev
;
164 struct net_device
*rt_dev
= rt
->dst
.dev
;
166 if (rt_idev
->dev
== dev
) {
167 rt
->rt6i_idev
= in6_dev_get(loopback_dev
);
168 in6_dev_put(rt_idev
);
172 rt
->dst
.dev
= loopback_dev
;
173 dev_hold(rt
->dst
.dev
);
177 spin_unlock_bh(&ul
->lock
);
181 static u32
*rt6_pcpu_cow_metrics(struct rt6_info
*rt
)
183 return dst_metrics_write_ptr(rt
->dst
.from
);
186 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
188 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
190 if (rt
->rt6i_flags
& RTF_PCPU
)
191 return rt6_pcpu_cow_metrics(rt
);
192 else if (rt
->rt6i_flags
& RTF_CACHE
)
195 return dst_cow_metrics_generic(dst
, old
);
198 static inline const void *choose_neigh_daddr(struct rt6_info
*rt
,
202 struct in6_addr
*p
= &rt
->rt6i_gateway
;
204 if (!ipv6_addr_any(p
))
205 return (const void *) p
;
207 return &ipv6_hdr(skb
)->daddr
;
211 static struct neighbour
*ip6_neigh_lookup(const struct dst_entry
*dst
,
215 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
218 daddr
= choose_neigh_daddr(rt
, skb
, daddr
);
219 n
= __ipv6_neigh_lookup(dst
->dev
, daddr
);
222 return neigh_create(&nd_tbl
, daddr
, dst
->dev
);
225 static void ip6_confirm_neigh(const struct dst_entry
*dst
, const void *daddr
)
227 struct net_device
*dev
= dst
->dev
;
228 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
230 daddr
= choose_neigh_daddr(rt
, NULL
, daddr
);
233 if (dev
->flags
& (IFF_NOARP
| IFF_LOOPBACK
))
235 if (ipv6_addr_is_multicast((const struct in6_addr
*)daddr
))
237 __ipv6_confirm_neigh(dev
, daddr
);
240 static struct dst_ops ip6_dst_ops_template
= {
244 .check
= ip6_dst_check
,
245 .default_advmss
= ip6_default_advmss
,
247 .cow_metrics
= ipv6_cow_metrics
,
248 .destroy
= ip6_dst_destroy
,
249 .ifdown
= ip6_dst_ifdown
,
250 .negative_advice
= ip6_negative_advice
,
251 .link_failure
= ip6_link_failure
,
252 .update_pmtu
= ip6_rt_update_pmtu
,
253 .redirect
= rt6_do_redirect
,
254 .local_out
= __ip6_local_out
,
255 .neigh_lookup
= ip6_neigh_lookup
,
256 .confirm_neigh
= ip6_confirm_neigh
,
259 static unsigned int ip6_blackhole_mtu(const struct dst_entry
*dst
)
261 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
263 return mtu
? : dst
->dev
->mtu
;
266 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
267 struct sk_buff
*skb
, u32 mtu
)
271 static void ip6_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
276 static struct dst_ops ip6_dst_blackhole_ops
= {
278 .destroy
= ip6_dst_destroy
,
279 .check
= ip6_dst_check
,
280 .mtu
= ip6_blackhole_mtu
,
281 .default_advmss
= ip6_default_advmss
,
282 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
283 .redirect
= ip6_rt_blackhole_redirect
,
284 .cow_metrics
= dst_cow_metrics_generic
,
285 .neigh_lookup
= ip6_neigh_lookup
,
288 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
289 [RTAX_HOPLIMIT
- 1] = 0,
292 static const struct rt6_info ip6_null_entry_template
= {
294 .__refcnt
= ATOMIC_INIT(1),
296 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
297 .error
= -ENETUNREACH
,
298 .input
= ip6_pkt_discard
,
299 .output
= ip6_pkt_discard_out
,
301 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
302 .rt6i_protocol
= RTPROT_KERNEL
,
303 .rt6i_metric
= ~(u32
) 0,
304 .rt6i_ref
= ATOMIC_INIT(1),
307 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
309 static const struct rt6_info ip6_prohibit_entry_template
= {
311 .__refcnt
= ATOMIC_INIT(1),
313 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
315 .input
= ip6_pkt_prohibit
,
316 .output
= ip6_pkt_prohibit_out
,
318 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
319 .rt6i_protocol
= RTPROT_KERNEL
,
320 .rt6i_metric
= ~(u32
) 0,
321 .rt6i_ref
= ATOMIC_INIT(1),
324 static const struct rt6_info ip6_blk_hole_entry_template
= {
326 .__refcnt
= ATOMIC_INIT(1),
328 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
330 .input
= dst_discard
,
331 .output
= dst_discard_out
,
333 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
334 .rt6i_protocol
= RTPROT_KERNEL
,
335 .rt6i_metric
= ~(u32
) 0,
336 .rt6i_ref
= ATOMIC_INIT(1),
341 static void rt6_info_init(struct rt6_info
*rt
)
343 struct dst_entry
*dst
= &rt
->dst
;
345 memset(dst
+ 1, 0, sizeof(*rt
) - sizeof(*dst
));
346 INIT_LIST_HEAD(&rt
->rt6i_siblings
);
347 INIT_LIST_HEAD(&rt
->rt6i_uncached
);
350 /* allocate dst with ip6_dst_ops */
351 static struct rt6_info
*__ip6_dst_alloc(struct net
*net
,
352 struct net_device
*dev
,
355 struct rt6_info
*rt
= dst_alloc(&net
->ipv6
.ip6_dst_ops
, dev
,
356 1, DST_OBSOLETE_FORCE_CHK
, flags
);
364 struct rt6_info
*ip6_dst_alloc(struct net
*net
,
365 struct net_device
*dev
,
368 struct rt6_info
*rt
= __ip6_dst_alloc(net
, dev
, flags
);
371 rt
->rt6i_pcpu
= alloc_percpu_gfp(struct rt6_info
*, GFP_ATOMIC
);
375 for_each_possible_cpu(cpu
) {
378 p
= per_cpu_ptr(rt
->rt6i_pcpu
, cpu
);
379 /* no one shares rt */
383 dst_release_immediate(&rt
->dst
);
390 EXPORT_SYMBOL(ip6_dst_alloc
);
392 static void ip6_dst_destroy(struct dst_entry
*dst
)
394 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
395 struct dst_entry
*from
= dst
->from
;
396 struct inet6_dev
*idev
;
398 dst_destroy_metrics_generic(dst
);
399 free_percpu(rt
->rt6i_pcpu
);
400 rt6_uncached_list_del(rt
);
402 idev
= rt
->rt6i_idev
;
404 rt
->rt6i_idev
= NULL
;
412 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
415 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
416 struct inet6_dev
*idev
= rt
->rt6i_idev
;
417 struct net_device
*loopback_dev
=
418 dev_net(dev
)->loopback_dev
;
420 if (idev
&& idev
->dev
!= loopback_dev
) {
421 struct inet6_dev
*loopback_idev
= in6_dev_get(loopback_dev
);
423 rt
->rt6i_idev
= loopback_idev
;
429 static bool __rt6_check_expired(const struct rt6_info
*rt
)
431 if (rt
->rt6i_flags
& RTF_EXPIRES
)
432 return time_after(jiffies
, rt
->dst
.expires
);
437 static bool rt6_check_expired(const struct rt6_info
*rt
)
439 if (rt
->rt6i_flags
& RTF_EXPIRES
) {
440 if (time_after(jiffies
, rt
->dst
.expires
))
442 } else if (rt
->dst
.from
) {
443 return rt
->dst
.obsolete
!= DST_OBSOLETE_FORCE_CHK
||
444 rt6_check_expired((struct rt6_info
*)rt
->dst
.from
);
449 /* Multipath route selection:
450 * Hash based function using packet header and flowlabel.
451 * Adapted from fib_info_hashfn()
453 static int rt6_info_hash_nhsfn(unsigned int candidate_count
,
454 const struct flowi6
*fl6
)
456 return get_hash_from_flowi6(fl6
) % candidate_count
;
459 static struct rt6_info
*rt6_multipath_select(struct rt6_info
*match
,
460 struct flowi6
*fl6
, int oif
,
463 struct rt6_info
*sibling
, *next_sibling
;
466 route_choosen
= rt6_info_hash_nhsfn(match
->rt6i_nsiblings
+ 1, fl6
);
467 /* Don't change the route, if route_choosen == 0
468 * (siblings does not include ourself)
471 list_for_each_entry_safe(sibling
, next_sibling
,
472 &match
->rt6i_siblings
, rt6i_siblings
) {
474 if (route_choosen
== 0) {
475 struct inet6_dev
*idev
= sibling
->rt6i_idev
;
477 if (!netif_carrier_ok(sibling
->dst
.dev
) &&
478 idev
->cnf
.ignore_routes_with_linkdown
)
480 if (rt6_score_route(sibling
, oif
, strict
) < 0)
490 * Route lookup. Any table->tb6_lock is implied.
493 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
495 const struct in6_addr
*saddr
,
499 struct rt6_info
*local
= NULL
;
500 struct rt6_info
*sprt
;
502 if (!oif
&& ipv6_addr_any(saddr
))
505 for (sprt
= rt
; sprt
; sprt
= sprt
->dst
.rt6_next
) {
506 struct net_device
*dev
= sprt
->dst
.dev
;
509 if (dev
->ifindex
== oif
)
511 if (dev
->flags
& IFF_LOOPBACK
) {
512 if (!sprt
->rt6i_idev
||
513 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
514 if (flags
& RT6_LOOKUP_F_IFACE
)
517 local
->rt6i_idev
->dev
->ifindex
== oif
)
523 if (ipv6_chk_addr(net
, saddr
, dev
,
524 flags
& RT6_LOOKUP_F_IFACE
))
533 if (flags
& RT6_LOOKUP_F_IFACE
)
534 return net
->ipv6
.ip6_null_entry
;
540 #ifdef CONFIG_IPV6_ROUTER_PREF
541 struct __rt6_probe_work
{
542 struct work_struct work
;
543 struct in6_addr target
;
544 struct net_device
*dev
;
547 static void rt6_probe_deferred(struct work_struct
*w
)
549 struct in6_addr mcaddr
;
550 struct __rt6_probe_work
*work
=
551 container_of(w
, struct __rt6_probe_work
, work
);
553 addrconf_addr_solict_mult(&work
->target
, &mcaddr
);
554 ndisc_send_ns(work
->dev
, &work
->target
, &mcaddr
, NULL
, 0);
559 static void rt6_probe(struct rt6_info
*rt
)
561 struct __rt6_probe_work
*work
;
562 struct neighbour
*neigh
;
564 * Okay, this does not seem to be appropriate
565 * for now, however, we need to check if it
566 * is really so; aka Router Reachability Probing.
568 * Router Reachability Probe MUST be rate-limited
569 * to no more than one per minute.
571 if (!rt
|| !(rt
->rt6i_flags
& RTF_GATEWAY
))
574 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
576 if (neigh
->nud_state
& NUD_VALID
)
580 write_lock(&neigh
->lock
);
581 if (!(neigh
->nud_state
& NUD_VALID
) &&
584 rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
585 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
587 __neigh_set_probe_once(neigh
);
589 write_unlock(&neigh
->lock
);
591 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
595 INIT_WORK(&work
->work
, rt6_probe_deferred
);
596 work
->target
= rt
->rt6i_gateway
;
597 dev_hold(rt
->dst
.dev
);
598 work
->dev
= rt
->dst
.dev
;
599 schedule_work(&work
->work
);
603 rcu_read_unlock_bh();
606 static inline void rt6_probe(struct rt6_info
*rt
)
612 * Default Router Selection (RFC 2461 6.3.6)
614 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
616 struct net_device
*dev
= rt
->dst
.dev
;
617 if (!oif
|| dev
->ifindex
== oif
)
619 if ((dev
->flags
& IFF_LOOPBACK
) &&
620 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
625 static inline enum rt6_nud_state
rt6_check_neigh(struct rt6_info
*rt
)
627 struct neighbour
*neigh
;
628 enum rt6_nud_state ret
= RT6_NUD_FAIL_HARD
;
630 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
631 !(rt
->rt6i_flags
& RTF_GATEWAY
))
632 return RT6_NUD_SUCCEED
;
635 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
637 read_lock(&neigh
->lock
);
638 if (neigh
->nud_state
& NUD_VALID
)
639 ret
= RT6_NUD_SUCCEED
;
640 #ifdef CONFIG_IPV6_ROUTER_PREF
641 else if (!(neigh
->nud_state
& NUD_FAILED
))
642 ret
= RT6_NUD_SUCCEED
;
644 ret
= RT6_NUD_FAIL_PROBE
;
646 read_unlock(&neigh
->lock
);
648 ret
= IS_ENABLED(CONFIG_IPV6_ROUTER_PREF
) ?
649 RT6_NUD_SUCCEED
: RT6_NUD_FAIL_DO_RR
;
651 rcu_read_unlock_bh();
656 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
661 m
= rt6_check_dev(rt
, oif
);
662 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
663 return RT6_NUD_FAIL_HARD
;
664 #ifdef CONFIG_IPV6_ROUTER_PREF
665 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
667 if (strict
& RT6_LOOKUP_F_REACHABLE
) {
668 int n
= rt6_check_neigh(rt
);
675 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
676 int *mpri
, struct rt6_info
*match
,
680 bool match_do_rr
= false;
681 struct inet6_dev
*idev
= rt
->rt6i_idev
;
682 struct net_device
*dev
= rt
->dst
.dev
;
684 if (dev
&& !netif_carrier_ok(dev
) &&
685 idev
->cnf
.ignore_routes_with_linkdown
&&
686 !(strict
& RT6_LOOKUP_F_IGNORE_LINKSTATE
))
689 if (rt6_check_expired(rt
))
692 m
= rt6_score_route(rt
, oif
, strict
);
693 if (m
== RT6_NUD_FAIL_DO_RR
) {
695 m
= 0; /* lowest valid score */
696 } else if (m
== RT6_NUD_FAIL_HARD
) {
700 if (strict
& RT6_LOOKUP_F_REACHABLE
)
703 /* note that m can be RT6_NUD_FAIL_PROBE at this point */
705 *do_rr
= match_do_rr
;
713 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
714 struct rt6_info
*rr_head
,
715 u32 metric
, int oif
, int strict
,
718 struct rt6_info
*rt
, *match
, *cont
;
723 for (rt
= rr_head
; rt
; rt
= rt
->dst
.rt6_next
) {
724 if (rt
->rt6i_metric
!= metric
) {
729 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
732 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
; rt
= rt
->dst
.rt6_next
) {
733 if (rt
->rt6i_metric
!= metric
) {
738 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
744 for (rt
= cont
; rt
; rt
= rt
->dst
.rt6_next
)
745 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
750 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
752 struct rt6_info
*match
, *rt0
;
758 fn
->rr_ptr
= rt0
= fn
->leaf
;
760 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
,
764 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
766 /* no entries matched; do round-robin */
767 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
774 net
= dev_net(rt0
->dst
.dev
);
775 return match
? match
: net
->ipv6
.ip6_null_entry
;
778 static bool rt6_is_gw_or_nonexthop(const struct rt6_info
*rt
)
780 return (rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
));
783 #ifdef CONFIG_IPV6_ROUTE_INFO
784 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
785 const struct in6_addr
*gwaddr
)
787 struct net
*net
= dev_net(dev
);
788 struct route_info
*rinfo
= (struct route_info
*) opt
;
789 struct in6_addr prefix_buf
, *prefix
;
791 unsigned long lifetime
;
794 if (len
< sizeof(struct route_info
)) {
798 /* Sanity check for prefix_len and length */
799 if (rinfo
->length
> 3) {
801 } else if (rinfo
->prefix_len
> 128) {
803 } else if (rinfo
->prefix_len
> 64) {
804 if (rinfo
->length
< 2) {
807 } else if (rinfo
->prefix_len
> 0) {
808 if (rinfo
->length
< 1) {
813 pref
= rinfo
->route_pref
;
814 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
817 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
819 if (rinfo
->length
== 3)
820 prefix
= (struct in6_addr
*)rinfo
->prefix
;
822 /* this function is safe */
823 ipv6_addr_prefix(&prefix_buf
,
824 (struct in6_addr
*)rinfo
->prefix
,
826 prefix
= &prefix_buf
;
829 if (rinfo
->prefix_len
== 0)
830 rt
= rt6_get_dflt_router(gwaddr
, dev
);
832 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
,
835 if (rt
&& !lifetime
) {
841 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
,
844 rt
->rt6i_flags
= RTF_ROUTEINFO
|
845 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
848 if (!addrconf_finite_timeout(lifetime
))
849 rt6_clean_expires(rt
);
851 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
859 static struct fib6_node
* fib6_backtrack(struct fib6_node
*fn
,
860 struct in6_addr
*saddr
)
862 struct fib6_node
*pn
;
864 if (fn
->fn_flags
& RTN_TL_ROOT
)
867 if (FIB6_SUBTREE(pn
) && FIB6_SUBTREE(pn
) != fn
)
868 fn
= fib6_lookup(FIB6_SUBTREE(pn
), NULL
, saddr
);
871 if (fn
->fn_flags
& RTN_RTINFO
)
876 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
877 struct fib6_table
*table
,
878 struct flowi6
*fl6
, int flags
)
880 struct fib6_node
*fn
;
883 read_lock_bh(&table
->tb6_lock
);
884 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
887 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
, fl6
->flowi6_oif
, flags
);
888 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
889 rt
= rt6_multipath_select(rt
, fl6
, fl6
->flowi6_oif
, flags
);
890 if (rt
== net
->ipv6
.ip6_null_entry
) {
891 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
895 dst_use(&rt
->dst
, jiffies
);
896 read_unlock_bh(&table
->tb6_lock
);
898 trace_fib6_table_lookup(net
, rt
, table
->tb6_id
, fl6
);
904 struct dst_entry
*ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
907 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
909 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
911 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
912 const struct in6_addr
*saddr
, int oif
, int strict
)
914 struct flowi6 fl6
= {
918 struct dst_entry
*dst
;
919 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
922 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
923 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
926 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
928 return (struct rt6_info
*) dst
;
934 EXPORT_SYMBOL(rt6_lookup
);
936 /* ip6_ins_rt is called with FREE table->tb6_lock.
937 * It takes new route entry, the addition fails by any reason the
939 * Caller must hold dst before calling it.
942 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
,
943 struct mx6_config
*mxc
,
944 struct netlink_ext_ack
*extack
)
947 struct fib6_table
*table
;
949 table
= rt
->rt6i_table
;
950 write_lock_bh(&table
->tb6_lock
);
951 err
= fib6_add(&table
->tb6_root
, rt
, info
, mxc
, extack
);
952 write_unlock_bh(&table
->tb6_lock
);
957 int ip6_ins_rt(struct rt6_info
*rt
)
959 struct nl_info info
= { .nl_net
= dev_net(rt
->dst
.dev
), };
960 struct mx6_config mxc
= { .mx
= NULL
, };
962 /* Hold dst to account for the reference from the fib6 tree */
964 return __ip6_ins_rt(rt
, &info
, &mxc
, NULL
);
967 static struct rt6_info
*ip6_rt_cache_alloc(struct rt6_info
*ort
,
968 const struct in6_addr
*daddr
,
969 const struct in6_addr
*saddr
)
977 if (ort
->rt6i_flags
& (RTF_CACHE
| RTF_PCPU
))
978 ort
= (struct rt6_info
*)ort
->dst
.from
;
980 rt
= __ip6_dst_alloc(dev_net(ort
->dst
.dev
), ort
->dst
.dev
, 0);
985 ip6_rt_copy_init(rt
, ort
);
986 rt
->rt6i_flags
|= RTF_CACHE
;
988 rt
->dst
.flags
|= DST_HOST
;
989 rt
->rt6i_dst
.addr
= *daddr
;
990 rt
->rt6i_dst
.plen
= 128;
992 if (!rt6_is_gw_or_nonexthop(ort
)) {
993 if (ort
->rt6i_dst
.plen
!= 128 &&
994 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
995 rt
->rt6i_flags
|= RTF_ANYCAST
;
996 #ifdef CONFIG_IPV6_SUBTREES
997 if (rt
->rt6i_src
.plen
&& saddr
) {
998 rt
->rt6i_src
.addr
= *saddr
;
999 rt
->rt6i_src
.plen
= 128;
1007 static struct rt6_info
*ip6_rt_pcpu_alloc(struct rt6_info
*rt
)
1009 struct rt6_info
*pcpu_rt
;
1011 pcpu_rt
= __ip6_dst_alloc(dev_net(rt
->dst
.dev
),
1012 rt
->dst
.dev
, rt
->dst
.flags
);
1016 ip6_rt_copy_init(pcpu_rt
, rt
);
1017 pcpu_rt
->rt6i_protocol
= rt
->rt6i_protocol
;
1018 pcpu_rt
->rt6i_flags
|= RTF_PCPU
;
1022 /* It should be called with read_lock_bh(&tb6_lock) acquired */
1023 static struct rt6_info
*rt6_get_pcpu_route(struct rt6_info
*rt
)
1025 struct rt6_info
*pcpu_rt
, **p
;
1027 p
= this_cpu_ptr(rt
->rt6i_pcpu
);
1031 dst_hold(&pcpu_rt
->dst
);
1032 rt6_dst_from_metrics_check(pcpu_rt
);
1037 static struct rt6_info
*rt6_make_pcpu_route(struct rt6_info
*rt
)
1039 struct fib6_table
*table
= rt
->rt6i_table
;
1040 struct rt6_info
*pcpu_rt
, *prev
, **p
;
1042 pcpu_rt
= ip6_rt_pcpu_alloc(rt
);
1044 struct net
*net
= dev_net(rt
->dst
.dev
);
1046 dst_hold(&net
->ipv6
.ip6_null_entry
->dst
);
1047 return net
->ipv6
.ip6_null_entry
;
1050 read_lock_bh(&table
->tb6_lock
);
1051 if (rt
->rt6i_pcpu
) {
1052 p
= this_cpu_ptr(rt
->rt6i_pcpu
);
1053 prev
= cmpxchg(p
, NULL
, pcpu_rt
);
1055 /* If someone did it before us, return prev instead */
1056 dst_release_immediate(&pcpu_rt
->dst
);
1060 /* rt has been removed from the fib6 tree
1061 * before we have a chance to acquire the read_lock.
1062 * In this case, don't brother to create a pcpu rt
1063 * since rt is going away anyway. The next
1064 * dst_check() will trigger a re-lookup.
1066 dst_release_immediate(&pcpu_rt
->dst
);
1069 dst_hold(&pcpu_rt
->dst
);
1070 rt6_dst_from_metrics_check(pcpu_rt
);
1071 read_unlock_bh(&table
->tb6_lock
);
1075 struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
,
1076 int oif
, struct flowi6
*fl6
, int flags
)
1078 struct fib6_node
*fn
, *saved_fn
;
1079 struct rt6_info
*rt
;
1082 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
1083 strict
|= flags
& RT6_LOOKUP_F_IGNORE_LINKSTATE
;
1084 if (net
->ipv6
.devconf_all
->forwarding
== 0)
1085 strict
|= RT6_LOOKUP_F_REACHABLE
;
1087 read_lock_bh(&table
->tb6_lock
);
1089 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1092 if (fl6
->flowi6_flags
& FLOWI_FLAG_SKIP_NH_OIF
)
1096 rt
= rt6_select(fn
, oif
, strict
);
1097 if (rt
->rt6i_nsiblings
)
1098 rt
= rt6_multipath_select(rt
, fl6
, oif
, strict
);
1099 if (rt
== net
->ipv6
.ip6_null_entry
) {
1100 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
1102 goto redo_rt6_select
;
1103 else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
1104 /* also consider unreachable route */
1105 strict
&= ~RT6_LOOKUP_F_REACHABLE
;
1107 goto redo_rt6_select
;
1112 if (rt
== net
->ipv6
.ip6_null_entry
|| (rt
->rt6i_flags
& RTF_CACHE
)) {
1113 dst_use(&rt
->dst
, jiffies
);
1114 read_unlock_bh(&table
->tb6_lock
);
1116 rt6_dst_from_metrics_check(rt
);
1118 trace_fib6_table_lookup(net
, rt
, table
->tb6_id
, fl6
);
1120 } else if (unlikely((fl6
->flowi6_flags
& FLOWI_FLAG_KNOWN_NH
) &&
1121 !(rt
->rt6i_flags
& RTF_GATEWAY
))) {
1122 /* Create a RTF_CACHE clone which will not be
1123 * owned by the fib6 tree. It is for the special case where
1124 * the daddr in the skb during the neighbor look-up is different
1125 * from the fl6->daddr used to look-up route here.
1128 struct rt6_info
*uncached_rt
;
1130 dst_use(&rt
->dst
, jiffies
);
1131 read_unlock_bh(&table
->tb6_lock
);
1133 uncached_rt
= ip6_rt_cache_alloc(rt
, &fl6
->daddr
, NULL
);
1134 dst_release(&rt
->dst
);
1137 /* Uncached_rt's refcnt is taken during ip6_rt_cache_alloc()
1138 * No need for another dst_hold()
1140 rt6_uncached_list_add(uncached_rt
);
1142 uncached_rt
= net
->ipv6
.ip6_null_entry
;
1143 dst_hold(&uncached_rt
->dst
);
1146 trace_fib6_table_lookup(net
, uncached_rt
, table
->tb6_id
, fl6
);
1150 /* Get a percpu copy */
1152 struct rt6_info
*pcpu_rt
;
1154 rt
->dst
.lastuse
= jiffies
;
1156 pcpu_rt
= rt6_get_pcpu_route(rt
);
1159 read_unlock_bh(&table
->tb6_lock
);
1161 /* We have to do the read_unlock first
1162 * because rt6_make_pcpu_route() may trigger
1163 * ip6_dst_gc() which will take the write_lock.
1166 read_unlock_bh(&table
->tb6_lock
);
1167 pcpu_rt
= rt6_make_pcpu_route(rt
);
1168 dst_release(&rt
->dst
);
1171 trace_fib6_table_lookup(net
, pcpu_rt
, table
->tb6_id
, fl6
);
1176 EXPORT_SYMBOL_GPL(ip6_pol_route
);
1178 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
1179 struct flowi6
*fl6
, int flags
)
1181 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
1184 struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
1185 struct net_device
*dev
,
1186 struct flowi6
*fl6
, int flags
)
1188 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
1189 flags
|= RT6_LOOKUP_F_IFACE
;
1191 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
1193 EXPORT_SYMBOL_GPL(ip6_route_input_lookup
);
1195 void ip6_route_input(struct sk_buff
*skb
)
1197 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1198 struct net
*net
= dev_net(skb
->dev
);
1199 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1200 struct ip_tunnel_info
*tun_info
;
1201 struct flowi6 fl6
= {
1202 .flowi6_iif
= skb
->dev
->ifindex
,
1203 .daddr
= iph
->daddr
,
1204 .saddr
= iph
->saddr
,
1205 .flowlabel
= ip6_flowinfo(iph
),
1206 .flowi6_mark
= skb
->mark
,
1207 .flowi6_proto
= iph
->nexthdr
,
1210 tun_info
= skb_tunnel_info(skb
);
1211 if (tun_info
&& !(tun_info
->mode
& IP_TUNNEL_INFO_TX
))
1212 fl6
.flowi6_tun_key
.tun_id
= tun_info
->key
.tun_id
;
1214 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
1217 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
1218 struct flowi6
*fl6
, int flags
)
1220 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1223 struct dst_entry
*ip6_route_output_flags(struct net
*net
, const struct sock
*sk
,
1224 struct flowi6
*fl6
, int flags
)
1228 if (rt6_need_strict(&fl6
->daddr
)) {
1229 struct dst_entry
*dst
;
1231 dst
= l3mdev_link_scope_lookup(net
, fl6
);
1236 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1238 any_src
= ipv6_addr_any(&fl6
->saddr
);
1239 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
) ||
1240 (fl6
->flowi6_oif
&& any_src
))
1241 flags
|= RT6_LOOKUP_F_IFACE
;
1244 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1246 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1248 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1250 EXPORT_SYMBOL_GPL(ip6_route_output_flags
);
1252 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1254 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1255 struct net_device
*loopback_dev
= net
->loopback_dev
;
1256 struct dst_entry
*new = NULL
;
1258 rt
= dst_alloc(&ip6_dst_blackhole_ops
, loopback_dev
, 1,
1259 DST_OBSOLETE_DEAD
, 0);
1265 new->input
= dst_discard
;
1266 new->output
= dst_discard_out
;
1268 dst_copy_metrics(new, &ort
->dst
);
1270 rt
->rt6i_idev
= in6_dev_get(loopback_dev
);
1271 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1272 rt
->rt6i_flags
= ort
->rt6i_flags
& ~RTF_PCPU
;
1273 rt
->rt6i_metric
= 0;
1275 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1276 #ifdef CONFIG_IPV6_SUBTREES
1277 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1281 dst_release(dst_orig
);
1282 return new ? new : ERR_PTR(-ENOMEM
);
1286 * Destination cache support functions
1289 static void rt6_dst_from_metrics_check(struct rt6_info
*rt
)
1292 dst_metrics_ptr(&rt
->dst
) != dst_metrics_ptr(rt
->dst
.from
))
1293 dst_init_metrics(&rt
->dst
, dst_metrics_ptr(rt
->dst
.from
), true);
1296 static struct dst_entry
*rt6_check(struct rt6_info
*rt
, u32 cookie
)
1300 if (!rt6_get_cookie_safe(rt
, &rt_cookie
) || rt_cookie
!= cookie
)
1303 if (rt6_check_expired(rt
))
1309 static struct dst_entry
*rt6_dst_from_check(struct rt6_info
*rt
, u32 cookie
)
1311 if (!__rt6_check_expired(rt
) &&
1312 rt
->dst
.obsolete
== DST_OBSOLETE_FORCE_CHK
&&
1313 rt6_check((struct rt6_info
*)(rt
->dst
.from
), cookie
))
1319 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1321 struct rt6_info
*rt
;
1323 rt
= (struct rt6_info
*) dst
;
1325 /* All IPV6 dsts are created with ->obsolete set to the value
1326 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1327 * into this function always.
1330 rt6_dst_from_metrics_check(rt
);
1332 if (rt
->rt6i_flags
& RTF_PCPU
||
1333 (unlikely(!list_empty(&rt
->rt6i_uncached
)) && rt
->dst
.from
))
1334 return rt6_dst_from_check(rt
, cookie
);
1336 return rt6_check(rt
, cookie
);
1339 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1341 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1344 if (rt
->rt6i_flags
& RTF_CACHE
) {
1345 if (rt6_check_expired(rt
)) {
1357 static void ip6_link_failure(struct sk_buff
*skb
)
1359 struct rt6_info
*rt
;
1361 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1363 rt
= (struct rt6_info
*) skb_dst(skb
);
1365 if (rt
->rt6i_flags
& RTF_CACHE
) {
1366 if (dst_hold_safe(&rt
->dst
))
1369 struct fib6_node
*fn
;
1372 fn
= rcu_dereference(rt
->rt6i_node
);
1373 if (fn
&& (rt
->rt6i_flags
& RTF_DEFAULT
))
1380 static void rt6_do_update_pmtu(struct rt6_info
*rt
, u32 mtu
)
1382 struct net
*net
= dev_net(rt
->dst
.dev
);
1384 rt
->rt6i_flags
|= RTF_MODIFIED
;
1385 rt
->rt6i_pmtu
= mtu
;
1386 rt6_update_expires(rt
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1389 static bool rt6_cache_allowed_for_pmtu(const struct rt6_info
*rt
)
1391 return !(rt
->rt6i_flags
& RTF_CACHE
) &&
1392 (rt
->rt6i_flags
& RTF_PCPU
||
1393 rcu_access_pointer(rt
->rt6i_node
));
1396 static void __ip6_rt_update_pmtu(struct dst_entry
*dst
, const struct sock
*sk
,
1397 const struct ipv6hdr
*iph
, u32 mtu
)
1399 const struct in6_addr
*daddr
, *saddr
;
1400 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1402 if (rt6
->rt6i_flags
& RTF_LOCAL
)
1405 if (dst_metric_locked(dst
, RTAX_MTU
))
1409 daddr
= &iph
->daddr
;
1410 saddr
= &iph
->saddr
;
1412 daddr
= &sk
->sk_v6_daddr
;
1413 saddr
= &inet6_sk(sk
)->saddr
;
1418 dst_confirm_neigh(dst
, daddr
);
1419 mtu
= max_t(u32
, mtu
, IPV6_MIN_MTU
);
1420 if (mtu
>= dst_mtu(dst
))
1423 if (!rt6_cache_allowed_for_pmtu(rt6
)) {
1424 rt6_do_update_pmtu(rt6
, mtu
);
1426 struct rt6_info
*nrt6
;
1428 nrt6
= ip6_rt_cache_alloc(rt6
, daddr
, saddr
);
1430 rt6_do_update_pmtu(nrt6
, mtu
);
1432 /* ip6_ins_rt(nrt6) will bump the
1433 * rt6->rt6i_node->fn_sernum
1434 * which will fail the next rt6_check() and
1435 * invalidate the sk->sk_dst_cache.
1438 /* Release the reference taken in
1439 * ip6_rt_cache_alloc()
1441 dst_release(&nrt6
->dst
);
1446 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1447 struct sk_buff
*skb
, u32 mtu
)
1449 __ip6_rt_update_pmtu(dst
, sk
, skb
? ipv6_hdr(skb
) : NULL
, mtu
);
1452 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1453 int oif
, u32 mark
, kuid_t uid
)
1455 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1456 struct dst_entry
*dst
;
1459 memset(&fl6
, 0, sizeof(fl6
));
1460 fl6
.flowi6_oif
= oif
;
1461 fl6
.flowi6_mark
= mark
? mark
: IP6_REPLY_MARK(net
, skb
->mark
);
1462 fl6
.daddr
= iph
->daddr
;
1463 fl6
.saddr
= iph
->saddr
;
1464 fl6
.flowlabel
= ip6_flowinfo(iph
);
1465 fl6
.flowi6_uid
= uid
;
1467 dst
= ip6_route_output(net
, NULL
, &fl6
);
1469 __ip6_rt_update_pmtu(dst
, NULL
, iph
, ntohl(mtu
));
1472 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1474 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1476 struct dst_entry
*dst
;
1478 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1479 sk
->sk_bound_dev_if
, sk
->sk_mark
, sk
->sk_uid
);
1481 dst
= __sk_dst_get(sk
);
1482 if (!dst
|| !dst
->obsolete
||
1483 dst
->ops
->check(dst
, inet6_sk(sk
)->dst_cookie
))
1487 if (!sock_owned_by_user(sk
) && !ipv6_addr_v4mapped(&sk
->sk_v6_daddr
))
1488 ip6_datagram_dst_update(sk
, false);
1491 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1493 /* Handle redirects */
1494 struct ip6rd_flowi
{
1496 struct in6_addr gateway
;
1499 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
1500 struct fib6_table
*table
,
1504 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl6
;
1505 struct rt6_info
*rt
;
1506 struct fib6_node
*fn
;
1508 /* Get the "current" route for this destination and
1509 * check if the redirect has come from appropriate router.
1511 * RFC 4861 specifies that redirects should only be
1512 * accepted if they come from the nexthop to the target.
1513 * Due to the way the routes are chosen, this notion
1514 * is a bit fuzzy and one might need to check all possible
1518 read_lock_bh(&table
->tb6_lock
);
1519 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1521 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1522 if (rt6_check_expired(rt
))
1526 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
1528 if (fl6
->flowi6_oif
!= rt
->dst
.dev
->ifindex
)
1530 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
))
1536 rt
= net
->ipv6
.ip6_null_entry
;
1537 else if (rt
->dst
.error
) {
1538 rt
= net
->ipv6
.ip6_null_entry
;
1542 if (rt
== net
->ipv6
.ip6_null_entry
) {
1543 fn
= fib6_backtrack(fn
, &fl6
->saddr
);
1551 read_unlock_bh(&table
->tb6_lock
);
1553 trace_fib6_table_lookup(net
, rt
, table
->tb6_id
, fl6
);
1557 static struct dst_entry
*ip6_route_redirect(struct net
*net
,
1558 const struct flowi6
*fl6
,
1559 const struct in6_addr
*gateway
)
1561 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1562 struct ip6rd_flowi rdfl
;
1565 rdfl
.gateway
= *gateway
;
1567 return fib6_rule_lookup(net
, &rdfl
.fl6
,
1568 flags
, __ip6_route_redirect
);
1571 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
,
1574 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1575 struct dst_entry
*dst
;
1578 memset(&fl6
, 0, sizeof(fl6
));
1579 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1580 fl6
.flowi6_oif
= oif
;
1581 fl6
.flowi6_mark
= mark
;
1582 fl6
.daddr
= iph
->daddr
;
1583 fl6
.saddr
= iph
->saddr
;
1584 fl6
.flowlabel
= ip6_flowinfo(iph
);
1585 fl6
.flowi6_uid
= uid
;
1587 dst
= ip6_route_redirect(net
, &fl6
, &ipv6_hdr(skb
)->saddr
);
1588 rt6_do_redirect(dst
, NULL
, skb
);
1591 EXPORT_SYMBOL_GPL(ip6_redirect
);
1593 void ip6_redirect_no_header(struct sk_buff
*skb
, struct net
*net
, int oif
,
1596 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1597 const struct rd_msg
*msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1598 struct dst_entry
*dst
;
1601 memset(&fl6
, 0, sizeof(fl6
));
1602 fl6
.flowi6_iif
= LOOPBACK_IFINDEX
;
1603 fl6
.flowi6_oif
= oif
;
1604 fl6
.flowi6_mark
= mark
;
1605 fl6
.daddr
= msg
->dest
;
1606 fl6
.saddr
= iph
->daddr
;
1607 fl6
.flowi6_uid
= sock_net_uid(net
, NULL
);
1609 dst
= ip6_route_redirect(net
, &fl6
, &iph
->saddr
);
1610 rt6_do_redirect(dst
, NULL
, skb
);
1614 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1616 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
,
1619 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1621 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1623 struct net_device
*dev
= dst
->dev
;
1624 unsigned int mtu
= dst_mtu(dst
);
1625 struct net
*net
= dev_net(dev
);
1627 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1629 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1630 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1633 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1634 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1635 * IPV6_MAXPLEN is also valid and means: "any MSS,
1636 * rely only on pmtu discovery"
1638 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1643 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1645 const struct rt6_info
*rt
= (const struct rt6_info
*)dst
;
1646 unsigned int mtu
= rt
->rt6i_pmtu
;
1647 struct inet6_dev
*idev
;
1652 mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1659 idev
= __in6_dev_get(dst
->dev
);
1661 mtu
= idev
->cnf
.mtu6
;
1665 mtu
= min_t(unsigned int, mtu
, IP6_MAX_MTU
);
1667 return mtu
- lwtunnel_headroom(dst
->lwtstate
, mtu
);
1670 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1673 struct dst_entry
*dst
;
1674 struct rt6_info
*rt
;
1675 struct inet6_dev
*idev
= in6_dev_get(dev
);
1676 struct net
*net
= dev_net(dev
);
1678 if (unlikely(!idev
))
1679 return ERR_PTR(-ENODEV
);
1681 rt
= ip6_dst_alloc(net
, dev
, 0);
1682 if (unlikely(!rt
)) {
1684 dst
= ERR_PTR(-ENOMEM
);
1688 rt
->dst
.flags
|= DST_HOST
;
1689 rt
->dst
.output
= ip6_output
;
1690 rt
->rt6i_gateway
= fl6
->daddr
;
1691 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1692 rt
->rt6i_dst
.plen
= 128;
1693 rt
->rt6i_idev
= idev
;
1694 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1696 /* Add this dst into uncached_list so that rt6_ifdown() can
1697 * do proper release of the net_device
1699 rt6_uncached_list_add(rt
);
1701 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1707 static int ip6_dst_gc(struct dst_ops
*ops
)
1709 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1710 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1711 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1712 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1713 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1714 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1717 entries
= dst_entries_get_fast(ops
);
1718 if (time_after(rt_last_gc
+ rt_min_interval
, jiffies
) &&
1719 entries
<= rt_max_size
)
1722 net
->ipv6
.ip6_rt_gc_expire
++;
1723 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
, true);
1724 entries
= dst_entries_get_slow(ops
);
1725 if (entries
< ops
->gc_thresh
)
1726 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1728 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1729 return entries
> rt_max_size
;
1732 static int ip6_convert_metrics(struct mx6_config
*mxc
,
1733 const struct fib6_config
*cfg
)
1735 bool ecn_ca
= false;
1743 mp
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_KERNEL
);
1747 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1748 int type
= nla_type(nla
);
1753 if (unlikely(type
> RTAX_MAX
))
1756 if (type
== RTAX_CC_ALGO
) {
1757 char tmp
[TCP_CA_NAME_MAX
];
1759 nla_strlcpy(tmp
, nla
, sizeof(tmp
));
1760 val
= tcp_ca_get_key_by_name(tmp
, &ecn_ca
);
1761 if (val
== TCP_CA_UNSPEC
)
1764 val
= nla_get_u32(nla
);
1766 if (type
== RTAX_HOPLIMIT
&& val
> 255)
1768 if (type
== RTAX_FEATURES
&& (val
& ~RTAX_FEATURE_MASK
))
1772 __set_bit(type
- 1, mxc
->mx_valid
);
1776 __set_bit(RTAX_FEATURES
- 1, mxc
->mx_valid
);
1777 mp
[RTAX_FEATURES
- 1] |= DST_FEATURE_ECN_CA
;
1787 static struct rt6_info
*ip6_nh_lookup_table(struct net
*net
,
1788 struct fib6_config
*cfg
,
1789 const struct in6_addr
*gw_addr
)
1791 struct flowi6 fl6
= {
1792 .flowi6_oif
= cfg
->fc_ifindex
,
1794 .saddr
= cfg
->fc_prefsrc
,
1796 struct fib6_table
*table
;
1797 struct rt6_info
*rt
;
1798 int flags
= RT6_LOOKUP_F_IFACE
| RT6_LOOKUP_F_IGNORE_LINKSTATE
;
1800 table
= fib6_get_table(net
, cfg
->fc_table
);
1804 if (!ipv6_addr_any(&cfg
->fc_prefsrc
))
1805 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1807 rt
= ip6_pol_route(net
, table
, cfg
->fc_ifindex
, &fl6
, flags
);
1809 /* if table lookup failed, fall back to full lookup */
1810 if (rt
== net
->ipv6
.ip6_null_entry
) {
1818 static struct rt6_info
*ip6_route_info_create(struct fib6_config
*cfg
,
1819 struct netlink_ext_ack
*extack
)
1821 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1822 struct rt6_info
*rt
= NULL
;
1823 struct net_device
*dev
= NULL
;
1824 struct inet6_dev
*idev
= NULL
;
1825 struct fib6_table
*table
;
1829 /* RTF_PCPU is an internal flag; can not be set by userspace */
1830 if (cfg
->fc_flags
& RTF_PCPU
) {
1831 NL_SET_ERR_MSG(extack
, "Userspace can not set RTF_PCPU");
1835 if (cfg
->fc_dst_len
> 128) {
1836 NL_SET_ERR_MSG(extack
, "Invalid prefix length");
1839 if (cfg
->fc_src_len
> 128) {
1840 NL_SET_ERR_MSG(extack
, "Invalid source address length");
1843 #ifndef CONFIG_IPV6_SUBTREES
1844 if (cfg
->fc_src_len
) {
1845 NL_SET_ERR_MSG(extack
,
1846 "Specifying source address requires IPV6_SUBTREES to be enabled");
1850 if (cfg
->fc_ifindex
) {
1852 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1855 idev
= in6_dev_get(dev
);
1860 if (cfg
->fc_metric
== 0)
1861 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1864 if (cfg
->fc_nlinfo
.nlh
&&
1865 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1866 table
= fib6_get_table(net
, cfg
->fc_table
);
1868 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1869 table
= fib6_new_table(net
, cfg
->fc_table
);
1872 table
= fib6_new_table(net
, cfg
->fc_table
);
1878 rt
= ip6_dst_alloc(net
, NULL
,
1879 (cfg
->fc_flags
& RTF_ADDRCONF
) ? 0 : DST_NOCOUNT
);
1886 if (cfg
->fc_flags
& RTF_EXPIRES
)
1887 rt6_set_expires(rt
, jiffies
+
1888 clock_t_to_jiffies(cfg
->fc_expires
));
1890 rt6_clean_expires(rt
);
1892 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1893 cfg
->fc_protocol
= RTPROT_BOOT
;
1894 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1896 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1898 if (addr_type
& IPV6_ADDR_MULTICAST
)
1899 rt
->dst
.input
= ip6_mc_input
;
1900 else if (cfg
->fc_flags
& RTF_LOCAL
)
1901 rt
->dst
.input
= ip6_input
;
1903 rt
->dst
.input
= ip6_forward
;
1905 rt
->dst
.output
= ip6_output
;
1907 if (cfg
->fc_encap
) {
1908 struct lwtunnel_state
*lwtstate
;
1910 err
= lwtunnel_build_state(cfg
->fc_encap_type
,
1911 cfg
->fc_encap
, AF_INET6
, cfg
,
1915 rt
->dst
.lwtstate
= lwtstate_get(lwtstate
);
1916 if (lwtunnel_output_redirect(rt
->dst
.lwtstate
)) {
1917 rt
->dst
.lwtstate
->orig_output
= rt
->dst
.output
;
1918 rt
->dst
.output
= lwtunnel_output
;
1920 if (lwtunnel_input_redirect(rt
->dst
.lwtstate
)) {
1921 rt
->dst
.lwtstate
->orig_input
= rt
->dst
.input
;
1922 rt
->dst
.input
= lwtunnel_input
;
1926 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1927 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1928 if (rt
->rt6i_dst
.plen
== 128)
1929 rt
->dst
.flags
|= DST_HOST
;
1931 #ifdef CONFIG_IPV6_SUBTREES
1932 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1933 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1936 rt
->rt6i_metric
= cfg
->fc_metric
;
1938 /* We cannot add true routes via loopback here,
1939 they would result in kernel looping; promote them to reject routes
1941 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1942 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1943 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1944 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1945 /* hold loopback dev/idev if we haven't done so. */
1946 if (dev
!= net
->loopback_dev
) {
1951 dev
= net
->loopback_dev
;
1953 idev
= in6_dev_get(dev
);
1959 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1960 switch (cfg
->fc_type
) {
1962 rt
->dst
.error
= -EINVAL
;
1963 rt
->dst
.output
= dst_discard_out
;
1964 rt
->dst
.input
= dst_discard
;
1967 rt
->dst
.error
= -EACCES
;
1968 rt
->dst
.output
= ip6_pkt_prohibit_out
;
1969 rt
->dst
.input
= ip6_pkt_prohibit
;
1972 case RTN_UNREACHABLE
:
1974 rt
->dst
.error
= (cfg
->fc_type
== RTN_THROW
) ? -EAGAIN
1975 : (cfg
->fc_type
== RTN_UNREACHABLE
)
1976 ? -EHOSTUNREACH
: -ENETUNREACH
;
1977 rt
->dst
.output
= ip6_pkt_discard_out
;
1978 rt
->dst
.input
= ip6_pkt_discard
;
1984 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1985 const struct in6_addr
*gw_addr
;
1988 gw_addr
= &cfg
->fc_gateway
;
1989 gwa_type
= ipv6_addr_type(gw_addr
);
1991 /* if gw_addr is local we will fail to detect this in case
1992 * address is still TENTATIVE (DAD in progress). rt6_lookup()
1993 * will return already-added prefix route via interface that
1994 * prefix route was assigned to, which might be non-loopback.
1997 if (ipv6_chk_addr_and_flags(net
, gw_addr
,
1998 gwa_type
& IPV6_ADDR_LINKLOCAL
?
1999 dev
: NULL
, 0, 0)) {
2000 NL_SET_ERR_MSG(extack
, "Invalid gateway address");
2003 rt
->rt6i_gateway
= *gw_addr
;
2005 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
2006 struct rt6_info
*grt
= NULL
;
2008 /* IPv6 strictly inhibits using not link-local
2009 addresses as nexthop address.
2010 Otherwise, router will not able to send redirects.
2011 It is very good, but in some (rare!) circumstances
2012 (SIT, PtP, NBMA NOARP links) it is handy to allow
2013 some exceptions. --ANK
2014 We allow IPv4-mapped nexthops to support RFC4798-type
2017 if (!(gwa_type
& (IPV6_ADDR_UNICAST
|
2018 IPV6_ADDR_MAPPED
))) {
2019 NL_SET_ERR_MSG(extack
,
2020 "Invalid gateway address");
2024 if (cfg
->fc_table
) {
2025 grt
= ip6_nh_lookup_table(net
, cfg
, gw_addr
);
2028 if (grt
->rt6i_flags
& RTF_GATEWAY
||
2029 (dev
&& dev
!= grt
->dst
.dev
)) {
2037 grt
= rt6_lookup(net
, gw_addr
, NULL
,
2038 cfg
->fc_ifindex
, 1);
2040 err
= -EHOSTUNREACH
;
2044 if (dev
!= grt
->dst
.dev
) {
2050 idev
= grt
->rt6i_idev
;
2052 in6_dev_hold(grt
->rt6i_idev
);
2054 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
2063 NL_SET_ERR_MSG(extack
, "Egress device not specified");
2065 } else if (dev
->flags
& IFF_LOOPBACK
) {
2066 NL_SET_ERR_MSG(extack
,
2067 "Egress device can not be loopback device for this route");
2076 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
2077 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
2078 NL_SET_ERR_MSG(extack
, "Invalid source address");
2082 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
2083 rt
->rt6i_prefsrc
.plen
= 128;
2085 rt
->rt6i_prefsrc
.plen
= 0;
2087 rt
->rt6i_flags
= cfg
->fc_flags
;
2091 rt
->rt6i_idev
= idev
;
2092 rt
->rt6i_table
= table
;
2094 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
2103 dst_release_immediate(&rt
->dst
);
2105 return ERR_PTR(err
);
2108 int ip6_route_add(struct fib6_config
*cfg
,
2109 struct netlink_ext_ack
*extack
)
2111 struct mx6_config mxc
= { .mx
= NULL
, };
2112 struct rt6_info
*rt
;
2115 rt
= ip6_route_info_create(cfg
, extack
);
2122 err
= ip6_convert_metrics(&mxc
, cfg
);
2126 err
= __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
, &mxc
, extack
);
2133 dst_release_immediate(&rt
->dst
);
2138 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
2141 struct fib6_table
*table
;
2142 struct net
*net
= dev_net(rt
->dst
.dev
);
2144 if (rt
== net
->ipv6
.ip6_null_entry
) {
2149 table
= rt
->rt6i_table
;
2150 write_lock_bh(&table
->tb6_lock
);
2151 err
= fib6_del(rt
, info
);
2152 write_unlock_bh(&table
->tb6_lock
);
2159 int ip6_del_rt(struct rt6_info
*rt
)
2161 struct nl_info info
= {
2162 .nl_net
= dev_net(rt
->dst
.dev
),
2164 return __ip6_del_rt(rt
, &info
);
2167 static int __ip6_del_rt_siblings(struct rt6_info
*rt
, struct fib6_config
*cfg
)
2169 struct nl_info
*info
= &cfg
->fc_nlinfo
;
2170 struct net
*net
= info
->nl_net
;
2171 struct sk_buff
*skb
= NULL
;
2172 struct fib6_table
*table
;
2175 if (rt
== net
->ipv6
.ip6_null_entry
)
2177 table
= rt
->rt6i_table
;
2178 write_lock_bh(&table
->tb6_lock
);
2180 if (rt
->rt6i_nsiblings
&& cfg
->fc_delete_all_nh
) {
2181 struct rt6_info
*sibling
, *next_sibling
;
2183 /* prefer to send a single notification with all hops */
2184 skb
= nlmsg_new(rt6_nlmsg_size(rt
), gfp_any());
2186 u32 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
2188 if (rt6_fill_node(net
, skb
, rt
,
2189 NULL
, NULL
, 0, RTM_DELROUTE
,
2190 info
->portid
, seq
, 0) < 0) {
2194 info
->skip_notify
= 1;
2197 list_for_each_entry_safe(sibling
, next_sibling
,
2200 err
= fib6_del(sibling
, info
);
2206 err
= fib6_del(rt
, info
);
2208 write_unlock_bh(&table
->tb6_lock
);
2213 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
2214 info
->nlh
, gfp_any());
2219 static int ip6_route_del(struct fib6_config
*cfg
,
2220 struct netlink_ext_ack
*extack
)
2222 struct fib6_table
*table
;
2223 struct fib6_node
*fn
;
2224 struct rt6_info
*rt
;
2227 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
2229 NL_SET_ERR_MSG(extack
, "FIB table does not exist");
2233 read_lock_bh(&table
->tb6_lock
);
2235 fn
= fib6_locate(&table
->tb6_root
,
2236 &cfg
->fc_dst
, cfg
->fc_dst_len
,
2237 &cfg
->fc_src
, cfg
->fc_src_len
);
2240 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2241 if ((rt
->rt6i_flags
& RTF_CACHE
) &&
2242 !(cfg
->fc_flags
& RTF_CACHE
))
2244 if (cfg
->fc_ifindex
&&
2246 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
2248 if (cfg
->fc_flags
& RTF_GATEWAY
&&
2249 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
2251 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
2253 if (cfg
->fc_protocol
&& cfg
->fc_protocol
!= rt
->rt6i_protocol
)
2256 read_unlock_bh(&table
->tb6_lock
);
2258 /* if gateway was specified only delete the one hop */
2259 if (cfg
->fc_flags
& RTF_GATEWAY
)
2260 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
2262 return __ip6_del_rt_siblings(rt
, cfg
);
2265 read_unlock_bh(&table
->tb6_lock
);
2270 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
2272 struct netevent_redirect netevent
;
2273 struct rt6_info
*rt
, *nrt
= NULL
;
2274 struct ndisc_options ndopts
;
2275 struct inet6_dev
*in6_dev
;
2276 struct neighbour
*neigh
;
2278 int optlen
, on_link
;
2281 optlen
= skb_tail_pointer(skb
) - skb_transport_header(skb
);
2282 optlen
-= sizeof(*msg
);
2285 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
2289 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
2291 if (ipv6_addr_is_multicast(&msg
->dest
)) {
2292 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
2297 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
2299 } else if (ipv6_addr_type(&msg
->target
) !=
2300 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
2301 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
2305 in6_dev
= __in6_dev_get(skb
->dev
);
2308 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
2312 * The IP source address of the Redirect MUST be the same as the current
2313 * first-hop router for the specified ICMP Destination Address.
2316 if (!ndisc_parse_options(skb
->dev
, msg
->opt
, optlen
, &ndopts
)) {
2317 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
2322 if (ndopts
.nd_opts_tgt_lladdr
) {
2323 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
2326 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
2331 rt
= (struct rt6_info
*) dst
;
2332 if (rt
->rt6i_flags
& RTF_REJECT
) {
2333 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
2337 /* Redirect received -> path was valid.
2338 * Look, redirects are sent only in response to data packets,
2339 * so that this nexthop apparently is reachable. --ANK
2341 dst_confirm_neigh(&rt
->dst
, &ipv6_hdr(skb
)->saddr
);
2343 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
2348 * We have finally decided to accept it.
2351 ndisc_update(skb
->dev
, neigh
, lladdr
, NUD_STALE
,
2352 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
2353 NEIGH_UPDATE_F_OVERRIDE
|
2354 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
2355 NEIGH_UPDATE_F_ISROUTER
)),
2356 NDISC_REDIRECT
, &ndopts
);
2358 nrt
= ip6_rt_cache_alloc(rt
, &msg
->dest
, NULL
);
2362 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
2364 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
2366 nrt
->rt6i_protocol
= RTPROT_REDIRECT
;
2367 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
2369 if (ip6_ins_rt(nrt
))
2372 netevent
.old
= &rt
->dst
;
2373 netevent
.new = &nrt
->dst
;
2374 netevent
.daddr
= &msg
->dest
;
2375 netevent
.neigh
= neigh
;
2376 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
2378 if (rt
->rt6i_flags
& RTF_CACHE
) {
2379 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
2384 /* Release the reference taken in
2385 * ip6_rt_cache_alloc()
2387 dst_release(&nrt
->dst
);
2390 neigh_release(neigh
);
2394 * Misc support functions
2397 static void rt6_set_from(struct rt6_info
*rt
, struct rt6_info
*from
)
2399 BUG_ON(from
->dst
.from
);
2401 rt
->rt6i_flags
&= ~RTF_EXPIRES
;
2402 dst_hold(&from
->dst
);
2403 rt
->dst
.from
= &from
->dst
;
2404 dst_init_metrics(&rt
->dst
, dst_metrics_ptr(&from
->dst
), true);
2407 static void ip6_rt_copy_init(struct rt6_info
*rt
, struct rt6_info
*ort
)
2409 rt
->dst
.input
= ort
->dst
.input
;
2410 rt
->dst
.output
= ort
->dst
.output
;
2411 rt
->rt6i_dst
= ort
->rt6i_dst
;
2412 rt
->dst
.error
= ort
->dst
.error
;
2413 rt
->rt6i_idev
= ort
->rt6i_idev
;
2415 in6_dev_hold(rt
->rt6i_idev
);
2416 rt
->dst
.lastuse
= jiffies
;
2417 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
2418 rt
->rt6i_flags
= ort
->rt6i_flags
;
2419 rt6_set_from(rt
, ort
);
2420 rt
->rt6i_metric
= ort
->rt6i_metric
;
2421 #ifdef CONFIG_IPV6_SUBTREES
2422 rt
->rt6i_src
= ort
->rt6i_src
;
2424 rt
->rt6i_prefsrc
= ort
->rt6i_prefsrc
;
2425 rt
->rt6i_table
= ort
->rt6i_table
;
2426 rt
->dst
.lwtstate
= lwtstate_get(ort
->dst
.lwtstate
);
2429 #ifdef CONFIG_IPV6_ROUTE_INFO
2430 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
2431 const struct in6_addr
*prefix
, int prefixlen
,
2432 const struct in6_addr
*gwaddr
,
2433 struct net_device
*dev
)
2435 u32 tb_id
= l3mdev_fib_table(dev
) ? : RT6_TABLE_INFO
;
2436 int ifindex
= dev
->ifindex
;
2437 struct fib6_node
*fn
;
2438 struct rt6_info
*rt
= NULL
;
2439 struct fib6_table
*table
;
2441 table
= fib6_get_table(net
, tb_id
);
2445 read_lock_bh(&table
->tb6_lock
);
2446 fn
= fib6_locate(&table
->tb6_root
, prefix
, prefixlen
, NULL
, 0);
2450 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2451 if (rt
->dst
.dev
->ifindex
!= ifindex
)
2453 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
2455 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
2461 read_unlock_bh(&table
->tb6_lock
);
2465 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
2466 const struct in6_addr
*prefix
, int prefixlen
,
2467 const struct in6_addr
*gwaddr
,
2468 struct net_device
*dev
,
2471 struct fib6_config cfg
= {
2472 .fc_metric
= IP6_RT_PRIO_USER
,
2473 .fc_ifindex
= dev
->ifindex
,
2474 .fc_dst_len
= prefixlen
,
2475 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
2476 RTF_UP
| RTF_PREF(pref
),
2477 .fc_protocol
= RTPROT_RA
,
2478 .fc_nlinfo
.portid
= 0,
2479 .fc_nlinfo
.nlh
= NULL
,
2480 .fc_nlinfo
.nl_net
= net
,
2483 cfg
.fc_table
= l3mdev_fib_table(dev
) ? : RT6_TABLE_INFO
,
2484 cfg
.fc_dst
= *prefix
;
2485 cfg
.fc_gateway
= *gwaddr
;
2487 /* We should treat it as a default route if prefix length is 0. */
2489 cfg
.fc_flags
|= RTF_DEFAULT
;
2491 ip6_route_add(&cfg
, NULL
);
2493 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, dev
);
2497 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
2499 u32 tb_id
= l3mdev_fib_table(dev
) ? : RT6_TABLE_DFLT
;
2500 struct rt6_info
*rt
;
2501 struct fib6_table
*table
;
2503 table
= fib6_get_table(dev_net(dev
), tb_id
);
2507 read_lock_bh(&table
->tb6_lock
);
2508 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2509 if (dev
== rt
->dst
.dev
&&
2510 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
2511 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
2516 read_unlock_bh(&table
->tb6_lock
);
2520 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
2521 struct net_device
*dev
,
2524 struct fib6_config cfg
= {
2525 .fc_table
= l3mdev_fib_table(dev
) ? : RT6_TABLE_DFLT
,
2526 .fc_metric
= IP6_RT_PRIO_USER
,
2527 .fc_ifindex
= dev
->ifindex
,
2528 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
2529 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
2530 .fc_protocol
= RTPROT_RA
,
2531 .fc_nlinfo
.portid
= 0,
2532 .fc_nlinfo
.nlh
= NULL
,
2533 .fc_nlinfo
.nl_net
= dev_net(dev
),
2536 cfg
.fc_gateway
= *gwaddr
;
2538 if (!ip6_route_add(&cfg
, NULL
)) {
2539 struct fib6_table
*table
;
2541 table
= fib6_get_table(dev_net(dev
), cfg
.fc_table
);
2543 table
->flags
|= RT6_TABLE_HAS_DFLT_ROUTER
;
2546 return rt6_get_dflt_router(gwaddr
, dev
);
2549 static void __rt6_purge_dflt_routers(struct fib6_table
*table
)
2551 struct rt6_info
*rt
;
2554 read_lock_bh(&table
->tb6_lock
);
2555 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2556 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
) &&
2557 (!rt
->rt6i_idev
|| rt
->rt6i_idev
->cnf
.accept_ra
!= 2)) {
2559 read_unlock_bh(&table
->tb6_lock
);
2564 read_unlock_bh(&table
->tb6_lock
);
2566 table
->flags
&= ~RT6_TABLE_HAS_DFLT_ROUTER
;
2569 void rt6_purge_dflt_routers(struct net
*net
)
2571 struct fib6_table
*table
;
2572 struct hlist_head
*head
;
2577 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
2578 head
= &net
->ipv6
.fib_table_hash
[h
];
2579 hlist_for_each_entry_rcu(table
, head
, tb6_hlist
) {
2580 if (table
->flags
& RT6_TABLE_HAS_DFLT_ROUTER
)
2581 __rt6_purge_dflt_routers(table
);
2588 static void rtmsg_to_fib6_config(struct net
*net
,
2589 struct in6_rtmsg
*rtmsg
,
2590 struct fib6_config
*cfg
)
2592 memset(cfg
, 0, sizeof(*cfg
));
2594 cfg
->fc_table
= l3mdev_fib_table_by_index(net
, rtmsg
->rtmsg_ifindex
) ?
2596 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
2597 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
2598 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
2599 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
2600 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
2601 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
2603 cfg
->fc_nlinfo
.nl_net
= net
;
2605 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
2606 cfg
->fc_src
= rtmsg
->rtmsg_src
;
2607 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
2610 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
2612 struct fib6_config cfg
;
2613 struct in6_rtmsg rtmsg
;
2617 case SIOCADDRT
: /* Add a route */
2618 case SIOCDELRT
: /* Delete a route */
2619 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2621 err
= copy_from_user(&rtmsg
, arg
,
2622 sizeof(struct in6_rtmsg
));
2626 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
2631 err
= ip6_route_add(&cfg
, NULL
);
2634 err
= ip6_route_del(&cfg
, NULL
);
2648 * Drop the packet on the floor
2651 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2654 struct dst_entry
*dst
= skb_dst(skb
);
2655 switch (ipstats_mib_noroutes
) {
2656 case IPSTATS_MIB_INNOROUTES
:
2657 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2658 if (type
== IPV6_ADDR_ANY
) {
2659 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2660 IPSTATS_MIB_INADDRERRORS
);
2664 case IPSTATS_MIB_OUTNOROUTES
:
2665 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2666 ipstats_mib_noroutes
);
2669 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2674 static int ip6_pkt_discard(struct sk_buff
*skb
)
2676 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2679 static int ip6_pkt_discard_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
2681 skb
->dev
= skb_dst(skb
)->dev
;
2682 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2685 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2687 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2690 static int ip6_pkt_prohibit_out(struct net
*net
, struct sock
*sk
, struct sk_buff
*skb
)
2692 skb
->dev
= skb_dst(skb
)->dev
;
2693 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2697 * Allocate a dst for local (unicast / anycast) address.
2700 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2701 const struct in6_addr
*addr
,
2705 struct net
*net
= dev_net(idev
->dev
);
2706 struct net_device
*dev
= net
->loopback_dev
;
2707 struct rt6_info
*rt
;
2709 /* use L3 Master device as loopback for host routes if device
2710 * is enslaved and address is not link local or multicast
2712 if (!rt6_need_strict(addr
))
2713 dev
= l3mdev_master_dev_rcu(idev
->dev
) ? : dev
;
2715 rt
= ip6_dst_alloc(net
, dev
, DST_NOCOUNT
);
2717 return ERR_PTR(-ENOMEM
);
2721 rt
->dst
.flags
|= DST_HOST
;
2722 rt
->dst
.input
= ip6_input
;
2723 rt
->dst
.output
= ip6_output
;
2724 rt
->rt6i_idev
= idev
;
2726 rt
->rt6i_protocol
= RTPROT_KERNEL
;
2727 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2729 rt
->rt6i_flags
|= RTF_ANYCAST
;
2731 rt
->rt6i_flags
|= RTF_LOCAL
;
2733 rt
->rt6i_gateway
= *addr
;
2734 rt
->rt6i_dst
.addr
= *addr
;
2735 rt
->rt6i_dst
.plen
= 128;
2736 tb_id
= l3mdev_fib_table(idev
->dev
) ? : RT6_TABLE_LOCAL
;
2737 rt
->rt6i_table
= fib6_get_table(net
, tb_id
);
2742 /* remove deleted ip from prefsrc entries */
2743 struct arg_dev_net_ip
{
2744 struct net_device
*dev
;
2746 struct in6_addr
*addr
;
2749 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2751 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2752 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2753 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2755 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2756 rt
!= net
->ipv6
.ip6_null_entry
&&
2757 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2758 /* remove prefsrc entry */
2759 rt
->rt6i_prefsrc
.plen
= 0;
2764 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2766 struct net
*net
= dev_net(ifp
->idev
->dev
);
2767 struct arg_dev_net_ip adni
= {
2768 .dev
= ifp
->idev
->dev
,
2772 fib6_clean_all(net
, fib6_remove_prefsrc
, &adni
);
2775 #define RTF_RA_ROUTER (RTF_ADDRCONF | RTF_DEFAULT | RTF_GATEWAY)
2776 #define RTF_CACHE_GATEWAY (RTF_GATEWAY | RTF_CACHE)
2778 /* Remove routers and update dst entries when gateway turn into host. */
2779 static int fib6_clean_tohost(struct rt6_info
*rt
, void *arg
)
2781 struct in6_addr
*gateway
= (struct in6_addr
*)arg
;
2783 if ((((rt
->rt6i_flags
& RTF_RA_ROUTER
) == RTF_RA_ROUTER
) ||
2784 ((rt
->rt6i_flags
& RTF_CACHE_GATEWAY
) == RTF_CACHE_GATEWAY
)) &&
2785 ipv6_addr_equal(gateway
, &rt
->rt6i_gateway
)) {
2791 void rt6_clean_tohost(struct net
*net
, struct in6_addr
*gateway
)
2793 fib6_clean_all(net
, fib6_clean_tohost
, gateway
);
2796 struct arg_dev_net
{
2797 struct net_device
*dev
;
2801 /* called with write lock held for table with rt */
2802 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2804 const struct arg_dev_net
*adn
= arg
;
2805 const struct net_device
*dev
= adn
->dev
;
2807 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2808 rt
!= adn
->net
->ipv6
.ip6_null_entry
&&
2809 (rt
->rt6i_nsiblings
== 0 ||
2810 (dev
&& netdev_unregistering(dev
)) ||
2811 !rt
->rt6i_idev
->cnf
.ignore_routes_with_linkdown
))
2817 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2819 struct arg_dev_net adn
= {
2824 fib6_clean_all(net
, fib6_ifdown
, &adn
);
2826 rt6_uncached_list_flush_dev(net
, dev
);
2829 struct rt6_mtu_change_arg
{
2830 struct net_device
*dev
;
2834 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2836 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2837 struct inet6_dev
*idev
;
2839 /* In IPv6 pmtu discovery is not optional,
2840 so that RTAX_MTU lock cannot disable it.
2841 We still use this lock to block changes
2842 caused by addrconf/ndisc.
2845 idev
= __in6_dev_get(arg
->dev
);
2849 /* For administrative MTU increase, there is no way to discover
2850 IPv6 PMTU increase, so PMTU increase should be updated here.
2851 Since RFC 1981 doesn't include administrative MTU increase
2852 update PMTU increase is a MUST. (i.e. jumbo frame)
2855 If new MTU is less than route PMTU, this new MTU will be the
2856 lowest MTU in the path, update the route PMTU to reflect PMTU
2857 decreases; if new MTU is greater than route PMTU, and the
2858 old MTU is the lowest MTU in the path, update the route PMTU
2859 to reflect the increase. In this case if the other nodes' MTU
2860 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2863 if (rt
->dst
.dev
== arg
->dev
&&
2864 dst_metric_raw(&rt
->dst
, RTAX_MTU
) &&
2865 !dst_metric_locked(&rt
->dst
, RTAX_MTU
)) {
2866 if (rt
->rt6i_flags
& RTF_CACHE
) {
2867 /* For RTF_CACHE with rt6i_pmtu == 0
2868 * (i.e. a redirected route),
2869 * the metrics of its rt->dst.from has already
2872 if (rt
->rt6i_pmtu
&& rt
->rt6i_pmtu
> arg
->mtu
)
2873 rt
->rt6i_pmtu
= arg
->mtu
;
2874 } else if (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2875 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2876 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
)) {
2877 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2883 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2885 struct rt6_mtu_change_arg arg
= {
2890 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, &arg
);
2893 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2894 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2895 [RTA_OIF
] = { .type
= NLA_U32
},
2896 [RTA_IIF
] = { .type
= NLA_U32
},
2897 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2898 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2899 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2900 [RTA_PREF
] = { .type
= NLA_U8
},
2901 [RTA_ENCAP_TYPE
] = { .type
= NLA_U16
},
2902 [RTA_ENCAP
] = { .type
= NLA_NESTED
},
2903 [RTA_EXPIRES
] = { .type
= NLA_U32
},
2904 [RTA_UID
] = { .type
= NLA_U32
},
2905 [RTA_MARK
] = { .type
= NLA_U32
},
2908 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2909 struct fib6_config
*cfg
,
2910 struct netlink_ext_ack
*extack
)
2913 struct nlattr
*tb
[RTA_MAX
+1];
2917 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
,
2923 rtm
= nlmsg_data(nlh
);
2924 memset(cfg
, 0, sizeof(*cfg
));
2926 cfg
->fc_table
= rtm
->rtm_table
;
2927 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2928 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2929 cfg
->fc_flags
= RTF_UP
;
2930 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2931 cfg
->fc_type
= rtm
->rtm_type
;
2933 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2934 rtm
->rtm_type
== RTN_BLACKHOLE
||
2935 rtm
->rtm_type
== RTN_PROHIBIT
||
2936 rtm
->rtm_type
== RTN_THROW
)
2937 cfg
->fc_flags
|= RTF_REJECT
;
2939 if (rtm
->rtm_type
== RTN_LOCAL
)
2940 cfg
->fc_flags
|= RTF_LOCAL
;
2942 if (rtm
->rtm_flags
& RTM_F_CLONED
)
2943 cfg
->fc_flags
|= RTF_CACHE
;
2945 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2946 cfg
->fc_nlinfo
.nlh
= nlh
;
2947 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2949 if (tb
[RTA_GATEWAY
]) {
2950 cfg
->fc_gateway
= nla_get_in6_addr(tb
[RTA_GATEWAY
]);
2951 cfg
->fc_flags
|= RTF_GATEWAY
;
2955 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2957 if (nla_len(tb
[RTA_DST
]) < plen
)
2960 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2964 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2966 if (nla_len(tb
[RTA_SRC
]) < plen
)
2969 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2972 if (tb
[RTA_PREFSRC
])
2973 cfg
->fc_prefsrc
= nla_get_in6_addr(tb
[RTA_PREFSRC
]);
2976 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2978 if (tb
[RTA_PRIORITY
])
2979 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2981 if (tb
[RTA_METRICS
]) {
2982 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2983 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2987 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2989 if (tb
[RTA_MULTIPATH
]) {
2990 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2991 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2993 err
= lwtunnel_valid_encap_type_attr(cfg
->fc_mp
,
2994 cfg
->fc_mp_len
, extack
);
3000 pref
= nla_get_u8(tb
[RTA_PREF
]);
3001 if (pref
!= ICMPV6_ROUTER_PREF_LOW
&&
3002 pref
!= ICMPV6_ROUTER_PREF_HIGH
)
3003 pref
= ICMPV6_ROUTER_PREF_MEDIUM
;
3004 cfg
->fc_flags
|= RTF_PREF(pref
);
3008 cfg
->fc_encap
= tb
[RTA_ENCAP
];
3010 if (tb
[RTA_ENCAP_TYPE
]) {
3011 cfg
->fc_encap_type
= nla_get_u16(tb
[RTA_ENCAP_TYPE
]);
3013 err
= lwtunnel_valid_encap_type(cfg
->fc_encap_type
, extack
);
3018 if (tb
[RTA_EXPIRES
]) {
3019 unsigned long timeout
= addrconf_timeout_fixup(nla_get_u32(tb
[RTA_EXPIRES
]), HZ
);
3021 if (addrconf_finite_timeout(timeout
)) {
3022 cfg
->fc_expires
= jiffies_to_clock_t(timeout
* HZ
);
3023 cfg
->fc_flags
|= RTF_EXPIRES
;
3033 struct rt6_info
*rt6_info
;
3034 struct fib6_config r_cfg
;
3035 struct mx6_config mxc
;
3036 struct list_head next
;
3039 static void ip6_print_replace_route_err(struct list_head
*rt6_nh_list
)
3043 list_for_each_entry(nh
, rt6_nh_list
, next
) {
3044 pr_warn("IPV6: multipath route replace failed (check consistency of installed routes): %pI6c nexthop %pI6c ifi %d\n",
3045 &nh
->r_cfg
.fc_dst
, &nh
->r_cfg
.fc_gateway
,
3046 nh
->r_cfg
.fc_ifindex
);
3050 static int ip6_route_info_append(struct list_head
*rt6_nh_list
,
3051 struct rt6_info
*rt
, struct fib6_config
*r_cfg
)
3056 list_for_each_entry(nh
, rt6_nh_list
, next
) {
3057 /* check if rt6_info already exists */
3058 if (rt6_duplicate_nexthop(nh
->rt6_info
, rt
))
3062 nh
= kzalloc(sizeof(*nh
), GFP_KERNEL
);
3066 err
= ip6_convert_metrics(&nh
->mxc
, r_cfg
);
3071 memcpy(&nh
->r_cfg
, r_cfg
, sizeof(*r_cfg
));
3072 list_add_tail(&nh
->next
, rt6_nh_list
);
3077 static void ip6_route_mpath_notify(struct rt6_info
*rt
,
3078 struct rt6_info
*rt_last
,
3079 struct nl_info
*info
,
3082 /* if this is an APPEND route, then rt points to the first route
3083 * inserted and rt_last points to last route inserted. Userspace
3084 * wants a consistent dump of the route which starts at the first
3085 * nexthop. Since sibling routes are always added at the end of
3086 * the list, find the first sibling of the last route appended
3088 if ((nlflags
& NLM_F_APPEND
) && rt_last
&& rt_last
->rt6i_nsiblings
) {
3089 rt
= list_first_entry(&rt_last
->rt6i_siblings
,
3095 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
, nlflags
);
3098 static int ip6_route_multipath_add(struct fib6_config
*cfg
,
3099 struct netlink_ext_ack
*extack
)
3101 struct rt6_info
*rt_notif
= NULL
, *rt_last
= NULL
;
3102 struct nl_info
*info
= &cfg
->fc_nlinfo
;
3103 struct fib6_config r_cfg
;
3104 struct rtnexthop
*rtnh
;
3105 struct rt6_info
*rt
;
3106 struct rt6_nh
*err_nh
;
3107 struct rt6_nh
*nh
, *nh_safe
;
3113 int replace
= (cfg
->fc_nlinfo
.nlh
&&
3114 (cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_REPLACE
));
3115 LIST_HEAD(rt6_nh_list
);
3117 nlflags
= replace
? NLM_F_REPLACE
: NLM_F_CREATE
;
3118 if (info
->nlh
&& info
->nlh
->nlmsg_flags
& NLM_F_APPEND
)
3119 nlflags
|= NLM_F_APPEND
;
3121 remaining
= cfg
->fc_mp_len
;
3122 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
3124 /* Parse a Multipath Entry and build a list (rt6_nh_list) of
3125 * rt6_info structs per nexthop
3127 while (rtnh_ok(rtnh
, remaining
)) {
3128 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
3129 if (rtnh
->rtnh_ifindex
)
3130 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
3132 attrlen
= rtnh_attrlen(rtnh
);
3134 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
3136 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
3138 r_cfg
.fc_gateway
= nla_get_in6_addr(nla
);
3139 r_cfg
.fc_flags
|= RTF_GATEWAY
;
3141 r_cfg
.fc_encap
= nla_find(attrs
, attrlen
, RTA_ENCAP
);
3142 nla
= nla_find(attrs
, attrlen
, RTA_ENCAP_TYPE
);
3144 r_cfg
.fc_encap_type
= nla_get_u16(nla
);
3147 rt
= ip6_route_info_create(&r_cfg
, extack
);
3154 err
= ip6_route_info_append(&rt6_nh_list
, rt
, &r_cfg
);
3156 dst_release_immediate(&rt
->dst
);
3160 rtnh
= rtnh_next(rtnh
, &remaining
);
3163 /* for add and replace send one notification with all nexthops.
3164 * Skip the notification in fib6_add_rt2node and send one with
3165 * the full route when done
3167 info
->skip_notify
= 1;
3170 list_for_each_entry(nh
, &rt6_nh_list
, next
) {
3171 rt_last
= nh
->rt6_info
;
3172 err
= __ip6_ins_rt(nh
->rt6_info
, info
, &nh
->mxc
, extack
);
3173 /* save reference to first route for notification */
3174 if (!rt_notif
&& !err
)
3175 rt_notif
= nh
->rt6_info
;
3177 /* nh->rt6_info is used or freed at this point, reset to NULL*/
3178 nh
->rt6_info
= NULL
;
3181 ip6_print_replace_route_err(&rt6_nh_list
);
3186 /* Because each route is added like a single route we remove
3187 * these flags after the first nexthop: if there is a collision,
3188 * we have already failed to add the first nexthop:
3189 * fib6_add_rt2node() has rejected it; when replacing, old
3190 * nexthops have been replaced by first new, the rest should
3193 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~(NLM_F_EXCL
|
3198 /* success ... tell user about new route */
3199 ip6_route_mpath_notify(rt_notif
, rt_last
, info
, nlflags
);
3203 /* send notification for routes that were added so that
3204 * the delete notifications sent by ip6_route_del are
3208 ip6_route_mpath_notify(rt_notif
, rt_last
, info
, nlflags
);
3210 /* Delete routes that were already added */
3211 list_for_each_entry(nh
, &rt6_nh_list
, next
) {
3214 ip6_route_del(&nh
->r_cfg
, extack
);
3218 list_for_each_entry_safe(nh
, nh_safe
, &rt6_nh_list
, next
) {
3220 dst_release_immediate(&nh
->rt6_info
->dst
);
3222 list_del(&nh
->next
);
3229 static int ip6_route_multipath_del(struct fib6_config
*cfg
,
3230 struct netlink_ext_ack
*extack
)
3232 struct fib6_config r_cfg
;
3233 struct rtnexthop
*rtnh
;
3236 int err
= 1, last_err
= 0;
3238 remaining
= cfg
->fc_mp_len
;
3239 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
3241 /* Parse a Multipath Entry */
3242 while (rtnh_ok(rtnh
, remaining
)) {
3243 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
3244 if (rtnh
->rtnh_ifindex
)
3245 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
3247 attrlen
= rtnh_attrlen(rtnh
);
3249 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
3251 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
3253 nla_memcpy(&r_cfg
.fc_gateway
, nla
, 16);
3254 r_cfg
.fc_flags
|= RTF_GATEWAY
;
3257 err
= ip6_route_del(&r_cfg
, extack
);
3261 rtnh
= rtnh_next(rtnh
, &remaining
);
3267 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3268 struct netlink_ext_ack
*extack
)
3270 struct fib6_config cfg
;
3273 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
, extack
);
3278 return ip6_route_multipath_del(&cfg
, extack
);
3280 cfg
.fc_delete_all_nh
= 1;
3281 return ip6_route_del(&cfg
, extack
);
3285 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
3286 struct netlink_ext_ack
*extack
)
3288 struct fib6_config cfg
;
3291 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
, extack
);
3296 return ip6_route_multipath_add(&cfg
, extack
);
3298 return ip6_route_add(&cfg
, extack
);
3301 static size_t rt6_nlmsg_size(struct rt6_info
*rt
)
3303 int nexthop_len
= 0;
3305 if (rt
->rt6i_nsiblings
) {
3306 nexthop_len
= nla_total_size(0) /* RTA_MULTIPATH */
3307 + NLA_ALIGN(sizeof(struct rtnexthop
))
3308 + nla_total_size(16) /* RTA_GATEWAY */
3309 + lwtunnel_get_encap_size(rt
->dst
.lwtstate
);
3311 nexthop_len
*= rt
->rt6i_nsiblings
;
3314 return NLMSG_ALIGN(sizeof(struct rtmsg
))
3315 + nla_total_size(16) /* RTA_SRC */
3316 + nla_total_size(16) /* RTA_DST */
3317 + nla_total_size(16) /* RTA_GATEWAY */
3318 + nla_total_size(16) /* RTA_PREFSRC */
3319 + nla_total_size(4) /* RTA_TABLE */
3320 + nla_total_size(4) /* RTA_IIF */
3321 + nla_total_size(4) /* RTA_OIF */
3322 + nla_total_size(4) /* RTA_PRIORITY */
3323 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
3324 + nla_total_size(sizeof(struct rta_cacheinfo
))
3325 + nla_total_size(TCP_CA_NAME_MAX
) /* RTAX_CC_ALGO */
3326 + nla_total_size(1) /* RTA_PREF */
3327 + lwtunnel_get_encap_size(rt
->dst
.lwtstate
)
3331 static int rt6_nexthop_info(struct sk_buff
*skb
, struct rt6_info
*rt
,
3332 unsigned int *flags
, bool skip_oif
)
3334 if (!netif_running(rt
->dst
.dev
) || !netif_carrier_ok(rt
->dst
.dev
)) {
3335 *flags
|= RTNH_F_LINKDOWN
;
3336 if (rt
->rt6i_idev
->cnf
.ignore_routes_with_linkdown
)
3337 *flags
|= RTNH_F_DEAD
;
3340 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
3341 if (nla_put_in6_addr(skb
, RTA_GATEWAY
, &rt
->rt6i_gateway
) < 0)
3342 goto nla_put_failure
;
3345 /* not needed for multipath encoding b/c it has a rtnexthop struct */
3346 if (!skip_oif
&& rt
->dst
.dev
&&
3347 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
3348 goto nla_put_failure
;
3350 if (rt
->dst
.lwtstate
&&
3351 lwtunnel_fill_encap(skb
, rt
->dst
.lwtstate
) < 0)
3352 goto nla_put_failure
;
3360 /* add multipath next hop */
3361 static int rt6_add_nexthop(struct sk_buff
*skb
, struct rt6_info
*rt
)
3363 struct rtnexthop
*rtnh
;
3364 unsigned int flags
= 0;
3366 rtnh
= nla_reserve_nohdr(skb
, sizeof(*rtnh
));
3368 goto nla_put_failure
;
3370 rtnh
->rtnh_hops
= 0;
3371 rtnh
->rtnh_ifindex
= rt
->dst
.dev
? rt
->dst
.dev
->ifindex
: 0;
3373 if (rt6_nexthop_info(skb
, rt
, &flags
, true) < 0)
3374 goto nla_put_failure
;
3376 rtnh
->rtnh_flags
= flags
;
3378 /* length of rtnetlink header + attributes */
3379 rtnh
->rtnh_len
= nlmsg_get_pos(skb
) - (void *)rtnh
;
3387 static int rt6_fill_node(struct net
*net
,
3388 struct sk_buff
*skb
, struct rt6_info
*rt
,
3389 struct in6_addr
*dst
, struct in6_addr
*src
,
3390 int iif
, int type
, u32 portid
, u32 seq
,
3393 u32 metrics
[RTAX_MAX
];
3395 struct nlmsghdr
*nlh
;
3399 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
3403 rtm
= nlmsg_data(nlh
);
3404 rtm
->rtm_family
= AF_INET6
;
3405 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
3406 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
3409 table
= rt
->rt6i_table
->tb6_id
;
3411 table
= RT6_TABLE_UNSPEC
;
3412 rtm
->rtm_table
= table
;
3413 if (nla_put_u32(skb
, RTA_TABLE
, table
))
3414 goto nla_put_failure
;
3415 if (rt
->rt6i_flags
& RTF_REJECT
) {
3416 switch (rt
->dst
.error
) {
3418 rtm
->rtm_type
= RTN_BLACKHOLE
;
3421 rtm
->rtm_type
= RTN_PROHIBIT
;
3424 rtm
->rtm_type
= RTN_THROW
;
3427 rtm
->rtm_type
= RTN_UNREACHABLE
;
3431 else if (rt
->rt6i_flags
& RTF_LOCAL
)
3432 rtm
->rtm_type
= RTN_LOCAL
;
3433 else if (rt
->rt6i_flags
& RTF_ANYCAST
)
3434 rtm
->rtm_type
= RTN_ANYCAST
;
3435 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
3436 rtm
->rtm_type
= RTN_LOCAL
;
3438 rtm
->rtm_type
= RTN_UNICAST
;
3440 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
3441 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
3443 if (rt
->rt6i_flags
& RTF_CACHE
)
3444 rtm
->rtm_flags
|= RTM_F_CLONED
;
3447 if (nla_put_in6_addr(skb
, RTA_DST
, dst
))
3448 goto nla_put_failure
;
3449 rtm
->rtm_dst_len
= 128;
3450 } else if (rtm
->rtm_dst_len
)
3451 if (nla_put_in6_addr(skb
, RTA_DST
, &rt
->rt6i_dst
.addr
))
3452 goto nla_put_failure
;
3453 #ifdef CONFIG_IPV6_SUBTREES
3455 if (nla_put_in6_addr(skb
, RTA_SRC
, src
))
3456 goto nla_put_failure
;
3457 rtm
->rtm_src_len
= 128;
3458 } else if (rtm
->rtm_src_len
&&
3459 nla_put_in6_addr(skb
, RTA_SRC
, &rt
->rt6i_src
.addr
))
3460 goto nla_put_failure
;
3463 #ifdef CONFIG_IPV6_MROUTE
3464 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
3465 int err
= ip6mr_get_route(net
, skb
, rtm
, portid
);
3470 goto nla_put_failure
;
3473 if (nla_put_u32(skb
, RTA_IIF
, iif
))
3474 goto nla_put_failure
;
3476 struct in6_addr saddr_buf
;
3477 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
3478 nla_put_in6_addr(skb
, RTA_PREFSRC
, &saddr_buf
))
3479 goto nla_put_failure
;
3482 if (rt
->rt6i_prefsrc
.plen
) {
3483 struct in6_addr saddr_buf
;
3484 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
3485 if (nla_put_in6_addr(skb
, RTA_PREFSRC
, &saddr_buf
))
3486 goto nla_put_failure
;
3489 memcpy(metrics
, dst_metrics_ptr(&rt
->dst
), sizeof(metrics
));
3491 metrics
[RTAX_MTU
- 1] = rt
->rt6i_pmtu
;
3492 if (rtnetlink_put_metrics(skb
, metrics
) < 0)
3493 goto nla_put_failure
;
3495 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
3496 goto nla_put_failure
;
3498 /* For multipath routes, walk the siblings list and add
3499 * each as a nexthop within RTA_MULTIPATH.
3501 if (rt
->rt6i_nsiblings
) {
3502 struct rt6_info
*sibling
, *next_sibling
;
3505 mp
= nla_nest_start(skb
, RTA_MULTIPATH
);
3507 goto nla_put_failure
;
3509 if (rt6_add_nexthop(skb
, rt
) < 0)
3510 goto nla_put_failure
;
3512 list_for_each_entry_safe(sibling
, next_sibling
,
3513 &rt
->rt6i_siblings
, rt6i_siblings
) {
3514 if (rt6_add_nexthop(skb
, sibling
) < 0)
3515 goto nla_put_failure
;
3518 nla_nest_end(skb
, mp
);
3520 if (rt6_nexthop_info(skb
, rt
, &rtm
->rtm_flags
, false) < 0)
3521 goto nla_put_failure
;
3524 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
3526 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
3527 goto nla_put_failure
;
3529 if (nla_put_u8(skb
, RTA_PREF
, IPV6_EXTRACT_PREF(rt
->rt6i_flags
)))
3530 goto nla_put_failure
;
3533 nlmsg_end(skb
, nlh
);
3537 nlmsg_cancel(skb
, nlh
);
3541 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
3543 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
3544 struct net
*net
= arg
->net
;
3546 if (rt
== net
->ipv6
.ip6_null_entry
)
3549 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
3550 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
3552 /* user wants prefix routes only */
3553 if (rtm
->rtm_flags
& RTM_F_PREFIX
&&
3554 !(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
3555 /* success since this is not a prefix route */
3560 return rt6_fill_node(net
,
3561 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
3562 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
3566 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
*nlh
,
3567 struct netlink_ext_ack
*extack
)
3569 struct net
*net
= sock_net(in_skb
->sk
);
3570 struct nlattr
*tb
[RTA_MAX
+1];
3571 int err
, iif
= 0, oif
= 0;
3572 struct dst_entry
*dst
;
3573 struct rt6_info
*rt
;
3574 struct sk_buff
*skb
;
3579 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
,
3585 memset(&fl6
, 0, sizeof(fl6
));
3586 rtm
= nlmsg_data(nlh
);
3587 fl6
.flowlabel
= ip6_make_flowinfo(rtm
->rtm_tos
, 0);
3588 fibmatch
= !!(rtm
->rtm_flags
& RTM_F_FIB_MATCH
);
3591 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
3594 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
3598 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
3601 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
3605 iif
= nla_get_u32(tb
[RTA_IIF
]);
3608 oif
= nla_get_u32(tb
[RTA_OIF
]);
3611 fl6
.flowi6_mark
= nla_get_u32(tb
[RTA_MARK
]);
3614 fl6
.flowi6_uid
= make_kuid(current_user_ns(),
3615 nla_get_u32(tb
[RTA_UID
]));
3617 fl6
.flowi6_uid
= iif
? INVALID_UID
: current_uid();
3620 struct net_device
*dev
;
3623 dev
= __dev_get_by_index(net
, iif
);
3629 fl6
.flowi6_iif
= iif
;
3631 if (!ipv6_addr_any(&fl6
.saddr
))
3632 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
3635 dst
= ip6_route_input_lookup(net
, dev
, &fl6
, flags
);
3637 fl6
.flowi6_oif
= oif
;
3640 dst
= ip6_route_output(net
, NULL
, &fl6
);
3644 dst
= ip6_route_lookup(net
, &fl6
, 0);
3646 rt
= container_of(dst
, struct rt6_info
, dst
);
3647 if (rt
->dst
.error
) {
3648 err
= rt
->dst
.error
;
3653 if (rt
== net
->ipv6
.ip6_null_entry
) {
3654 err
= rt
->dst
.error
;
3659 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
3666 skb_dst_set(skb
, &rt
->dst
);
3668 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, iif
,
3669 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
3672 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
3673 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
3680 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
3685 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
,
3686 unsigned int nlm_flags
)
3688 struct sk_buff
*skb
;
3689 struct net
*net
= info
->nl_net
;
3694 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
3696 skb
= nlmsg_new(rt6_nlmsg_size(rt
), gfp_any());
3700 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
3701 event
, info
->portid
, seq
, nlm_flags
);
3703 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
3704 WARN_ON(err
== -EMSGSIZE
);
3708 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
3709 info
->nlh
, gfp_any());
3713 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
3716 static int ip6_route_dev_notify(struct notifier_block
*this,
3717 unsigned long event
, void *ptr
)
3719 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
3720 struct net
*net
= dev_net(dev
);
3722 if (!(dev
->flags
& IFF_LOOPBACK
))
3725 if (event
== NETDEV_REGISTER
) {
3726 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
3727 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
3728 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3729 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
3730 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
3731 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
3732 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
3734 } else if (event
== NETDEV_UNREGISTER
&&
3735 dev
->reg_state
!= NETREG_UNREGISTERED
) {
3736 /* NETDEV_UNREGISTER could be fired for multiple times by
3737 * netdev_wait_allrefs(). Make sure we only call this once.
3739 in6_dev_put_clear(&net
->ipv6
.ip6_null_entry
->rt6i_idev
);
3740 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3741 in6_dev_put_clear(&net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
);
3742 in6_dev_put_clear(&net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
);
3753 #ifdef CONFIG_PROC_FS
3755 static const struct file_operations ipv6_route_proc_fops
= {
3756 .owner
= THIS_MODULE
,
3757 .open
= ipv6_route_open
,
3759 .llseek
= seq_lseek
,
3760 .release
= seq_release_net
,
3763 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
3765 struct net
*net
= (struct net
*)seq
->private;
3766 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
3767 net
->ipv6
.rt6_stats
->fib_nodes
,
3768 net
->ipv6
.rt6_stats
->fib_route_nodes
,
3769 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
3770 net
->ipv6
.rt6_stats
->fib_rt_entries
,
3771 net
->ipv6
.rt6_stats
->fib_rt_cache
,
3772 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
3773 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
3778 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
3780 return single_open_net(inode
, file
, rt6_stats_seq_show
);
3783 static const struct file_operations rt6_stats_seq_fops
= {
3784 .owner
= THIS_MODULE
,
3785 .open
= rt6_stats_seq_open
,
3787 .llseek
= seq_lseek
,
3788 .release
= single_release_net
,
3790 #endif /* CONFIG_PROC_FS */
3792 #ifdef CONFIG_SYSCTL
3795 int ipv6_sysctl_rtcache_flush(struct ctl_table
*ctl
, int write
,
3796 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
3803 net
= (struct net
*)ctl
->extra1
;
3804 delay
= net
->ipv6
.sysctl
.flush_delay
;
3805 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
3806 fib6_run_gc(delay
<= 0 ? 0 : (unsigned long)delay
, net
, delay
> 0);
3810 struct ctl_table ipv6_route_table_template
[] = {
3812 .procname
= "flush",
3813 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
3814 .maxlen
= sizeof(int),
3816 .proc_handler
= ipv6_sysctl_rtcache_flush
3819 .procname
= "gc_thresh",
3820 .data
= &ip6_dst_ops_template
.gc_thresh
,
3821 .maxlen
= sizeof(int),
3823 .proc_handler
= proc_dointvec
,
3826 .procname
= "max_size",
3827 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
3828 .maxlen
= sizeof(int),
3830 .proc_handler
= proc_dointvec
,
3833 .procname
= "gc_min_interval",
3834 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
3835 .maxlen
= sizeof(int),
3837 .proc_handler
= proc_dointvec_jiffies
,
3840 .procname
= "gc_timeout",
3841 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
3842 .maxlen
= sizeof(int),
3844 .proc_handler
= proc_dointvec_jiffies
,
3847 .procname
= "gc_interval",
3848 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
3849 .maxlen
= sizeof(int),
3851 .proc_handler
= proc_dointvec_jiffies
,
3854 .procname
= "gc_elasticity",
3855 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
3856 .maxlen
= sizeof(int),
3858 .proc_handler
= proc_dointvec
,
3861 .procname
= "mtu_expires",
3862 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
3863 .maxlen
= sizeof(int),
3865 .proc_handler
= proc_dointvec_jiffies
,
3868 .procname
= "min_adv_mss",
3869 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
3870 .maxlen
= sizeof(int),
3872 .proc_handler
= proc_dointvec
,
3875 .procname
= "gc_min_interval_ms",
3876 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
3877 .maxlen
= sizeof(int),
3879 .proc_handler
= proc_dointvec_ms_jiffies
,
3884 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
3886 struct ctl_table
*table
;
3888 table
= kmemdup(ipv6_route_table_template
,
3889 sizeof(ipv6_route_table_template
),
3893 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
3894 table
[0].extra1
= net
;
3895 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
3896 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
3897 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
3898 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
3899 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
3900 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
3901 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
3902 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
3903 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
3905 /* Don't export sysctls to unprivileged users */
3906 if (net
->user_ns
!= &init_user_ns
)
3907 table
[0].procname
= NULL
;
3914 static int __net_init
ip6_route_net_init(struct net
*net
)
3918 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
3919 sizeof(net
->ipv6
.ip6_dst_ops
));
3921 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
3922 goto out_ip6_dst_ops
;
3924 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
3925 sizeof(*net
->ipv6
.ip6_null_entry
),
3927 if (!net
->ipv6
.ip6_null_entry
)
3928 goto out_ip6_dst_entries
;
3929 net
->ipv6
.ip6_null_entry
->dst
.path
=
3930 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
3931 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3932 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
3933 ip6_template_metrics
, true);
3935 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3936 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
3937 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
3939 if (!net
->ipv6
.ip6_prohibit_entry
)
3940 goto out_ip6_null_entry
;
3941 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
3942 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
3943 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3944 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
3945 ip6_template_metrics
, true);
3947 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
3948 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
3950 if (!net
->ipv6
.ip6_blk_hole_entry
)
3951 goto out_ip6_prohibit_entry
;
3952 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
3953 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
3954 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3955 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
3956 ip6_template_metrics
, true);
3959 net
->ipv6
.sysctl
.flush_delay
= 0;
3960 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
3961 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
3962 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
3963 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
3964 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
3965 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
3966 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
3968 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
3974 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3975 out_ip6_prohibit_entry
:
3976 kfree(net
->ipv6
.ip6_prohibit_entry
);
3978 kfree(net
->ipv6
.ip6_null_entry
);
3980 out_ip6_dst_entries
:
3981 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3986 static void __net_exit
ip6_route_net_exit(struct net
*net
)
3988 kfree(net
->ipv6
.ip6_null_entry
);
3989 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3990 kfree(net
->ipv6
.ip6_prohibit_entry
);
3991 kfree(net
->ipv6
.ip6_blk_hole_entry
);
3993 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3996 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3998 #ifdef CONFIG_PROC_FS
3999 proc_create("ipv6_route", 0, net
->proc_net
, &ipv6_route_proc_fops
);
4000 proc_create("rt6_stats", S_IRUGO
, net
->proc_net
, &rt6_stats_seq_fops
);
4005 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
4007 #ifdef CONFIG_PROC_FS
4008 remove_proc_entry("ipv6_route", net
->proc_net
);
4009 remove_proc_entry("rt6_stats", net
->proc_net
);
4013 static struct pernet_operations ip6_route_net_ops
= {
4014 .init
= ip6_route_net_init
,
4015 .exit
= ip6_route_net_exit
,
4018 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
4020 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
4024 inet_peer_base_init(bp
);
4025 net
->ipv6
.peers
= bp
;
4029 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
4031 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
4033 net
->ipv6
.peers
= NULL
;
4034 inetpeer_invalidate_tree(bp
);
4038 static struct pernet_operations ipv6_inetpeer_ops
= {
4039 .init
= ipv6_inetpeer_init
,
4040 .exit
= ipv6_inetpeer_exit
,
4043 static struct pernet_operations ip6_route_net_late_ops
= {
4044 .init
= ip6_route_net_init_late
,
4045 .exit
= ip6_route_net_exit_late
,
4048 static struct notifier_block ip6_route_dev_notifier
= {
4049 .notifier_call
= ip6_route_dev_notify
,
4050 .priority
= ADDRCONF_NOTIFY_PRIORITY
- 10,
4053 void __init
ip6_route_init_special_entries(void)
4055 /* Registering of the loopback is done before this portion of code,
4056 * the loopback reference in rt6_info will not be taken, do it
4057 * manually for init_net */
4058 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
4059 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
4060 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
4061 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
4062 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
4063 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
4064 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
4068 int __init
ip6_route_init(void)
4074 ip6_dst_ops_template
.kmem_cachep
=
4075 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
4076 SLAB_HWCACHE_ALIGN
, NULL
);
4077 if (!ip6_dst_ops_template
.kmem_cachep
)
4080 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
4082 goto out_kmem_cache
;
4084 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
4086 goto out_dst_entries
;
4088 ret
= register_pernet_subsys(&ip6_route_net_ops
);
4090 goto out_register_inetpeer
;
4092 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
4096 goto out_register_subsys
;
4102 ret
= fib6_rules_init();
4106 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
4108 goto fib6_rules_init
;
4111 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
4112 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
4113 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
4114 goto out_register_late_subsys
;
4116 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
4118 goto out_register_late_subsys
;
4120 for_each_possible_cpu(cpu
) {
4121 struct uncached_list
*ul
= per_cpu_ptr(&rt6_uncached_list
, cpu
);
4123 INIT_LIST_HEAD(&ul
->head
);
4124 spin_lock_init(&ul
->lock
);
4130 out_register_late_subsys
:
4131 unregister_pernet_subsys(&ip6_route_net_late_ops
);
4133 fib6_rules_cleanup();
4138 out_register_subsys
:
4139 unregister_pernet_subsys(&ip6_route_net_ops
);
4140 out_register_inetpeer
:
4141 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
4143 dst_entries_destroy(&ip6_dst_blackhole_ops
);
4145 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
4149 void ip6_route_cleanup(void)
4151 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
4152 unregister_pernet_subsys(&ip6_route_net_late_ops
);
4153 fib6_rules_cleanup();
4156 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
4157 unregister_pernet_subsys(&ip6_route_net_ops
);
4158 dst_entries_destroy(&ip6_dst_blackhole_ops
);
4159 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);