2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
16 * YOSHIFUJI Hideaki @USAGI
17 * reworked default router selection.
18 * - respect outgoing interface
19 * - select from (probably) reachable routers (i.e.
20 * routers in REACHABLE, STALE, DELAY or PROBE states).
21 * - always select the same router if it is (probably)
22 * reachable. otherwise, round-robin the list.
24 * Fixed routing subtrees.
27 #define pr_fmt(fmt) "IPv6: " fmt
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/export.h>
32 #include <linux/types.h>
33 #include <linux/times.h>
34 #include <linux/socket.h>
35 #include <linux/sockios.h>
36 #include <linux/net.h>
37 #include <linux/route.h>
38 #include <linux/netdevice.h>
39 #include <linux/in6.h>
40 #include <linux/mroute6.h>
41 #include <linux/init.h>
42 #include <linux/if_arp.h>
43 #include <linux/proc_fs.h>
44 #include <linux/seq_file.h>
45 #include <linux/nsproxy.h>
46 #include <linux/slab.h>
47 #include <net/net_namespace.h>
50 #include <net/ip6_fib.h>
51 #include <net/ip6_route.h>
52 #include <net/ndisc.h>
53 #include <net/addrconf.h>
55 #include <linux/rtnetlink.h>
58 #include <net/netevent.h>
59 #include <net/netlink.h>
60 #include <net/nexthop.h>
62 #include <asm/uaccess.h>
65 #include <linux/sysctl.h>
69 RT6_NUD_FAIL_HARD
= -2,
70 RT6_NUD_FAIL_SOFT
= -1,
74 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
75 const struct in6_addr
*dest
);
76 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
77 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
78 static unsigned int ip6_mtu(const struct dst_entry
*dst
);
79 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
80 static void ip6_dst_destroy(struct dst_entry
*);
81 static void ip6_dst_ifdown(struct dst_entry
*,
82 struct net_device
*dev
, int how
);
83 static int ip6_dst_gc(struct dst_ops
*ops
);
85 static int ip6_pkt_discard(struct sk_buff
*skb
);
86 static int ip6_pkt_discard_out(struct sk_buff
*skb
);
87 static void ip6_link_failure(struct sk_buff
*skb
);
88 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
89 struct sk_buff
*skb
, u32 mtu
);
90 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
92 static int rt6_score_route(struct rt6_info
*rt
, int oif
, int strict
);
94 #ifdef CONFIG_IPV6_ROUTE_INFO
95 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
96 const struct in6_addr
*prefix
, int prefixlen
,
97 const struct in6_addr
*gwaddr
, int ifindex
,
99 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
100 const struct in6_addr
*prefix
, int prefixlen
,
101 const struct in6_addr
*gwaddr
, int ifindex
);
104 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
106 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
107 struct inet_peer
*peer
;
110 if (!(rt
->dst
.flags
& DST_HOST
))
113 peer
= rt6_get_peer_create(rt
);
115 u32
*old_p
= __DST_METRICS_PTR(old
);
116 unsigned long prev
, new;
119 if (inet_metrics_new(peer
))
120 memcpy(p
, old_p
, sizeof(u32
) * RTAX_MAX
);
122 new = (unsigned long) p
;
123 prev
= cmpxchg(&dst
->_metrics
, old
, new);
126 p
= __DST_METRICS_PTR(prev
);
127 if (prev
& DST_METRICS_READ_ONLY
)
134 static inline const void *choose_neigh_daddr(struct rt6_info
*rt
,
138 struct in6_addr
*p
= &rt
->rt6i_gateway
;
140 if (!ipv6_addr_any(p
))
141 return (const void *) p
;
143 return &ipv6_hdr(skb
)->daddr
;
147 static struct neighbour
*ip6_neigh_lookup(const struct dst_entry
*dst
,
151 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
154 daddr
= choose_neigh_daddr(rt
, skb
, daddr
);
155 n
= __ipv6_neigh_lookup(dst
->dev
, daddr
);
158 return neigh_create(&nd_tbl
, daddr
, dst
->dev
);
161 static struct dst_ops ip6_dst_ops_template
= {
163 .protocol
= cpu_to_be16(ETH_P_IPV6
),
166 .check
= ip6_dst_check
,
167 .default_advmss
= ip6_default_advmss
,
169 .cow_metrics
= ipv6_cow_metrics
,
170 .destroy
= ip6_dst_destroy
,
171 .ifdown
= ip6_dst_ifdown
,
172 .negative_advice
= ip6_negative_advice
,
173 .link_failure
= ip6_link_failure
,
174 .update_pmtu
= ip6_rt_update_pmtu
,
175 .redirect
= rt6_do_redirect
,
176 .local_out
= __ip6_local_out
,
177 .neigh_lookup
= ip6_neigh_lookup
,
180 static unsigned int ip6_blackhole_mtu(const struct dst_entry
*dst
)
182 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
184 return mtu
? : dst
->dev
->mtu
;
187 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
188 struct sk_buff
*skb
, u32 mtu
)
192 static void ip6_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
197 static u32
*ip6_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
203 static struct dst_ops ip6_dst_blackhole_ops
= {
205 .protocol
= cpu_to_be16(ETH_P_IPV6
),
206 .destroy
= ip6_dst_destroy
,
207 .check
= ip6_dst_check
,
208 .mtu
= ip6_blackhole_mtu
,
209 .default_advmss
= ip6_default_advmss
,
210 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
211 .redirect
= ip6_rt_blackhole_redirect
,
212 .cow_metrics
= ip6_rt_blackhole_cow_metrics
,
213 .neigh_lookup
= ip6_neigh_lookup
,
216 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
217 [RTAX_HOPLIMIT
- 1] = 0,
220 static const struct rt6_info ip6_null_entry_template
= {
222 .__refcnt
= ATOMIC_INIT(1),
224 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
225 .error
= -ENETUNREACH
,
226 .input
= ip6_pkt_discard
,
227 .output
= ip6_pkt_discard_out
,
229 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
230 .rt6i_protocol
= RTPROT_KERNEL
,
231 .rt6i_metric
= ~(u32
) 0,
232 .rt6i_ref
= ATOMIC_INIT(1),
235 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
237 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
238 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
);
240 static const struct rt6_info ip6_prohibit_entry_template
= {
242 .__refcnt
= ATOMIC_INIT(1),
244 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
246 .input
= ip6_pkt_prohibit
,
247 .output
= ip6_pkt_prohibit_out
,
249 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
250 .rt6i_protocol
= RTPROT_KERNEL
,
251 .rt6i_metric
= ~(u32
) 0,
252 .rt6i_ref
= ATOMIC_INIT(1),
255 static const struct rt6_info ip6_blk_hole_entry_template
= {
257 .__refcnt
= ATOMIC_INIT(1),
259 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
261 .input
= dst_discard
,
262 .output
= dst_discard
,
264 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
265 .rt6i_protocol
= RTPROT_KERNEL
,
266 .rt6i_metric
= ~(u32
) 0,
267 .rt6i_ref
= ATOMIC_INIT(1),
272 /* allocate dst with ip6_dst_ops */
273 static inline struct rt6_info
*ip6_dst_alloc(struct net
*net
,
274 struct net_device
*dev
,
276 struct fib6_table
*table
)
278 struct rt6_info
*rt
= dst_alloc(&net
->ipv6
.ip6_dst_ops
, dev
,
279 0, DST_OBSOLETE_FORCE_CHK
, flags
);
282 struct dst_entry
*dst
= &rt
->dst
;
284 memset(dst
+ 1, 0, sizeof(*rt
) - sizeof(*dst
));
285 rt6_init_peer(rt
, table
? &table
->tb6_peers
: net
->ipv6
.peers
);
286 rt
->rt6i_genid
= rt_genid_ipv6(net
);
287 INIT_LIST_HEAD(&rt
->rt6i_siblings
);
292 static void ip6_dst_destroy(struct dst_entry
*dst
)
294 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
295 struct inet6_dev
*idev
= rt
->rt6i_idev
;
296 struct dst_entry
*from
= dst
->from
;
298 if (!(rt
->dst
.flags
& DST_HOST
))
299 dst_destroy_metrics_generic(dst
);
302 rt
->rt6i_idev
= NULL
;
309 if (rt6_has_peer(rt
)) {
310 struct inet_peer
*peer
= rt6_peer_ptr(rt
);
315 void rt6_bind_peer(struct rt6_info
*rt
, int create
)
317 struct inet_peer_base
*base
;
318 struct inet_peer
*peer
;
320 base
= inetpeer_base_ptr(rt
->_rt6i_peer
);
324 peer
= inet_getpeer_v6(base
, &rt
->rt6i_dst
.addr
, create
);
326 if (!rt6_set_peer(rt
, peer
))
331 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
334 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
335 struct inet6_dev
*idev
= rt
->rt6i_idev
;
336 struct net_device
*loopback_dev
=
337 dev_net(dev
)->loopback_dev
;
339 if (dev
!= loopback_dev
) {
340 if (idev
&& idev
->dev
== dev
) {
341 struct inet6_dev
*loopback_idev
=
342 in6_dev_get(loopback_dev
);
344 rt
->rt6i_idev
= loopback_idev
;
351 static bool rt6_check_expired(const struct rt6_info
*rt
)
353 if (rt
->rt6i_flags
& RTF_EXPIRES
) {
354 if (time_after(jiffies
, rt
->dst
.expires
))
356 } else if (rt
->dst
.from
) {
357 return rt6_check_expired((struct rt6_info
*) rt
->dst
.from
);
362 static bool rt6_need_strict(const struct in6_addr
*daddr
)
364 return ipv6_addr_type(daddr
) &
365 (IPV6_ADDR_MULTICAST
| IPV6_ADDR_LINKLOCAL
| IPV6_ADDR_LOOPBACK
);
368 /* Multipath route selection:
369 * Hash based function using packet header and flowlabel.
370 * Adapted from fib_info_hashfn()
372 static int rt6_info_hash_nhsfn(unsigned int candidate_count
,
373 const struct flowi6
*fl6
)
375 unsigned int val
= fl6
->flowi6_proto
;
377 val
^= ipv6_addr_hash(&fl6
->daddr
);
378 val
^= ipv6_addr_hash(&fl6
->saddr
);
380 /* Work only if this not encapsulated */
381 switch (fl6
->flowi6_proto
) {
385 val
^= (__force u16
)fl6
->fl6_sport
;
386 val
^= (__force u16
)fl6
->fl6_dport
;
390 val
^= (__force u16
)fl6
->fl6_icmp_type
;
391 val
^= (__force u16
)fl6
->fl6_icmp_code
;
394 /* RFC6438 recommands to use flowlabel */
395 val
^= (__force u32
)fl6
->flowlabel
;
397 /* Perhaps, we need to tune, this function? */
398 val
= val
^ (val
>> 7) ^ (val
>> 12);
399 return val
% candidate_count
;
402 static struct rt6_info
*rt6_multipath_select(struct rt6_info
*match
,
403 struct flowi6
*fl6
, int oif
,
406 struct rt6_info
*sibling
, *next_sibling
;
409 route_choosen
= rt6_info_hash_nhsfn(match
->rt6i_nsiblings
+ 1, fl6
);
410 /* Don't change the route, if route_choosen == 0
411 * (siblings does not include ourself)
414 list_for_each_entry_safe(sibling
, next_sibling
,
415 &match
->rt6i_siblings
, rt6i_siblings
) {
417 if (route_choosen
== 0) {
418 if (rt6_score_route(sibling
, oif
, strict
) < 0)
428 * Route lookup. Any table->tb6_lock is implied.
431 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
433 const struct in6_addr
*saddr
,
437 struct rt6_info
*local
= NULL
;
438 struct rt6_info
*sprt
;
440 if (!oif
&& ipv6_addr_any(saddr
))
443 for (sprt
= rt
; sprt
; sprt
= sprt
->dst
.rt6_next
) {
444 struct net_device
*dev
= sprt
->dst
.dev
;
447 if (dev
->ifindex
== oif
)
449 if (dev
->flags
& IFF_LOOPBACK
) {
450 if (!sprt
->rt6i_idev
||
451 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
452 if (flags
& RT6_LOOKUP_F_IFACE
&& oif
)
454 if (local
&& (!oif
||
455 local
->rt6i_idev
->dev
->ifindex
== oif
))
461 if (ipv6_chk_addr(net
, saddr
, dev
,
462 flags
& RT6_LOOKUP_F_IFACE
))
471 if (flags
& RT6_LOOKUP_F_IFACE
)
472 return net
->ipv6
.ip6_null_entry
;
478 #ifdef CONFIG_IPV6_ROUTER_PREF
479 struct __rt6_probe_work
{
480 struct work_struct work
;
481 struct in6_addr target
;
482 struct net_device
*dev
;
485 static void rt6_probe_deferred(struct work_struct
*w
)
487 struct in6_addr mcaddr
;
488 struct __rt6_probe_work
*work
=
489 container_of(w
, struct __rt6_probe_work
, work
);
491 addrconf_addr_solict_mult(&work
->target
, &mcaddr
);
492 ndisc_send_ns(work
->dev
, NULL
, &work
->target
, &mcaddr
, NULL
);
497 static void rt6_probe(struct rt6_info
*rt
)
499 struct neighbour
*neigh
;
501 * Okay, this does not seem to be appropriate
502 * for now, however, we need to check if it
503 * is really so; aka Router Reachability Probing.
505 * Router Reachability Probe MUST be rate-limited
506 * to no more than one per minute.
508 if (!rt
|| !(rt
->rt6i_flags
& RTF_GATEWAY
))
511 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
513 write_lock(&neigh
->lock
);
514 if (neigh
->nud_state
& NUD_VALID
)
519 time_after(jiffies
, neigh
->updated
+ rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
520 struct __rt6_probe_work
*work
;
522 work
= kmalloc(sizeof(*work
), GFP_ATOMIC
);
525 neigh
->updated
= jiffies
;
528 write_unlock(&neigh
->lock
);
531 INIT_WORK(&work
->work
, rt6_probe_deferred
);
532 work
->target
= rt
->rt6i_gateway
;
533 dev_hold(rt
->dst
.dev
);
534 work
->dev
= rt
->dst
.dev
;
535 schedule_work(&work
->work
);
539 write_unlock(&neigh
->lock
);
541 rcu_read_unlock_bh();
544 static inline void rt6_probe(struct rt6_info
*rt
)
550 * Default Router Selection (RFC 2461 6.3.6)
552 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
554 struct net_device
*dev
= rt
->dst
.dev
;
555 if (!oif
|| dev
->ifindex
== oif
)
557 if ((dev
->flags
& IFF_LOOPBACK
) &&
558 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
563 static inline enum rt6_nud_state
rt6_check_neigh(struct rt6_info
*rt
)
565 struct neighbour
*neigh
;
566 enum rt6_nud_state ret
= RT6_NUD_FAIL_HARD
;
568 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
569 !(rt
->rt6i_flags
& RTF_GATEWAY
))
570 return RT6_NUD_SUCCEED
;
573 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
575 read_lock(&neigh
->lock
);
576 if (neigh
->nud_state
& NUD_VALID
)
577 ret
= RT6_NUD_SUCCEED
;
578 #ifdef CONFIG_IPV6_ROUTER_PREF
579 else if (!(neigh
->nud_state
& NUD_FAILED
))
580 ret
= RT6_NUD_SUCCEED
;
582 read_unlock(&neigh
->lock
);
584 ret
= IS_ENABLED(CONFIG_IPV6_ROUTER_PREF
) ?
585 RT6_NUD_SUCCEED
: RT6_NUD_FAIL_SOFT
;
587 rcu_read_unlock_bh();
592 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
597 m
= rt6_check_dev(rt
, oif
);
598 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
599 return RT6_NUD_FAIL_HARD
;
600 #ifdef CONFIG_IPV6_ROUTER_PREF
601 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
603 if (strict
& RT6_LOOKUP_F_REACHABLE
) {
604 int n
= rt6_check_neigh(rt
);
611 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
612 int *mpri
, struct rt6_info
*match
,
616 bool match_do_rr
= false;
618 if (rt6_check_expired(rt
))
621 m
= rt6_score_route(rt
, oif
, strict
);
622 if (m
== RT6_NUD_FAIL_SOFT
) {
624 m
= 0; /* lowest valid score */
629 if (strict
& RT6_LOOKUP_F_REACHABLE
)
633 *do_rr
= match_do_rr
;
641 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
642 struct rt6_info
*rr_head
,
643 u32 metric
, int oif
, int strict
,
646 struct rt6_info
*rt
, *match
;
650 for (rt
= rr_head
; rt
&& rt
->rt6i_metric
== metric
;
651 rt
= rt
->dst
.rt6_next
)
652 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
653 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
&& rt
->rt6i_metric
== metric
;
654 rt
= rt
->dst
.rt6_next
)
655 match
= find_match(rt
, oif
, strict
, &mpri
, match
, do_rr
);
660 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
662 struct rt6_info
*match
, *rt0
;
668 fn
->rr_ptr
= rt0
= fn
->leaf
;
670 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
,
674 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
676 /* no entries matched; do round-robin */
677 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
684 net
= dev_net(rt0
->dst
.dev
);
685 return match
? match
: net
->ipv6
.ip6_null_entry
;
688 #ifdef CONFIG_IPV6_ROUTE_INFO
689 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
690 const struct in6_addr
*gwaddr
)
692 struct net
*net
= dev_net(dev
);
693 struct route_info
*rinfo
= (struct route_info
*) opt
;
694 struct in6_addr prefix_buf
, *prefix
;
696 unsigned long lifetime
;
699 if (len
< sizeof(struct route_info
)) {
703 /* Sanity check for prefix_len and length */
704 if (rinfo
->length
> 3) {
706 } else if (rinfo
->prefix_len
> 128) {
708 } else if (rinfo
->prefix_len
> 64) {
709 if (rinfo
->length
< 2) {
712 } else if (rinfo
->prefix_len
> 0) {
713 if (rinfo
->length
< 1) {
718 pref
= rinfo
->route_pref
;
719 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
722 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
724 if (rinfo
->length
== 3)
725 prefix
= (struct in6_addr
*)rinfo
->prefix
;
727 /* this function is safe */
728 ipv6_addr_prefix(&prefix_buf
,
729 (struct in6_addr
*)rinfo
->prefix
,
731 prefix
= &prefix_buf
;
734 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
,
737 if (rt
&& !lifetime
) {
743 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
746 rt
->rt6i_flags
= RTF_ROUTEINFO
|
747 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
750 if (!addrconf_finite_timeout(lifetime
))
751 rt6_clean_expires(rt
);
753 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
761 #define BACKTRACK(__net, saddr) \
763 if (rt == __net->ipv6.ip6_null_entry) { \
764 struct fib6_node *pn; \
766 if (fn->fn_flags & RTN_TL_ROOT) \
769 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
770 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
773 if (fn->fn_flags & RTN_RTINFO) \
779 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
780 struct fib6_table
*table
,
781 struct flowi6
*fl6
, int flags
)
783 struct fib6_node
*fn
;
786 read_lock_bh(&table
->tb6_lock
);
787 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
790 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
, fl6
->flowi6_oif
, flags
);
791 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
792 rt
= rt6_multipath_select(rt
, fl6
, fl6
->flowi6_oif
, flags
);
793 BACKTRACK(net
, &fl6
->saddr
);
795 dst_use(&rt
->dst
, jiffies
);
796 read_unlock_bh(&table
->tb6_lock
);
801 struct dst_entry
* ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
804 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
806 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
808 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
809 const struct in6_addr
*saddr
, int oif
, int strict
)
811 struct flowi6 fl6
= {
815 struct dst_entry
*dst
;
816 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
819 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
820 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
823 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
825 return (struct rt6_info
*) dst
;
832 EXPORT_SYMBOL(rt6_lookup
);
834 /* ip6_ins_rt is called with FREE table->tb6_lock.
835 It takes new route entry, the addition fails by any reason the
836 route is freed. In any case, if caller does not hold it, it may
840 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
)
843 struct fib6_table
*table
;
845 table
= rt
->rt6i_table
;
846 write_lock_bh(&table
->tb6_lock
);
847 err
= fib6_add(&table
->tb6_root
, rt
, info
);
848 write_unlock_bh(&table
->tb6_lock
);
853 int ip6_ins_rt(struct rt6_info
*rt
)
855 struct nl_info info
= {
856 .nl_net
= dev_net(rt
->dst
.dev
),
858 return __ip6_ins_rt(rt
, &info
);
861 static struct rt6_info
*rt6_alloc_cow(struct rt6_info
*ort
,
862 const struct in6_addr
*daddr
,
863 const struct in6_addr
*saddr
)
871 rt
= ip6_rt_copy(ort
, daddr
);
874 if (ort
->rt6i_dst
.plen
!= 128 &&
875 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
876 rt
->rt6i_flags
|= RTF_ANYCAST
;
878 rt
->rt6i_flags
|= RTF_CACHE
;
880 #ifdef CONFIG_IPV6_SUBTREES
881 if (rt
->rt6i_src
.plen
&& saddr
) {
882 rt
->rt6i_src
.addr
= *saddr
;
883 rt
->rt6i_src
.plen
= 128;
891 static struct rt6_info
*rt6_alloc_clone(struct rt6_info
*ort
,
892 const struct in6_addr
*daddr
)
894 struct rt6_info
*rt
= ip6_rt_copy(ort
, daddr
);
897 rt
->rt6i_flags
|= RTF_CACHE
;
901 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
902 struct flowi6
*fl6
, int flags
)
904 struct fib6_node
*fn
;
905 struct rt6_info
*rt
, *nrt
;
909 int reachable
= net
->ipv6
.devconf_all
->forwarding
? 0 : RT6_LOOKUP_F_REACHABLE
;
911 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
914 read_lock_bh(&table
->tb6_lock
);
917 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
920 rt
= rt6_select(fn
, oif
, strict
| reachable
);
921 if (rt
->rt6i_nsiblings
)
922 rt
= rt6_multipath_select(rt
, fl6
, oif
, strict
| reachable
);
923 BACKTRACK(net
, &fl6
->saddr
);
924 if (rt
== net
->ipv6
.ip6_null_entry
||
925 rt
->rt6i_flags
& RTF_CACHE
)
929 read_unlock_bh(&table
->tb6_lock
);
931 if (!(rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
)))
932 nrt
= rt6_alloc_cow(rt
, &fl6
->daddr
, &fl6
->saddr
);
933 else if (!(rt
->dst
.flags
& DST_HOST
))
934 nrt
= rt6_alloc_clone(rt
, &fl6
->daddr
);
939 rt
= nrt
? : net
->ipv6
.ip6_null_entry
;
943 err
= ip6_ins_rt(nrt
);
952 * Race condition! In the gap, when table->tb6_lock was
953 * released someone could insert this route. Relookup.
964 read_unlock_bh(&table
->tb6_lock
);
966 rt
->dst
.lastuse
= jiffies
;
972 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
973 struct flowi6
*fl6
, int flags
)
975 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
978 static struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
979 struct net_device
*dev
,
980 struct flowi6
*fl6
, int flags
)
982 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
983 flags
|= RT6_LOOKUP_F_IFACE
;
985 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
988 void ip6_route_input(struct sk_buff
*skb
)
990 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
991 struct net
*net
= dev_net(skb
->dev
);
992 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
993 struct flowi6 fl6
= {
994 .flowi6_iif
= skb
->dev
->ifindex
,
997 .flowlabel
= ip6_flowinfo(iph
),
998 .flowi6_mark
= skb
->mark
,
999 .flowi6_proto
= iph
->nexthdr
,
1002 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
1005 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
1006 struct flowi6
*fl6
, int flags
)
1008 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1011 struct dst_entry
* ip6_route_output(struct net
*net
, const struct sock
*sk
,
1016 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1018 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
))
1019 flags
|= RT6_LOOKUP_F_IFACE
;
1021 if (!ipv6_addr_any(&fl6
->saddr
))
1022 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1024 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1026 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1029 EXPORT_SYMBOL(ip6_route_output
);
1031 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1033 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1034 struct dst_entry
*new = NULL
;
1036 rt
= dst_alloc(&ip6_dst_blackhole_ops
, ort
->dst
.dev
, 1, DST_OBSOLETE_NONE
, 0);
1040 memset(new + 1, 0, sizeof(*rt
) - sizeof(*new));
1041 rt6_init_peer(rt
, net
->ipv6
.peers
);
1044 new->input
= dst_discard
;
1045 new->output
= dst_discard
;
1047 if (dst_metrics_read_only(&ort
->dst
))
1048 new->_metrics
= ort
->dst
._metrics
;
1050 dst_copy_metrics(new, &ort
->dst
);
1051 rt
->rt6i_idev
= ort
->rt6i_idev
;
1053 in6_dev_hold(rt
->rt6i_idev
);
1055 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1056 rt
->rt6i_flags
= ort
->rt6i_flags
;
1057 rt
->rt6i_metric
= 0;
1059 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1060 #ifdef CONFIG_IPV6_SUBTREES
1061 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1067 dst_release(dst_orig
);
1068 return new ? new : ERR_PTR(-ENOMEM
);
1072 * Destination cache support functions
1075 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1077 struct rt6_info
*rt
;
1079 rt
= (struct rt6_info
*) dst
;
1081 /* All IPV6 dsts are created with ->obsolete set to the value
1082 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1083 * into this function always.
1085 if (rt
->rt6i_genid
!= rt_genid_ipv6(dev_net(rt
->dst
.dev
)))
1088 if (!rt
->rt6i_node
|| (rt
->rt6i_node
->fn_sernum
!= cookie
))
1091 if (rt6_check_expired(rt
))
1097 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1099 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1102 if (rt
->rt6i_flags
& RTF_CACHE
) {
1103 if (rt6_check_expired(rt
)) {
1115 static void ip6_link_failure(struct sk_buff
*skb
)
1117 struct rt6_info
*rt
;
1119 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1121 rt
= (struct rt6_info
*) skb_dst(skb
);
1123 if (rt
->rt6i_flags
& RTF_CACHE
) {
1127 } else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
)) {
1128 rt
->rt6i_node
->fn_sernum
= -1;
1133 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1134 struct sk_buff
*skb
, u32 mtu
)
1136 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1139 if (mtu
< dst_mtu(dst
) && rt6
->rt6i_dst
.plen
== 128) {
1140 struct net
*net
= dev_net(dst
->dev
);
1142 rt6
->rt6i_flags
|= RTF_MODIFIED
;
1143 if (mtu
< IPV6_MIN_MTU
) {
1144 u32 features
= dst_metric(dst
, RTAX_FEATURES
);
1146 features
|= RTAX_FEATURE_ALLFRAG
;
1147 dst_metric_set(dst
, RTAX_FEATURES
, features
);
1149 dst_metric_set(dst
, RTAX_MTU
, mtu
);
1150 rt6_update_expires(rt6
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1154 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1157 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1158 struct dst_entry
*dst
;
1161 memset(&fl6
, 0, sizeof(fl6
));
1162 fl6
.flowi6_oif
= oif
;
1163 fl6
.flowi6_mark
= mark
;
1164 fl6
.daddr
= iph
->daddr
;
1165 fl6
.saddr
= iph
->saddr
;
1166 fl6
.flowlabel
= ip6_flowinfo(iph
);
1168 dst
= ip6_route_output(net
, NULL
, &fl6
);
1170 ip6_rt_update_pmtu(dst
, NULL
, skb
, ntohl(mtu
));
1173 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1175 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1177 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1178 sk
->sk_bound_dev_if
, sk
->sk_mark
);
1180 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1182 /* Handle redirects */
1183 struct ip6rd_flowi
{
1185 struct in6_addr gateway
;
1188 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
1189 struct fib6_table
*table
,
1193 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl6
;
1194 struct rt6_info
*rt
;
1195 struct fib6_node
*fn
;
1197 /* Get the "current" route for this destination and
1198 * check if the redirect has come from approriate router.
1200 * RFC 4861 specifies that redirects should only be
1201 * accepted if they come from the nexthop to the target.
1202 * Due to the way the routes are chosen, this notion
1203 * is a bit fuzzy and one might need to check all possible
1207 read_lock_bh(&table
->tb6_lock
);
1208 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
1210 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1211 if (rt6_check_expired(rt
))
1215 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
1217 if (fl6
->flowi6_oif
!= rt
->dst
.dev
->ifindex
)
1219 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
))
1225 rt
= net
->ipv6
.ip6_null_entry
;
1226 else if (rt
->dst
.error
) {
1227 rt
= net
->ipv6
.ip6_null_entry
;
1230 BACKTRACK(net
, &fl6
->saddr
);
1234 read_unlock_bh(&table
->tb6_lock
);
1239 static struct dst_entry
*ip6_route_redirect(struct net
*net
,
1240 const struct flowi6
*fl6
,
1241 const struct in6_addr
*gateway
)
1243 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1244 struct ip6rd_flowi rdfl
;
1247 rdfl
.gateway
= *gateway
;
1249 return fib6_rule_lookup(net
, &rdfl
.fl6
,
1250 flags
, __ip6_route_redirect
);
1253 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
)
1255 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1256 struct dst_entry
*dst
;
1259 memset(&fl6
, 0, sizeof(fl6
));
1260 fl6
.flowi6_oif
= oif
;
1261 fl6
.flowi6_mark
= mark
;
1262 fl6
.daddr
= iph
->daddr
;
1263 fl6
.saddr
= iph
->saddr
;
1264 fl6
.flowlabel
= ip6_flowinfo(iph
);
1266 dst
= ip6_route_redirect(net
, &fl6
, &ipv6_hdr(skb
)->saddr
);
1267 rt6_do_redirect(dst
, NULL
, skb
);
1270 EXPORT_SYMBOL_GPL(ip6_redirect
);
1272 void ip6_redirect_no_header(struct sk_buff
*skb
, struct net
*net
, int oif
,
1275 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
1276 const struct rd_msg
*msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1277 struct dst_entry
*dst
;
1280 memset(&fl6
, 0, sizeof(fl6
));
1281 fl6
.flowi6_oif
= oif
;
1282 fl6
.flowi6_mark
= mark
;
1283 fl6
.daddr
= msg
->dest
;
1284 fl6
.saddr
= iph
->daddr
;
1286 dst
= ip6_route_redirect(net
, &fl6
, &iph
->saddr
);
1287 rt6_do_redirect(dst
, NULL
, skb
);
1291 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1293 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
);
1295 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1297 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1299 struct net_device
*dev
= dst
->dev
;
1300 unsigned int mtu
= dst_mtu(dst
);
1301 struct net
*net
= dev_net(dev
);
1303 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1305 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1306 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1309 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1310 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1311 * IPV6_MAXPLEN is also valid and means: "any MSS,
1312 * rely only on pmtu discovery"
1314 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1319 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1321 struct inet6_dev
*idev
;
1322 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1330 idev
= __in6_dev_get(dst
->dev
);
1332 mtu
= idev
->cnf
.mtu6
;
1338 static struct dst_entry
*icmp6_dst_gc_list
;
1339 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1341 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1344 struct dst_entry
*dst
;
1345 struct rt6_info
*rt
;
1346 struct inet6_dev
*idev
= in6_dev_get(dev
);
1347 struct net
*net
= dev_net(dev
);
1349 if (unlikely(!idev
))
1350 return ERR_PTR(-ENODEV
);
1352 rt
= ip6_dst_alloc(net
, dev
, 0, NULL
);
1353 if (unlikely(!rt
)) {
1355 dst
= ERR_PTR(-ENOMEM
);
1359 rt
->dst
.flags
|= DST_HOST
;
1360 rt
->dst
.output
= ip6_output
;
1361 atomic_set(&rt
->dst
.__refcnt
, 1);
1362 rt
->rt6i_gateway
= fl6
->daddr
;
1363 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1364 rt
->rt6i_dst
.plen
= 128;
1365 rt
->rt6i_idev
= idev
;
1366 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1368 spin_lock_bh(&icmp6_dst_lock
);
1369 rt
->dst
.next
= icmp6_dst_gc_list
;
1370 icmp6_dst_gc_list
= &rt
->dst
;
1371 spin_unlock_bh(&icmp6_dst_lock
);
1373 fib6_force_start_gc(net
);
1375 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1381 int icmp6_dst_gc(void)
1383 struct dst_entry
*dst
, **pprev
;
1386 spin_lock_bh(&icmp6_dst_lock
);
1387 pprev
= &icmp6_dst_gc_list
;
1389 while ((dst
= *pprev
) != NULL
) {
1390 if (!atomic_read(&dst
->__refcnt
)) {
1399 spin_unlock_bh(&icmp6_dst_lock
);
1404 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1407 struct dst_entry
*dst
, **pprev
;
1409 spin_lock_bh(&icmp6_dst_lock
);
1410 pprev
= &icmp6_dst_gc_list
;
1411 while ((dst
= *pprev
) != NULL
) {
1412 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1413 if (func(rt
, arg
)) {
1420 spin_unlock_bh(&icmp6_dst_lock
);
1423 static int ip6_dst_gc(struct dst_ops
*ops
)
1425 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1426 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1427 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1428 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1429 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1430 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1433 entries
= dst_entries_get_fast(ops
);
1434 if (time_after(rt_last_gc
+ rt_min_interval
, jiffies
) &&
1435 entries
<= rt_max_size
)
1438 net
->ipv6
.ip6_rt_gc_expire
++;
1439 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
, entries
> rt_max_size
);
1440 entries
= dst_entries_get_slow(ops
);
1441 if (entries
< ops
->gc_thresh
)
1442 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1444 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1445 return entries
> rt_max_size
;
1452 int ip6_route_add(struct fib6_config
*cfg
)
1455 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1456 struct rt6_info
*rt
= NULL
;
1457 struct net_device
*dev
= NULL
;
1458 struct inet6_dev
*idev
= NULL
;
1459 struct fib6_table
*table
;
1462 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1464 #ifndef CONFIG_IPV6_SUBTREES
1465 if (cfg
->fc_src_len
)
1468 if (cfg
->fc_ifindex
) {
1470 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1473 idev
= in6_dev_get(dev
);
1478 if (cfg
->fc_metric
== 0)
1479 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1482 if (cfg
->fc_nlinfo
.nlh
&&
1483 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1484 table
= fib6_get_table(net
, cfg
->fc_table
);
1486 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1487 table
= fib6_new_table(net
, cfg
->fc_table
);
1490 table
= fib6_new_table(net
, cfg
->fc_table
);
1496 rt
= ip6_dst_alloc(net
, NULL
, DST_NOCOUNT
, table
);
1503 if (cfg
->fc_flags
& RTF_EXPIRES
)
1504 rt6_set_expires(rt
, jiffies
+
1505 clock_t_to_jiffies(cfg
->fc_expires
));
1507 rt6_clean_expires(rt
);
1509 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1510 cfg
->fc_protocol
= RTPROT_BOOT
;
1511 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1513 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1515 if (addr_type
& IPV6_ADDR_MULTICAST
)
1516 rt
->dst
.input
= ip6_mc_input
;
1517 else if (cfg
->fc_flags
& RTF_LOCAL
)
1518 rt
->dst
.input
= ip6_input
;
1520 rt
->dst
.input
= ip6_forward
;
1522 rt
->dst
.output
= ip6_output
;
1524 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1525 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1526 if (rt
->rt6i_dst
.plen
== 128)
1527 rt
->dst
.flags
|= DST_HOST
;
1529 if (!(rt
->dst
.flags
& DST_HOST
) && cfg
->fc_mx
) {
1530 u32
*metrics
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_KERNEL
);
1535 dst_init_metrics(&rt
->dst
, metrics
, 0);
1537 #ifdef CONFIG_IPV6_SUBTREES
1538 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1539 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1542 rt
->rt6i_metric
= cfg
->fc_metric
;
1544 /* We cannot add true routes via loopback here,
1545 they would result in kernel looping; promote them to reject routes
1547 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1548 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1549 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1550 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1551 /* hold loopback dev/idev if we haven't done so. */
1552 if (dev
!= net
->loopback_dev
) {
1557 dev
= net
->loopback_dev
;
1559 idev
= in6_dev_get(dev
);
1565 rt
->dst
.output
= ip6_pkt_discard_out
;
1566 rt
->dst
.input
= ip6_pkt_discard
;
1567 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1568 switch (cfg
->fc_type
) {
1570 rt
->dst
.error
= -EINVAL
;
1573 rt
->dst
.error
= -EACCES
;
1576 rt
->dst
.error
= -EAGAIN
;
1579 rt
->dst
.error
= -ENETUNREACH
;
1585 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1586 const struct in6_addr
*gw_addr
;
1589 gw_addr
= &cfg
->fc_gateway
;
1590 rt
->rt6i_gateway
= *gw_addr
;
1591 gwa_type
= ipv6_addr_type(gw_addr
);
1593 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1594 struct rt6_info
*grt
;
1596 /* IPv6 strictly inhibits using not link-local
1597 addresses as nexthop address.
1598 Otherwise, router will not able to send redirects.
1599 It is very good, but in some (rare!) circumstances
1600 (SIT, PtP, NBMA NOARP links) it is handy to allow
1601 some exceptions. --ANK
1604 if (!(gwa_type
& IPV6_ADDR_UNICAST
))
1607 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1609 err
= -EHOSTUNREACH
;
1613 if (dev
!= grt
->dst
.dev
) {
1619 idev
= grt
->rt6i_idev
;
1621 in6_dev_hold(grt
->rt6i_idev
);
1623 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
1631 if (!dev
|| (dev
->flags
& IFF_LOOPBACK
))
1639 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
1640 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
1644 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
1645 rt
->rt6i_prefsrc
.plen
= 128;
1647 rt
->rt6i_prefsrc
.plen
= 0;
1649 rt
->rt6i_flags
= cfg
->fc_flags
;
1656 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1657 int type
= nla_type(nla
);
1660 if (type
> RTAX_MAX
) {
1665 dst_metric_set(&rt
->dst
, type
, nla_get_u32(nla
));
1671 rt
->rt6i_idev
= idev
;
1672 rt
->rt6i_table
= table
;
1674 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1676 return __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
);
1688 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1691 struct fib6_table
*table
;
1692 struct net
*net
= dev_net(rt
->dst
.dev
);
1694 if (rt
== net
->ipv6
.ip6_null_entry
) {
1699 table
= rt
->rt6i_table
;
1700 write_lock_bh(&table
->tb6_lock
);
1701 err
= fib6_del(rt
, info
);
1702 write_unlock_bh(&table
->tb6_lock
);
1709 int ip6_del_rt(struct rt6_info
*rt
)
1711 struct nl_info info
= {
1712 .nl_net
= dev_net(rt
->dst
.dev
),
1714 return __ip6_del_rt(rt
, &info
);
1717 static int ip6_route_del(struct fib6_config
*cfg
)
1719 struct fib6_table
*table
;
1720 struct fib6_node
*fn
;
1721 struct rt6_info
*rt
;
1724 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1728 read_lock_bh(&table
->tb6_lock
);
1730 fn
= fib6_locate(&table
->tb6_root
,
1731 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1732 &cfg
->fc_src
, cfg
->fc_src_len
);
1735 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1736 if (cfg
->fc_ifindex
&&
1738 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
1740 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1741 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1743 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1746 read_unlock_bh(&table
->tb6_lock
);
1748 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1751 read_unlock_bh(&table
->tb6_lock
);
1756 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
1758 struct net
*net
= dev_net(skb
->dev
);
1759 struct netevent_redirect netevent
;
1760 struct rt6_info
*rt
, *nrt
= NULL
;
1761 struct ndisc_options ndopts
;
1762 struct inet6_dev
*in6_dev
;
1763 struct neighbour
*neigh
;
1765 int optlen
, on_link
;
1768 optlen
= skb_tail_pointer(skb
) - skb_transport_header(skb
);
1769 optlen
-= sizeof(*msg
);
1772 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1776 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1778 if (ipv6_addr_is_multicast(&msg
->dest
)) {
1779 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1784 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
1786 } else if (ipv6_addr_type(&msg
->target
) !=
1787 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
1788 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1792 in6_dev
= __in6_dev_get(skb
->dev
);
1795 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
1799 * The IP source address of the Redirect MUST be the same as the current
1800 * first-hop router for the specified ICMP Destination Address.
1803 if (!ndisc_parse_options(msg
->opt
, optlen
, &ndopts
)) {
1804 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1809 if (ndopts
.nd_opts_tgt_lladdr
) {
1810 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
1813 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1818 rt
= (struct rt6_info
*) dst
;
1819 if (rt
== net
->ipv6
.ip6_null_entry
) {
1820 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1824 /* Redirect received -> path was valid.
1825 * Look, redirects are sent only in response to data packets,
1826 * so that this nexthop apparently is reachable. --ANK
1828 dst_confirm(&rt
->dst
);
1830 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
1835 * We have finally decided to accept it.
1838 neigh_update(neigh
, lladdr
, NUD_STALE
,
1839 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
1840 NEIGH_UPDATE_F_OVERRIDE
|
1841 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
1842 NEIGH_UPDATE_F_ISROUTER
))
1845 nrt
= ip6_rt_copy(rt
, &msg
->dest
);
1849 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
1851 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
1853 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
1855 if (ip6_ins_rt(nrt
))
1858 netevent
.old
= &rt
->dst
;
1859 netevent
.new = &nrt
->dst
;
1860 netevent
.daddr
= &msg
->dest
;
1861 netevent
.neigh
= neigh
;
1862 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
1864 if (rt
->rt6i_flags
& RTF_CACHE
) {
1865 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
1870 neigh_release(neigh
);
1874 * Misc support functions
1877 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
1878 const struct in6_addr
*dest
)
1880 struct net
*net
= dev_net(ort
->dst
.dev
);
1881 struct rt6_info
*rt
= ip6_dst_alloc(net
, ort
->dst
.dev
, 0,
1885 rt
->dst
.input
= ort
->dst
.input
;
1886 rt
->dst
.output
= ort
->dst
.output
;
1887 rt
->dst
.flags
|= DST_HOST
;
1889 rt
->rt6i_dst
.addr
= *dest
;
1890 rt
->rt6i_dst
.plen
= 128;
1891 dst_copy_metrics(&rt
->dst
, &ort
->dst
);
1892 rt
->dst
.error
= ort
->dst
.error
;
1893 rt
->rt6i_idev
= ort
->rt6i_idev
;
1895 in6_dev_hold(rt
->rt6i_idev
);
1896 rt
->dst
.lastuse
= jiffies
;
1898 if (ort
->rt6i_flags
& RTF_GATEWAY
)
1899 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1901 rt
->rt6i_gateway
= *dest
;
1902 rt
->rt6i_flags
= ort
->rt6i_flags
;
1903 if ((ort
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) ==
1904 (RTF_DEFAULT
| RTF_ADDRCONF
))
1905 rt6_set_from(rt
, ort
);
1906 rt
->rt6i_metric
= 0;
1908 #ifdef CONFIG_IPV6_SUBTREES
1909 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1911 memcpy(&rt
->rt6i_prefsrc
, &ort
->rt6i_prefsrc
, sizeof(struct rt6key
));
1912 rt
->rt6i_table
= ort
->rt6i_table
;
1917 #ifdef CONFIG_IPV6_ROUTE_INFO
1918 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
1919 const struct in6_addr
*prefix
, int prefixlen
,
1920 const struct in6_addr
*gwaddr
, int ifindex
)
1922 struct fib6_node
*fn
;
1923 struct rt6_info
*rt
= NULL
;
1924 struct fib6_table
*table
;
1926 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
1930 read_lock_bh(&table
->tb6_lock
);
1931 fn
= fib6_locate(&table
->tb6_root
, prefix
,prefixlen
, NULL
, 0);
1935 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1936 if (rt
->dst
.dev
->ifindex
!= ifindex
)
1938 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
1940 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
1946 read_unlock_bh(&table
->tb6_lock
);
1950 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
1951 const struct in6_addr
*prefix
, int prefixlen
,
1952 const struct in6_addr
*gwaddr
, int ifindex
,
1955 struct fib6_config cfg
= {
1956 .fc_table
= RT6_TABLE_INFO
,
1957 .fc_metric
= IP6_RT_PRIO_USER
,
1958 .fc_ifindex
= ifindex
,
1959 .fc_dst_len
= prefixlen
,
1960 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
1961 RTF_UP
| RTF_PREF(pref
),
1962 .fc_nlinfo
.portid
= 0,
1963 .fc_nlinfo
.nlh
= NULL
,
1964 .fc_nlinfo
.nl_net
= net
,
1967 cfg
.fc_dst
= *prefix
;
1968 cfg
.fc_gateway
= *gwaddr
;
1970 /* We should treat it as a default route if prefix length is 0. */
1972 cfg
.fc_flags
|= RTF_DEFAULT
;
1974 ip6_route_add(&cfg
);
1976 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
1980 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
1982 struct rt6_info
*rt
;
1983 struct fib6_table
*table
;
1985 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
1989 read_lock_bh(&table
->tb6_lock
);
1990 for (rt
= table
->tb6_root
.leaf
; rt
; rt
=rt
->dst
.rt6_next
) {
1991 if (dev
== rt
->dst
.dev
&&
1992 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
1993 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
1998 read_unlock_bh(&table
->tb6_lock
);
2002 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
2003 struct net_device
*dev
,
2006 struct fib6_config cfg
= {
2007 .fc_table
= RT6_TABLE_DFLT
,
2008 .fc_metric
= IP6_RT_PRIO_USER
,
2009 .fc_ifindex
= dev
->ifindex
,
2010 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
2011 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
2012 .fc_nlinfo
.portid
= 0,
2013 .fc_nlinfo
.nlh
= NULL
,
2014 .fc_nlinfo
.nl_net
= dev_net(dev
),
2017 cfg
.fc_gateway
= *gwaddr
;
2019 ip6_route_add(&cfg
);
2021 return rt6_get_dflt_router(gwaddr
, dev
);
2024 void rt6_purge_dflt_routers(struct net
*net
)
2026 struct rt6_info
*rt
;
2027 struct fib6_table
*table
;
2029 /* NOTE: Keep consistent with rt6_get_dflt_router */
2030 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
2035 read_lock_bh(&table
->tb6_lock
);
2036 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
2037 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
) &&
2038 (!rt
->rt6i_idev
|| rt
->rt6i_idev
->cnf
.accept_ra
!= 2)) {
2040 read_unlock_bh(&table
->tb6_lock
);
2045 read_unlock_bh(&table
->tb6_lock
);
2048 static void rtmsg_to_fib6_config(struct net
*net
,
2049 struct in6_rtmsg
*rtmsg
,
2050 struct fib6_config
*cfg
)
2052 memset(cfg
, 0, sizeof(*cfg
));
2054 cfg
->fc_table
= RT6_TABLE_MAIN
;
2055 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
2056 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
2057 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
2058 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
2059 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
2060 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
2062 cfg
->fc_nlinfo
.nl_net
= net
;
2064 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
2065 cfg
->fc_src
= rtmsg
->rtmsg_src
;
2066 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
2069 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
2071 struct fib6_config cfg
;
2072 struct in6_rtmsg rtmsg
;
2076 case SIOCADDRT
: /* Add a route */
2077 case SIOCDELRT
: /* Delete a route */
2078 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2080 err
= copy_from_user(&rtmsg
, arg
,
2081 sizeof(struct in6_rtmsg
));
2085 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
2090 err
= ip6_route_add(&cfg
);
2093 err
= ip6_route_del(&cfg
);
2107 * Drop the packet on the floor
2110 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2113 struct dst_entry
*dst
= skb_dst(skb
);
2114 switch (ipstats_mib_noroutes
) {
2115 case IPSTATS_MIB_INNOROUTES
:
2116 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2117 if (type
== IPV6_ADDR_ANY
) {
2118 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2119 IPSTATS_MIB_INADDRERRORS
);
2123 case IPSTATS_MIB_OUTNOROUTES
:
2124 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2125 ipstats_mib_noroutes
);
2128 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2133 static int ip6_pkt_discard(struct sk_buff
*skb
)
2135 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2138 static int ip6_pkt_discard_out(struct sk_buff
*skb
)
2140 skb
->dev
= skb_dst(skb
)->dev
;
2141 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2144 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2146 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2148 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2151 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
)
2153 skb
->dev
= skb_dst(skb
)->dev
;
2154 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2160 * Allocate a dst for local (unicast / anycast) address.
2163 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2164 const struct in6_addr
*addr
,
2167 struct net
*net
= dev_net(idev
->dev
);
2168 struct rt6_info
*rt
= ip6_dst_alloc(net
, net
->loopback_dev
, 0, NULL
);
2171 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2172 return ERR_PTR(-ENOMEM
);
2177 rt
->dst
.flags
|= DST_HOST
;
2178 rt
->dst
.input
= ip6_input
;
2179 rt
->dst
.output
= ip6_output
;
2180 rt
->rt6i_idev
= idev
;
2182 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2184 rt
->rt6i_flags
|= RTF_ANYCAST
;
2186 rt
->rt6i_flags
|= RTF_LOCAL
;
2188 rt
->rt6i_gateway
= *addr
;
2189 rt
->rt6i_dst
.addr
= *addr
;
2190 rt
->rt6i_dst
.plen
= 128;
2191 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2193 atomic_set(&rt
->dst
.__refcnt
, 1);
2198 int ip6_route_get_saddr(struct net
*net
,
2199 struct rt6_info
*rt
,
2200 const struct in6_addr
*daddr
,
2202 struct in6_addr
*saddr
)
2204 struct inet6_dev
*idev
= ip6_dst_idev((struct dst_entry
*)rt
);
2206 if (rt
->rt6i_prefsrc
.plen
)
2207 *saddr
= rt
->rt6i_prefsrc
.addr
;
2209 err
= ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2210 daddr
, prefs
, saddr
);
2214 /* remove deleted ip from prefsrc entries */
2215 struct arg_dev_net_ip
{
2216 struct net_device
*dev
;
2218 struct in6_addr
*addr
;
2221 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2223 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2224 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2225 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2227 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2228 rt
!= net
->ipv6
.ip6_null_entry
&&
2229 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2230 /* remove prefsrc entry */
2231 rt
->rt6i_prefsrc
.plen
= 0;
2236 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2238 struct net
*net
= dev_net(ifp
->idev
->dev
);
2239 struct arg_dev_net_ip adni
= {
2240 .dev
= ifp
->idev
->dev
,
2244 fib6_clean_all(net
, fib6_remove_prefsrc
, 0, &adni
);
2247 struct arg_dev_net
{
2248 struct net_device
*dev
;
2252 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2254 const struct arg_dev_net
*adn
= arg
;
2255 const struct net_device
*dev
= adn
->dev
;
2257 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2258 rt
!= adn
->net
->ipv6
.ip6_null_entry
)
2264 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2266 struct arg_dev_net adn
= {
2271 fib6_clean_all(net
, fib6_ifdown
, 0, &adn
);
2272 icmp6_clean_all(fib6_ifdown
, &adn
);
2275 struct rt6_mtu_change_arg
{
2276 struct net_device
*dev
;
2280 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2282 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2283 struct inet6_dev
*idev
;
2285 /* In IPv6 pmtu discovery is not optional,
2286 so that RTAX_MTU lock cannot disable it.
2287 We still use this lock to block changes
2288 caused by addrconf/ndisc.
2291 idev
= __in6_dev_get(arg
->dev
);
2295 /* For administrative MTU increase, there is no way to discover
2296 IPv6 PMTU increase, so PMTU increase should be updated here.
2297 Since RFC 1981 doesn't include administrative MTU increase
2298 update PMTU increase is a MUST. (i.e. jumbo frame)
2301 If new MTU is less than route PMTU, this new MTU will be the
2302 lowest MTU in the path, update the route PMTU to reflect PMTU
2303 decreases; if new MTU is greater than route PMTU, and the
2304 old MTU is the lowest MTU in the path, update the route PMTU
2305 to reflect the increase. In this case if the other nodes' MTU
2306 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2309 if (rt
->dst
.dev
== arg
->dev
&&
2310 !dst_metric_locked(&rt
->dst
, RTAX_MTU
) &&
2311 (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2312 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2313 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
))) {
2314 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2319 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2321 struct rt6_mtu_change_arg arg
= {
2326 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, 0, &arg
);
2329 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2330 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2331 [RTA_OIF
] = { .type
= NLA_U32
},
2332 [RTA_IIF
] = { .type
= NLA_U32
},
2333 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2334 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2335 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2338 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2339 struct fib6_config
*cfg
)
2342 struct nlattr
*tb
[RTA_MAX
+1];
2345 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2350 rtm
= nlmsg_data(nlh
);
2351 memset(cfg
, 0, sizeof(*cfg
));
2353 cfg
->fc_table
= rtm
->rtm_table
;
2354 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2355 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2356 cfg
->fc_flags
= RTF_UP
;
2357 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2358 cfg
->fc_type
= rtm
->rtm_type
;
2360 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2361 rtm
->rtm_type
== RTN_BLACKHOLE
||
2362 rtm
->rtm_type
== RTN_PROHIBIT
||
2363 rtm
->rtm_type
== RTN_THROW
)
2364 cfg
->fc_flags
|= RTF_REJECT
;
2366 if (rtm
->rtm_type
== RTN_LOCAL
)
2367 cfg
->fc_flags
|= RTF_LOCAL
;
2369 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2370 cfg
->fc_nlinfo
.nlh
= nlh
;
2371 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2373 if (tb
[RTA_GATEWAY
]) {
2374 nla_memcpy(&cfg
->fc_gateway
, tb
[RTA_GATEWAY
], 16);
2375 cfg
->fc_flags
|= RTF_GATEWAY
;
2379 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2381 if (nla_len(tb
[RTA_DST
]) < plen
)
2384 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2388 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2390 if (nla_len(tb
[RTA_SRC
]) < plen
)
2393 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2396 if (tb
[RTA_PREFSRC
])
2397 nla_memcpy(&cfg
->fc_prefsrc
, tb
[RTA_PREFSRC
], 16);
2400 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2402 if (tb
[RTA_PRIORITY
])
2403 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2405 if (tb
[RTA_METRICS
]) {
2406 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2407 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2411 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2413 if (tb
[RTA_MULTIPATH
]) {
2414 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2415 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2423 static int ip6_route_multipath(struct fib6_config
*cfg
, int add
)
2425 struct fib6_config r_cfg
;
2426 struct rtnexthop
*rtnh
;
2429 int err
= 0, last_err
= 0;
2432 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
2433 remaining
= cfg
->fc_mp_len
;
2435 /* Parse a Multipath Entry */
2436 while (rtnh_ok(rtnh
, remaining
)) {
2437 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
2438 if (rtnh
->rtnh_ifindex
)
2439 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
2441 attrlen
= rtnh_attrlen(rtnh
);
2443 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
2445 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
2447 nla_memcpy(&r_cfg
.fc_gateway
, nla
, 16);
2448 r_cfg
.fc_flags
|= RTF_GATEWAY
;
2451 err
= add
? ip6_route_add(&r_cfg
) : ip6_route_del(&r_cfg
);
2454 /* If we are trying to remove a route, do not stop the
2455 * loop when ip6_route_del() fails (because next hop is
2456 * already gone), we should try to remove all next hops.
2459 /* If add fails, we should try to delete all
2460 * next hops that have been already added.
2466 /* Because each route is added like a single route we remove
2467 * this flag after the first nexthop (if there is a collision,
2468 * we have already fail to add the first nexthop:
2469 * fib6_add_rt2node() has reject it).
2471 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~NLM_F_EXCL
;
2472 rtnh
= rtnh_next(rtnh
, &remaining
);
2478 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
)
2480 struct fib6_config cfg
;
2483 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2488 return ip6_route_multipath(&cfg
, 0);
2490 return ip6_route_del(&cfg
);
2493 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
)
2495 struct fib6_config cfg
;
2498 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2503 return ip6_route_multipath(&cfg
, 1);
2505 return ip6_route_add(&cfg
);
2508 static inline size_t rt6_nlmsg_size(void)
2510 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2511 + nla_total_size(16) /* RTA_SRC */
2512 + nla_total_size(16) /* RTA_DST */
2513 + nla_total_size(16) /* RTA_GATEWAY */
2514 + nla_total_size(16) /* RTA_PREFSRC */
2515 + nla_total_size(4) /* RTA_TABLE */
2516 + nla_total_size(4) /* RTA_IIF */
2517 + nla_total_size(4) /* RTA_OIF */
2518 + nla_total_size(4) /* RTA_PRIORITY */
2519 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2520 + nla_total_size(sizeof(struct rta_cacheinfo
));
2523 static int rt6_fill_node(struct net
*net
,
2524 struct sk_buff
*skb
, struct rt6_info
*rt
,
2525 struct in6_addr
*dst
, struct in6_addr
*src
,
2526 int iif
, int type
, u32 portid
, u32 seq
,
2527 int prefix
, int nowait
, unsigned int flags
)
2530 struct nlmsghdr
*nlh
;
2534 if (prefix
) { /* user wants prefix routes only */
2535 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2536 /* success since this is not a prefix route */
2541 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
2545 rtm
= nlmsg_data(nlh
);
2546 rtm
->rtm_family
= AF_INET6
;
2547 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2548 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2551 table
= rt
->rt6i_table
->tb6_id
;
2553 table
= RT6_TABLE_UNSPEC
;
2554 rtm
->rtm_table
= table
;
2555 if (nla_put_u32(skb
, RTA_TABLE
, table
))
2556 goto nla_put_failure
;
2557 if (rt
->rt6i_flags
& RTF_REJECT
) {
2558 switch (rt
->dst
.error
) {
2560 rtm
->rtm_type
= RTN_BLACKHOLE
;
2563 rtm
->rtm_type
= RTN_PROHIBIT
;
2566 rtm
->rtm_type
= RTN_THROW
;
2569 rtm
->rtm_type
= RTN_UNREACHABLE
;
2573 else if (rt
->rt6i_flags
& RTF_LOCAL
)
2574 rtm
->rtm_type
= RTN_LOCAL
;
2575 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
2576 rtm
->rtm_type
= RTN_LOCAL
;
2578 rtm
->rtm_type
= RTN_UNICAST
;
2580 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2581 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2582 if (rt
->rt6i_flags
& RTF_DYNAMIC
)
2583 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2584 else if (rt
->rt6i_flags
& RTF_ADDRCONF
) {
2585 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ROUTEINFO
))
2586 rtm
->rtm_protocol
= RTPROT_RA
;
2588 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2591 if (rt
->rt6i_flags
& RTF_CACHE
)
2592 rtm
->rtm_flags
|= RTM_F_CLONED
;
2595 if (nla_put(skb
, RTA_DST
, 16, dst
))
2596 goto nla_put_failure
;
2597 rtm
->rtm_dst_len
= 128;
2598 } else if (rtm
->rtm_dst_len
)
2599 if (nla_put(skb
, RTA_DST
, 16, &rt
->rt6i_dst
.addr
))
2600 goto nla_put_failure
;
2601 #ifdef CONFIG_IPV6_SUBTREES
2603 if (nla_put(skb
, RTA_SRC
, 16, src
))
2604 goto nla_put_failure
;
2605 rtm
->rtm_src_len
= 128;
2606 } else if (rtm
->rtm_src_len
&&
2607 nla_put(skb
, RTA_SRC
, 16, &rt
->rt6i_src
.addr
))
2608 goto nla_put_failure
;
2611 #ifdef CONFIG_IPV6_MROUTE
2612 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2613 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2618 goto nla_put_failure
;
2620 if (err
== -EMSGSIZE
)
2621 goto nla_put_failure
;
2626 if (nla_put_u32(skb
, RTA_IIF
, iif
))
2627 goto nla_put_failure
;
2629 struct in6_addr saddr_buf
;
2630 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
2631 nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2632 goto nla_put_failure
;
2635 if (rt
->rt6i_prefsrc
.plen
) {
2636 struct in6_addr saddr_buf
;
2637 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
2638 if (nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2639 goto nla_put_failure
;
2642 if (rtnetlink_put_metrics(skb
, dst_metrics_ptr(&rt
->dst
)) < 0)
2643 goto nla_put_failure
;
2645 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2646 if (nla_put(skb
, RTA_GATEWAY
, 16, &rt
->rt6i_gateway
) < 0)
2647 goto nla_put_failure
;
2651 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2652 goto nla_put_failure
;
2653 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
2654 goto nla_put_failure
;
2656 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
2658 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
2659 goto nla_put_failure
;
2661 return nlmsg_end(skb
, nlh
);
2664 nlmsg_cancel(skb
, nlh
);
2668 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2670 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2673 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2674 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2675 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2679 return rt6_fill_node(arg
->net
,
2680 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2681 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
2682 prefix
, 0, NLM_F_MULTI
);
2685 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
* nlh
)
2687 struct net
*net
= sock_net(in_skb
->sk
);
2688 struct nlattr
*tb
[RTA_MAX
+1];
2689 struct rt6_info
*rt
;
2690 struct sk_buff
*skb
;
2693 int err
, iif
= 0, oif
= 0;
2695 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2700 memset(&fl6
, 0, sizeof(fl6
));
2703 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2706 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
2710 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
2713 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
2717 iif
= nla_get_u32(tb
[RTA_IIF
]);
2720 oif
= nla_get_u32(tb
[RTA_OIF
]);
2723 struct net_device
*dev
;
2726 dev
= __dev_get_by_index(net
, iif
);
2732 fl6
.flowi6_iif
= iif
;
2734 if (!ipv6_addr_any(&fl6
.saddr
))
2735 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
2737 rt
= (struct rt6_info
*)ip6_route_input_lookup(net
, dev
, &fl6
,
2740 fl6
.flowi6_oif
= oif
;
2742 rt
= (struct rt6_info
*)ip6_route_output(net
, NULL
, &fl6
);
2745 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2752 /* Reserve room for dummy headers, this skb can pass
2753 through good chunk of routing engine.
2755 skb_reset_mac_header(skb
);
2756 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
2758 skb_dst_set(skb
, &rt
->dst
);
2760 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
2761 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
2762 nlh
->nlmsg_seq
, 0, 0, 0);
2768 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2773 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
2775 struct sk_buff
*skb
;
2776 struct net
*net
= info
->nl_net
;
2781 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
2783 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
2787 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
2788 event
, info
->portid
, seq
, 0, 0, 0);
2790 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2791 WARN_ON(err
== -EMSGSIZE
);
2795 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
2796 info
->nlh
, gfp_any());
2800 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
2803 static int ip6_route_dev_notify(struct notifier_block
*this,
2804 unsigned long event
, void *ptr
)
2806 struct net_device
*dev
= netdev_notifier_info_to_dev(ptr
);
2807 struct net
*net
= dev_net(dev
);
2809 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
2810 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
2811 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
2812 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2813 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
2814 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
2815 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
2816 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
2827 #ifdef CONFIG_PROC_FS
2829 static const struct file_operations ipv6_route_proc_fops
= {
2830 .owner
= THIS_MODULE
,
2831 .open
= ipv6_route_open
,
2833 .llseek
= seq_lseek
,
2834 .release
= seq_release_net
,
2837 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
2839 struct net
*net
= (struct net
*)seq
->private;
2840 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
2841 net
->ipv6
.rt6_stats
->fib_nodes
,
2842 net
->ipv6
.rt6_stats
->fib_route_nodes
,
2843 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
2844 net
->ipv6
.rt6_stats
->fib_rt_entries
,
2845 net
->ipv6
.rt6_stats
->fib_rt_cache
,
2846 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
2847 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
2852 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
2854 return single_open_net(inode
, file
, rt6_stats_seq_show
);
2857 static const struct file_operations rt6_stats_seq_fops
= {
2858 .owner
= THIS_MODULE
,
2859 .open
= rt6_stats_seq_open
,
2861 .llseek
= seq_lseek
,
2862 .release
= single_release_net
,
2864 #endif /* CONFIG_PROC_FS */
2866 #ifdef CONFIG_SYSCTL
2869 int ipv6_sysctl_rtcache_flush(struct ctl_table
*ctl
, int write
,
2870 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2877 net
= (struct net
*)ctl
->extra1
;
2878 delay
= net
->ipv6
.sysctl
.flush_delay
;
2879 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
2880 fib6_run_gc(delay
<= 0 ? 0 : (unsigned long)delay
, net
, delay
> 0);
2884 struct ctl_table ipv6_route_table_template
[] = {
2886 .procname
= "flush",
2887 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
2888 .maxlen
= sizeof(int),
2890 .proc_handler
= ipv6_sysctl_rtcache_flush
2893 .procname
= "gc_thresh",
2894 .data
= &ip6_dst_ops_template
.gc_thresh
,
2895 .maxlen
= sizeof(int),
2897 .proc_handler
= proc_dointvec
,
2900 .procname
= "max_size",
2901 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
2902 .maxlen
= sizeof(int),
2904 .proc_handler
= proc_dointvec
,
2907 .procname
= "gc_min_interval",
2908 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2909 .maxlen
= sizeof(int),
2911 .proc_handler
= proc_dointvec_jiffies
,
2914 .procname
= "gc_timeout",
2915 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
2916 .maxlen
= sizeof(int),
2918 .proc_handler
= proc_dointvec_jiffies
,
2921 .procname
= "gc_interval",
2922 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
2923 .maxlen
= sizeof(int),
2925 .proc_handler
= proc_dointvec_jiffies
,
2928 .procname
= "gc_elasticity",
2929 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
2930 .maxlen
= sizeof(int),
2932 .proc_handler
= proc_dointvec
,
2935 .procname
= "mtu_expires",
2936 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
2937 .maxlen
= sizeof(int),
2939 .proc_handler
= proc_dointvec_jiffies
,
2942 .procname
= "min_adv_mss",
2943 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
2944 .maxlen
= sizeof(int),
2946 .proc_handler
= proc_dointvec
,
2949 .procname
= "gc_min_interval_ms",
2950 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2951 .maxlen
= sizeof(int),
2953 .proc_handler
= proc_dointvec_ms_jiffies
,
2958 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
2960 struct ctl_table
*table
;
2962 table
= kmemdup(ipv6_route_table_template
,
2963 sizeof(ipv6_route_table_template
),
2967 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
2968 table
[0].extra1
= net
;
2969 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
2970 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
2971 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2972 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2973 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
2974 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2975 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
2976 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2977 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2979 /* Don't export sysctls to unprivileged users */
2980 if (net
->user_ns
!= &init_user_ns
)
2981 table
[0].procname
= NULL
;
2988 static int __net_init
ip6_route_net_init(struct net
*net
)
2992 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
2993 sizeof(net
->ipv6
.ip6_dst_ops
));
2995 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
2996 goto out_ip6_dst_ops
;
2998 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
2999 sizeof(*net
->ipv6
.ip6_null_entry
),
3001 if (!net
->ipv6
.ip6_null_entry
)
3002 goto out_ip6_dst_entries
;
3003 net
->ipv6
.ip6_null_entry
->dst
.path
=
3004 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
3005 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3006 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
3007 ip6_template_metrics
, true);
3009 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3010 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
3011 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
3013 if (!net
->ipv6
.ip6_prohibit_entry
)
3014 goto out_ip6_null_entry
;
3015 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
3016 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
3017 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3018 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
3019 ip6_template_metrics
, true);
3021 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
3022 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
3024 if (!net
->ipv6
.ip6_blk_hole_entry
)
3025 goto out_ip6_prohibit_entry
;
3026 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
3027 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
3028 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3029 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
3030 ip6_template_metrics
, true);
3033 net
->ipv6
.sysctl
.flush_delay
= 0;
3034 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
3035 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
3036 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
3037 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
3038 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
3039 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
3040 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
3042 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
3048 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3049 out_ip6_prohibit_entry
:
3050 kfree(net
->ipv6
.ip6_prohibit_entry
);
3052 kfree(net
->ipv6
.ip6_null_entry
);
3054 out_ip6_dst_entries
:
3055 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3060 static void __net_exit
ip6_route_net_exit(struct net
*net
)
3062 kfree(net
->ipv6
.ip6_null_entry
);
3063 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3064 kfree(net
->ipv6
.ip6_prohibit_entry
);
3065 kfree(net
->ipv6
.ip6_blk_hole_entry
);
3067 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3070 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3072 #ifdef CONFIG_PROC_FS
3073 proc_create("ipv6_route", 0, net
->proc_net
, &ipv6_route_proc_fops
);
3074 proc_create("rt6_stats", S_IRUGO
, net
->proc_net
, &rt6_stats_seq_fops
);
3079 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
3081 #ifdef CONFIG_PROC_FS
3082 remove_proc_entry("ipv6_route", net
->proc_net
);
3083 remove_proc_entry("rt6_stats", net
->proc_net
);
3087 static struct pernet_operations ip6_route_net_ops
= {
3088 .init
= ip6_route_net_init
,
3089 .exit
= ip6_route_net_exit
,
3092 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
3094 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
3098 inet_peer_base_init(bp
);
3099 net
->ipv6
.peers
= bp
;
3103 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
3105 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
3107 net
->ipv6
.peers
= NULL
;
3108 inetpeer_invalidate_tree(bp
);
3112 static struct pernet_operations ipv6_inetpeer_ops
= {
3113 .init
= ipv6_inetpeer_init
,
3114 .exit
= ipv6_inetpeer_exit
,
3117 static struct pernet_operations ip6_route_net_late_ops
= {
3118 .init
= ip6_route_net_init_late
,
3119 .exit
= ip6_route_net_exit_late
,
3122 static struct notifier_block ip6_route_dev_notifier
= {
3123 .notifier_call
= ip6_route_dev_notify
,
3127 int __init
ip6_route_init(void)
3132 ip6_dst_ops_template
.kmem_cachep
=
3133 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
3134 SLAB_HWCACHE_ALIGN
, NULL
);
3135 if (!ip6_dst_ops_template
.kmem_cachep
)
3138 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
3140 goto out_kmem_cache
;
3142 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
3144 goto out_dst_entries
;
3146 ret
= register_pernet_subsys(&ip6_route_net_ops
);
3148 goto out_register_inetpeer
;
3150 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
3152 /* Registering of the loopback is done before this portion of code,
3153 * the loopback reference in rt6_info will not be taken, do it
3154 * manually for init_net */
3155 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
3156 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3157 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3158 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
3159 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3160 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
3161 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3165 goto out_register_subsys
;
3171 ret
= fib6_rules_init();
3175 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
3177 goto fib6_rules_init
;
3180 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
3181 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
3182 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
3183 goto out_register_late_subsys
;
3185 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
3187 goto out_register_late_subsys
;
3192 out_register_late_subsys
:
3193 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3195 fib6_rules_cleanup();
3200 out_register_subsys
:
3201 unregister_pernet_subsys(&ip6_route_net_ops
);
3202 out_register_inetpeer
:
3203 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3205 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3207 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
3211 void ip6_route_cleanup(void)
3213 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
3214 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3215 fib6_rules_cleanup();
3218 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3219 unregister_pernet_subsys(&ip6_route_net_ops
);
3220 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3221 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);