2 * Linux INET6 implementation
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * $Id: route.c,v 1.56 2001/10/31 21:55:55 davem Exp $
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
18 * YOSHIFUJI Hideaki @USAGI
19 * reworked default router selection.
20 * - respect outgoing interface
21 * - select from (probably) reachable routers (i.e.
22 * routers in REACHABLE, STALE, DELAY or PROBE states).
23 * - always select the same router if it is (probably)
24 * reachable. otherwise, round-robin the list.
26 * Fixed routing subtrees.
29 #include <linux/capability.h>
30 #include <linux/errno.h>
31 #include <linux/types.h>
32 #include <linux/times.h>
33 #include <linux/socket.h>
34 #include <linux/sockios.h>
35 #include <linux/net.h>
36 #include <linux/route.h>
37 #include <linux/netdevice.h>
38 #include <linux/in6.h>
39 #include <linux/init.h>
40 #include <linux/if_arp.h>
41 #include <linux/proc_fs.h>
42 #include <linux/seq_file.h>
43 #include <linux/nsproxy.h>
44 #include <net/net_namespace.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
52 #include <linux/rtnetlink.h>
55 #include <net/netevent.h>
56 #include <net/netlink.h>
58 #include <asm/uaccess.h>
61 #include <linux/sysctl.h>
64 /* Set to 3 to get tracing. */
68 #define RDBG(x) printk x
69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
72 #define RT6_TRACE(x...) do { ; } while (0)
75 #define CLONE_OFFLINK_ROUTE 0
77 static struct rt6_info
* ip6_rt_copy(struct rt6_info
*ort
);
78 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
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
, u32 mtu
);
90 #ifdef CONFIG_IPV6_ROUTE_INFO
91 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
92 struct in6_addr
*prefix
, int prefixlen
,
93 struct in6_addr
*gwaddr
, int ifindex
,
95 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
96 struct in6_addr
*prefix
, int prefixlen
,
97 struct in6_addr
*gwaddr
, int ifindex
);
100 static struct dst_ops ip6_dst_ops_template
= {
102 .protocol
= __constant_htons(ETH_P_IPV6
),
105 .check
= ip6_dst_check
,
106 .destroy
= ip6_dst_destroy
,
107 .ifdown
= ip6_dst_ifdown
,
108 .negative_advice
= ip6_negative_advice
,
109 .link_failure
= ip6_link_failure
,
110 .update_pmtu
= ip6_rt_update_pmtu
,
111 .local_out
= ip6_local_out
,
112 .entry_size
= sizeof(struct rt6_info
),
113 .entries
= ATOMIC_INIT(0),
116 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, u32 mtu
)
120 static struct dst_ops ip6_dst_blackhole_ops
= {
122 .protocol
= __constant_htons(ETH_P_IPV6
),
123 .destroy
= ip6_dst_destroy
,
124 .check
= ip6_dst_check
,
125 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
126 .entry_size
= sizeof(struct rt6_info
),
127 .entries
= ATOMIC_INIT(0),
130 static struct rt6_info ip6_null_entry_template
= {
133 .__refcnt
= ATOMIC_INIT(1),
136 .error
= -ENETUNREACH
,
137 .metrics
= { [RTAX_HOPLIMIT
- 1] = 255, },
138 .input
= ip6_pkt_discard
,
139 .output
= ip6_pkt_discard_out
,
142 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
143 .rt6i_metric
= ~(u32
) 0,
144 .rt6i_ref
= ATOMIC_INIT(1),
147 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
149 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
150 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
);
152 struct rt6_info ip6_prohibit_entry_template
= {
155 .__refcnt
= ATOMIC_INIT(1),
159 .metrics
= { [RTAX_HOPLIMIT
- 1] = 255, },
160 .input
= ip6_pkt_prohibit
,
161 .output
= ip6_pkt_prohibit_out
,
164 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
165 .rt6i_metric
= ~(u32
) 0,
166 .rt6i_ref
= ATOMIC_INIT(1),
169 static struct rt6_info ip6_blk_hole_entry_template
= {
172 .__refcnt
= ATOMIC_INIT(1),
176 .metrics
= { [RTAX_HOPLIMIT
- 1] = 255, },
177 .input
= dst_discard
,
178 .output
= dst_discard
,
181 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
182 .rt6i_metric
= ~(u32
) 0,
183 .rt6i_ref
= ATOMIC_INIT(1),
188 /* allocate dst with ip6_dst_ops */
189 static inline struct rt6_info
*ip6_dst_alloc(struct dst_ops
*ops
)
191 return (struct rt6_info
*)dst_alloc(ops
);
194 static void ip6_dst_destroy(struct dst_entry
*dst
)
196 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
197 struct inet6_dev
*idev
= rt
->rt6i_idev
;
200 rt
->rt6i_idev
= NULL
;
205 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
208 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
209 struct inet6_dev
*idev
= rt
->rt6i_idev
;
210 struct net_device
*loopback_dev
=
211 dev
->nd_net
->loopback_dev
;
213 if (dev
!= loopback_dev
&& idev
!= NULL
&& idev
->dev
== dev
) {
214 struct inet6_dev
*loopback_idev
=
215 in6_dev_get(loopback_dev
);
216 if (loopback_idev
!= NULL
) {
217 rt
->rt6i_idev
= loopback_idev
;
223 static __inline__
int rt6_check_expired(const struct rt6_info
*rt
)
225 return (rt
->rt6i_flags
& RTF_EXPIRES
&&
226 time_after(jiffies
, rt
->rt6i_expires
));
229 static inline int rt6_need_strict(struct in6_addr
*daddr
)
231 return (ipv6_addr_type(daddr
) &
232 (IPV6_ADDR_MULTICAST
| IPV6_ADDR_LINKLOCAL
));
236 * Route lookup. Any table->tb6_lock is implied.
239 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
244 struct rt6_info
*local
= NULL
;
245 struct rt6_info
*sprt
;
248 for (sprt
= rt
; sprt
; sprt
= sprt
->u
.dst
.rt6_next
) {
249 struct net_device
*dev
= sprt
->rt6i_dev
;
250 if (dev
->ifindex
== oif
)
252 if (dev
->flags
& IFF_LOOPBACK
) {
253 if (sprt
->rt6i_idev
== NULL
||
254 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
257 if (local
&& (!oif
||
258 local
->rt6i_idev
->dev
->ifindex
== oif
))
269 return net
->ipv6
.ip6_null_entry
;
274 #ifdef CONFIG_IPV6_ROUTER_PREF
275 static void rt6_probe(struct rt6_info
*rt
)
277 struct neighbour
*neigh
= rt
? rt
->rt6i_nexthop
: NULL
;
279 * Okay, this does not seem to be appropriate
280 * for now, however, we need to check if it
281 * is really so; aka Router Reachability Probing.
283 * Router Reachability Probe MUST be rate-limited
284 * to no more than one per minute.
286 if (!neigh
|| (neigh
->nud_state
& NUD_VALID
))
288 read_lock_bh(&neigh
->lock
);
289 if (!(neigh
->nud_state
& NUD_VALID
) &&
290 time_after(jiffies
, neigh
->updated
+ rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
291 struct in6_addr mcaddr
;
292 struct in6_addr
*target
;
294 neigh
->updated
= jiffies
;
295 read_unlock_bh(&neigh
->lock
);
297 target
= (struct in6_addr
*)&neigh
->primary_key
;
298 addrconf_addr_solict_mult(target
, &mcaddr
);
299 ndisc_send_ns(rt
->rt6i_dev
, NULL
, target
, &mcaddr
, NULL
);
301 read_unlock_bh(&neigh
->lock
);
304 static inline void rt6_probe(struct rt6_info
*rt
)
311 * Default Router Selection (RFC 2461 6.3.6)
313 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
315 struct net_device
*dev
= rt
->rt6i_dev
;
316 if (!oif
|| dev
->ifindex
== oif
)
318 if ((dev
->flags
& IFF_LOOPBACK
) &&
319 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
324 static inline int rt6_check_neigh(struct rt6_info
*rt
)
326 struct neighbour
*neigh
= rt
->rt6i_nexthop
;
328 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
329 !(rt
->rt6i_flags
& RTF_GATEWAY
))
332 read_lock_bh(&neigh
->lock
);
333 if (neigh
->nud_state
& NUD_VALID
)
335 #ifdef CONFIG_IPV6_ROUTER_PREF
336 else if (neigh
->nud_state
& NUD_FAILED
)
341 read_unlock_bh(&neigh
->lock
);
347 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
352 m
= rt6_check_dev(rt
, oif
);
353 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
355 #ifdef CONFIG_IPV6_ROUTER_PREF
356 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
358 n
= rt6_check_neigh(rt
);
359 if (!n
&& (strict
& RT6_LOOKUP_F_REACHABLE
))
364 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
365 int *mpri
, struct rt6_info
*match
)
369 if (rt6_check_expired(rt
))
372 m
= rt6_score_route(rt
, oif
, strict
);
377 if (strict
& RT6_LOOKUP_F_REACHABLE
)
381 } else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
389 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
390 struct rt6_info
*rr_head
,
391 u32 metric
, int oif
, int strict
)
393 struct rt6_info
*rt
, *match
;
397 for (rt
= rr_head
; rt
&& rt
->rt6i_metric
== metric
;
398 rt
= rt
->u
.dst
.rt6_next
)
399 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
400 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
&& rt
->rt6i_metric
== metric
;
401 rt
= rt
->u
.dst
.rt6_next
)
402 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
407 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
409 struct rt6_info
*match
, *rt0
;
412 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
413 __FUNCTION__
, fn
->leaf
, oif
);
417 fn
->rr_ptr
= rt0
= fn
->leaf
;
419 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
);
422 (strict
& RT6_LOOKUP_F_REACHABLE
)) {
423 struct rt6_info
*next
= rt0
->u
.dst
.rt6_next
;
425 /* no entries matched; do round-robin */
426 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
433 RT6_TRACE("%s() => %p\n",
434 __FUNCTION__
, match
);
436 net
= rt0
->rt6i_dev
->nd_net
;
437 return (match
? match
: net
->ipv6
.ip6_null_entry
);
440 #ifdef CONFIG_IPV6_ROUTE_INFO
441 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
442 struct in6_addr
*gwaddr
)
444 struct net
*net
= dev
->nd_net
;
445 struct route_info
*rinfo
= (struct route_info
*) opt
;
446 struct in6_addr prefix_buf
, *prefix
;
451 if (len
< sizeof(struct route_info
)) {
455 /* Sanity check for prefix_len and length */
456 if (rinfo
->length
> 3) {
458 } else if (rinfo
->prefix_len
> 128) {
460 } else if (rinfo
->prefix_len
> 64) {
461 if (rinfo
->length
< 2) {
464 } else if (rinfo
->prefix_len
> 0) {
465 if (rinfo
->length
< 1) {
470 pref
= rinfo
->route_pref
;
471 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
472 pref
= ICMPV6_ROUTER_PREF_MEDIUM
;
474 lifetime
= ntohl(rinfo
->lifetime
);
475 if (lifetime
== 0xffffffff) {
477 } else if (lifetime
> 0x7fffffff/HZ
) {
478 /* Avoid arithmetic overflow */
479 lifetime
= 0x7fffffff/HZ
- 1;
482 if (rinfo
->length
== 3)
483 prefix
= (struct in6_addr
*)rinfo
->prefix
;
485 /* this function is safe */
486 ipv6_addr_prefix(&prefix_buf
,
487 (struct in6_addr
*)rinfo
->prefix
,
489 prefix
= &prefix_buf
;
492 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
,
495 if (rt
&& !lifetime
) {
501 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
504 rt
->rt6i_flags
= RTF_ROUTEINFO
|
505 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
508 if (lifetime
== 0xffffffff) {
509 rt
->rt6i_flags
&= ~RTF_EXPIRES
;
511 rt
->rt6i_expires
= jiffies
+ HZ
* lifetime
;
512 rt
->rt6i_flags
|= RTF_EXPIRES
;
514 dst_release(&rt
->u
.dst
);
520 #define BACKTRACK(__net, saddr) \
522 if (rt == __net->ipv6.ip6_null_entry) { \
523 struct fib6_node *pn; \
525 if (fn->fn_flags & RTN_TL_ROOT) \
528 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
529 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
532 if (fn->fn_flags & RTN_RTINFO) \
538 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
539 struct fib6_table
*table
,
540 struct flowi
*fl
, int flags
)
542 struct fib6_node
*fn
;
545 read_lock_bh(&table
->tb6_lock
);
546 fn
= fib6_lookup(&table
->tb6_root
, &fl
->fl6_dst
, &fl
->fl6_src
);
549 rt
= rt6_device_match(net
, rt
, fl
->oif
, flags
);
550 BACKTRACK(net
, &fl
->fl6_src
);
552 dst_use(&rt
->u
.dst
, jiffies
);
553 read_unlock_bh(&table
->tb6_lock
);
558 struct rt6_info
*rt6_lookup(struct net
*net
, struct in6_addr
*daddr
,
559 struct in6_addr
*saddr
, int oif
, int strict
)
569 struct dst_entry
*dst
;
570 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
573 memcpy(&fl
.fl6_src
, saddr
, sizeof(*saddr
));
574 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
577 dst
= fib6_rule_lookup(net
, &fl
, flags
, ip6_pol_route_lookup
);
579 return (struct rt6_info
*) dst
;
586 EXPORT_SYMBOL(rt6_lookup
);
588 /* ip6_ins_rt is called with FREE table->tb6_lock.
589 It takes new route entry, the addition fails by any reason the
590 route is freed. In any case, if caller does not hold it, it may
594 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
)
597 struct fib6_table
*table
;
599 table
= rt
->rt6i_table
;
600 write_lock_bh(&table
->tb6_lock
);
601 err
= fib6_add(&table
->tb6_root
, rt
, info
);
602 write_unlock_bh(&table
->tb6_lock
);
607 int ip6_ins_rt(struct rt6_info
*rt
)
609 struct nl_info info
= {
610 .nl_net
= rt
->rt6i_dev
->nd_net
,
612 return __ip6_ins_rt(rt
, &info
);
615 static struct rt6_info
*rt6_alloc_cow(struct rt6_info
*ort
, struct in6_addr
*daddr
,
616 struct in6_addr
*saddr
)
624 rt
= ip6_rt_copy(ort
);
627 if (!(rt
->rt6i_flags
&RTF_GATEWAY
)) {
628 if (rt
->rt6i_dst
.plen
!= 128 &&
629 ipv6_addr_equal(&rt
->rt6i_dst
.addr
, daddr
))
630 rt
->rt6i_flags
|= RTF_ANYCAST
;
631 ipv6_addr_copy(&rt
->rt6i_gateway
, daddr
);
634 ipv6_addr_copy(&rt
->rt6i_dst
.addr
, daddr
);
635 rt
->rt6i_dst
.plen
= 128;
636 rt
->rt6i_flags
|= RTF_CACHE
;
637 rt
->u
.dst
.flags
|= DST_HOST
;
639 #ifdef CONFIG_IPV6_SUBTREES
640 if (rt
->rt6i_src
.plen
&& saddr
) {
641 ipv6_addr_copy(&rt
->rt6i_src
.addr
, saddr
);
642 rt
->rt6i_src
.plen
= 128;
646 rt
->rt6i_nexthop
= ndisc_get_neigh(rt
->rt6i_dev
, &rt
->rt6i_gateway
);
653 static struct rt6_info
*rt6_alloc_clone(struct rt6_info
*ort
, struct in6_addr
*daddr
)
655 struct rt6_info
*rt
= ip6_rt_copy(ort
);
657 ipv6_addr_copy(&rt
->rt6i_dst
.addr
, daddr
);
658 rt
->rt6i_dst
.plen
= 128;
659 rt
->rt6i_flags
|= RTF_CACHE
;
660 rt
->u
.dst
.flags
|= DST_HOST
;
661 rt
->rt6i_nexthop
= neigh_clone(ort
->rt6i_nexthop
);
666 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
667 struct flowi
*fl
, int flags
)
669 struct fib6_node
*fn
;
670 struct rt6_info
*rt
, *nrt
;
674 int reachable
= ipv6_devconf
.forwarding
? 0 : RT6_LOOKUP_F_REACHABLE
;
676 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
679 read_lock_bh(&table
->tb6_lock
);
682 fn
= fib6_lookup(&table
->tb6_root
, &fl
->fl6_dst
, &fl
->fl6_src
);
685 rt
= rt6_select(fn
, oif
, strict
| reachable
);
687 BACKTRACK(net
, &fl
->fl6_src
);
688 if (rt
== net
->ipv6
.ip6_null_entry
||
689 rt
->rt6i_flags
& RTF_CACHE
)
692 dst_hold(&rt
->u
.dst
);
693 read_unlock_bh(&table
->tb6_lock
);
695 if (!rt
->rt6i_nexthop
&& !(rt
->rt6i_flags
& RTF_NONEXTHOP
))
696 nrt
= rt6_alloc_cow(rt
, &fl
->fl6_dst
, &fl
->fl6_src
);
698 #if CLONE_OFFLINK_ROUTE
699 nrt
= rt6_alloc_clone(rt
, &fl
->fl6_dst
);
705 dst_release(&rt
->u
.dst
);
706 rt
= nrt
? : net
->ipv6
.ip6_null_entry
;
708 dst_hold(&rt
->u
.dst
);
710 err
= ip6_ins_rt(nrt
);
719 * Race condition! In the gap, when table->tb6_lock was
720 * released someone could insert this route. Relookup.
722 dst_release(&rt
->u
.dst
);
730 dst_hold(&rt
->u
.dst
);
731 read_unlock_bh(&table
->tb6_lock
);
733 rt
->u
.dst
.lastuse
= jiffies
;
739 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
740 struct flowi
*fl
, int flags
)
742 return ip6_pol_route(net
, table
, fl
->iif
, fl
, flags
);
745 void ip6_route_input(struct sk_buff
*skb
)
747 struct ipv6hdr
*iph
= ipv6_hdr(skb
);
748 struct net
*net
= skb
->dev
->nd_net
;
749 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
751 .iif
= skb
->dev
->ifindex
,
756 .flowlabel
= (* (__be32
*) iph
)&IPV6_FLOWINFO_MASK
,
760 .proto
= iph
->nexthdr
,
763 if (rt6_need_strict(&iph
->daddr
))
764 flags
|= RT6_LOOKUP_F_IFACE
;
766 skb
->dst
= fib6_rule_lookup(net
, &fl
, flags
, ip6_pol_route_input
);
769 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
770 struct flowi
*fl
, int flags
)
772 return ip6_pol_route(net
, table
, fl
->oif
, fl
, flags
);
775 struct dst_entry
* ip6_route_output(struct net
*net
, struct sock
*sk
,
780 if (rt6_need_strict(&fl
->fl6_dst
))
781 flags
|= RT6_LOOKUP_F_IFACE
;
783 if (!ipv6_addr_any(&fl
->fl6_src
))
784 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
786 return fib6_rule_lookup(net
, fl
, flags
, ip6_pol_route_output
);
789 EXPORT_SYMBOL(ip6_route_output
);
791 int ip6_dst_blackhole(struct sock
*sk
, struct dst_entry
**dstp
, struct flowi
*fl
)
793 struct rt6_info
*ort
= (struct rt6_info
*) *dstp
;
794 struct rt6_info
*rt
= (struct rt6_info
*)
795 dst_alloc(&ip6_dst_blackhole_ops
);
796 struct dst_entry
*new = NULL
;
801 atomic_set(&new->__refcnt
, 1);
803 new->input
= dst_discard
;
804 new->output
= dst_discard
;
806 memcpy(new->metrics
, ort
->u
.dst
.metrics
, RTAX_MAX
*sizeof(u32
));
807 new->dev
= ort
->u
.dst
.dev
;
810 rt
->rt6i_idev
= ort
->rt6i_idev
;
812 in6_dev_hold(rt
->rt6i_idev
);
813 rt
->rt6i_expires
= 0;
815 ipv6_addr_copy(&rt
->rt6i_gateway
, &ort
->rt6i_gateway
);
816 rt
->rt6i_flags
= ort
->rt6i_flags
& ~RTF_EXPIRES
;
819 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
820 #ifdef CONFIG_IPV6_SUBTREES
821 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
829 return (new ? 0 : -ENOMEM
);
831 EXPORT_SYMBOL_GPL(ip6_dst_blackhole
);
834 * Destination cache support functions
837 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
841 rt
= (struct rt6_info
*) dst
;
843 if (rt
&& rt
->rt6i_node
&& (rt
->rt6i_node
->fn_sernum
== cookie
))
849 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
851 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
854 if (rt
->rt6i_flags
& RTF_CACHE
)
862 static void ip6_link_failure(struct sk_buff
*skb
)
866 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0, skb
->dev
);
868 rt
= (struct rt6_info
*) skb
->dst
;
870 if (rt
->rt6i_flags
&RTF_CACHE
) {
871 dst_set_expires(&rt
->u
.dst
, 0);
872 rt
->rt6i_flags
|= RTF_EXPIRES
;
873 } else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
))
874 rt
->rt6i_node
->fn_sernum
= -1;
878 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, u32 mtu
)
880 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
882 if (mtu
< dst_mtu(dst
) && rt6
->rt6i_dst
.plen
== 128) {
883 rt6
->rt6i_flags
|= RTF_MODIFIED
;
884 if (mtu
< IPV6_MIN_MTU
) {
886 dst
->metrics
[RTAX_FEATURES
-1] |= RTAX_FEATURE_ALLFRAG
;
888 dst
->metrics
[RTAX_MTU
-1] = mtu
;
889 call_netevent_notifiers(NETEVENT_PMTU_UPDATE
, dst
);
893 static int ipv6_get_mtu(struct net_device
*dev
);
895 static inline unsigned int ipv6_advmss(struct net
*net
, unsigned int mtu
)
897 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
899 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
900 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
903 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
904 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
905 * IPV6_MAXPLEN is also valid and means: "any MSS,
906 * rely only on pmtu discovery"
908 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
913 static struct dst_entry
*icmp6_dst_gc_list
;
914 static DEFINE_SPINLOCK(icmp6_dst_lock
);
916 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
917 struct neighbour
*neigh
,
918 struct in6_addr
*addr
)
921 struct inet6_dev
*idev
= in6_dev_get(dev
);
922 struct net
*net
= dev
->nd_net
;
924 if (unlikely(idev
== NULL
))
927 rt
= ip6_dst_alloc(net
->ipv6
.ip6_dst_ops
);
928 if (unlikely(rt
== NULL
)) {
937 neigh
= ndisc_get_neigh(dev
, addr
);
940 rt
->rt6i_idev
= idev
;
941 rt
->rt6i_nexthop
= neigh
;
942 atomic_set(&rt
->u
.dst
.__refcnt
, 1);
943 rt
->u
.dst
.metrics
[RTAX_HOPLIMIT
-1] = 255;
944 rt
->u
.dst
.metrics
[RTAX_MTU
-1] = ipv6_get_mtu(rt
->rt6i_dev
);
945 rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] = ipv6_advmss(net
, dst_mtu(&rt
->u
.dst
));
946 rt
->u
.dst
.output
= ip6_output
;
948 #if 0 /* there's no chance to use these for ndisc */
949 rt
->u
.dst
.flags
= ipv6_addr_type(addr
) & IPV6_ADDR_UNICAST
952 ipv6_addr_copy(&rt
->rt6i_dst
.addr
, addr
);
953 rt
->rt6i_dst
.plen
= 128;
956 spin_lock_bh(&icmp6_dst_lock
);
957 rt
->u
.dst
.next
= icmp6_dst_gc_list
;
958 icmp6_dst_gc_list
= &rt
->u
.dst
;
959 spin_unlock_bh(&icmp6_dst_lock
);
961 fib6_force_start_gc(net
);
967 int icmp6_dst_gc(int *more
)
969 struct dst_entry
*dst
, *next
, **pprev
;
975 spin_lock_bh(&icmp6_dst_lock
);
976 pprev
= &icmp6_dst_gc_list
;
978 while ((dst
= *pprev
) != NULL
) {
979 if (!atomic_read(&dst
->__refcnt
)) {
989 spin_unlock_bh(&icmp6_dst_lock
);
994 static int ip6_dst_gc(struct dst_ops
*ops
)
996 unsigned long now
= jiffies
;
997 struct net
*net
= ops
->dst_net
;
998 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
999 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1000 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1001 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1002 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1004 if (time_after(rt_last_gc
+ rt_min_interval
, now
) &&
1005 atomic_read(&ops
->entries
) <= rt_max_size
)
1008 net
->ipv6
.ip6_rt_gc_expire
++;
1009 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
);
1010 net
->ipv6
.ip6_rt_last_gc
= now
;
1011 if (atomic_read(&ops
->entries
) < ops
->gc_thresh
)
1012 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1014 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1015 return (atomic_read(&ops
->entries
) > rt_max_size
);
1018 /* Clean host part of a prefix. Not necessary in radix tree,
1019 but results in cleaner routing tables.
1021 Remove it only when all the things will work!
1024 static int ipv6_get_mtu(struct net_device
*dev
)
1026 int mtu
= IPV6_MIN_MTU
;
1027 struct inet6_dev
*idev
;
1029 idev
= in6_dev_get(dev
);
1031 mtu
= idev
->cnf
.mtu6
;
1037 int ipv6_get_hoplimit(struct net_device
*dev
)
1039 int hoplimit
= ipv6_devconf
.hop_limit
;
1040 struct inet6_dev
*idev
;
1042 idev
= in6_dev_get(dev
);
1044 hoplimit
= idev
->cnf
.hop_limit
;
1054 int ip6_route_add(struct fib6_config
*cfg
)
1057 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1058 struct rt6_info
*rt
= NULL
;
1059 struct net_device
*dev
= NULL
;
1060 struct inet6_dev
*idev
= NULL
;
1061 struct fib6_table
*table
;
1064 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1066 #ifndef CONFIG_IPV6_SUBTREES
1067 if (cfg
->fc_src_len
)
1070 if (cfg
->fc_ifindex
) {
1072 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1075 idev
= in6_dev_get(dev
);
1080 if (cfg
->fc_metric
== 0)
1081 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1083 table
= fib6_new_table(net
, cfg
->fc_table
);
1084 if (table
== NULL
) {
1089 rt
= ip6_dst_alloc(net
->ipv6
.ip6_dst_ops
);
1096 rt
->u
.dst
.obsolete
= -1;
1097 rt
->rt6i_expires
= jiffies
+ clock_t_to_jiffies(cfg
->fc_expires
);
1099 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1100 cfg
->fc_protocol
= RTPROT_BOOT
;
1101 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1103 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1105 if (addr_type
& IPV6_ADDR_MULTICAST
)
1106 rt
->u
.dst
.input
= ip6_mc_input
;
1108 rt
->u
.dst
.input
= ip6_forward
;
1110 rt
->u
.dst
.output
= ip6_output
;
1112 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1113 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1114 if (rt
->rt6i_dst
.plen
== 128)
1115 rt
->u
.dst
.flags
= DST_HOST
;
1117 #ifdef CONFIG_IPV6_SUBTREES
1118 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1119 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1122 rt
->rt6i_metric
= cfg
->fc_metric
;
1124 /* We cannot add true routes via loopback here,
1125 they would result in kernel looping; promote them to reject routes
1127 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1128 (dev
&& (dev
->flags
&IFF_LOOPBACK
) && !(addr_type
&IPV6_ADDR_LOOPBACK
))) {
1129 /* hold loopback dev/idev if we haven't done so. */
1130 if (dev
!= net
->loopback_dev
) {
1135 dev
= net
->loopback_dev
;
1137 idev
= in6_dev_get(dev
);
1143 rt
->u
.dst
.output
= ip6_pkt_discard_out
;
1144 rt
->u
.dst
.input
= ip6_pkt_discard
;
1145 rt
->u
.dst
.error
= -ENETUNREACH
;
1146 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1150 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1151 struct in6_addr
*gw_addr
;
1154 gw_addr
= &cfg
->fc_gateway
;
1155 ipv6_addr_copy(&rt
->rt6i_gateway
, gw_addr
);
1156 gwa_type
= ipv6_addr_type(gw_addr
);
1158 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1159 struct rt6_info
*grt
;
1161 /* IPv6 strictly inhibits using not link-local
1162 addresses as nexthop address.
1163 Otherwise, router will not able to send redirects.
1164 It is very good, but in some (rare!) circumstances
1165 (SIT, PtP, NBMA NOARP links) it is handy to allow
1166 some exceptions. --ANK
1169 if (!(gwa_type
&IPV6_ADDR_UNICAST
))
1172 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1174 err
= -EHOSTUNREACH
;
1178 if (dev
!= grt
->rt6i_dev
) {
1179 dst_release(&grt
->u
.dst
);
1183 dev
= grt
->rt6i_dev
;
1184 idev
= grt
->rt6i_idev
;
1186 in6_dev_hold(grt
->rt6i_idev
);
1188 if (!(grt
->rt6i_flags
&RTF_GATEWAY
))
1190 dst_release(&grt
->u
.dst
);
1196 if (dev
== NULL
|| (dev
->flags
&IFF_LOOPBACK
))
1204 if (cfg
->fc_flags
& (RTF_GATEWAY
| RTF_NONEXTHOP
)) {
1205 rt
->rt6i_nexthop
= __neigh_lookup_errno(&nd_tbl
, &rt
->rt6i_gateway
, dev
);
1206 if (IS_ERR(rt
->rt6i_nexthop
)) {
1207 err
= PTR_ERR(rt
->rt6i_nexthop
);
1208 rt
->rt6i_nexthop
= NULL
;
1213 rt
->rt6i_flags
= cfg
->fc_flags
;
1220 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1221 int type
= nla_type(nla
);
1224 if (type
> RTAX_MAX
) {
1229 rt
->u
.dst
.metrics
[type
- 1] = nla_get_u32(nla
);
1234 if (rt
->u
.dst
.metrics
[RTAX_HOPLIMIT
-1] == 0)
1235 rt
->u
.dst
.metrics
[RTAX_HOPLIMIT
-1] = -1;
1236 if (!rt
->u
.dst
.metrics
[RTAX_MTU
-1])
1237 rt
->u
.dst
.metrics
[RTAX_MTU
-1] = ipv6_get_mtu(dev
);
1238 if (!rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1])
1239 rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] = ipv6_advmss(net
, dst_mtu(&rt
->u
.dst
));
1240 rt
->u
.dst
.dev
= dev
;
1241 rt
->rt6i_idev
= idev
;
1242 rt
->rt6i_table
= table
;
1244 cfg
->fc_nlinfo
.nl_net
= dev
->nd_net
;
1246 return __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
);
1254 dst_free(&rt
->u
.dst
);
1258 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1261 struct fib6_table
*table
;
1262 struct net
*net
= rt
->rt6i_dev
->nd_net
;
1264 if (rt
== net
->ipv6
.ip6_null_entry
)
1267 table
= rt
->rt6i_table
;
1268 write_lock_bh(&table
->tb6_lock
);
1270 err
= fib6_del(rt
, info
);
1271 dst_release(&rt
->u
.dst
);
1273 write_unlock_bh(&table
->tb6_lock
);
1278 int ip6_del_rt(struct rt6_info
*rt
)
1280 struct nl_info info
= {
1281 .nl_net
= rt
->rt6i_dev
->nd_net
,
1283 return __ip6_del_rt(rt
, &info
);
1286 static int ip6_route_del(struct fib6_config
*cfg
)
1288 struct fib6_table
*table
;
1289 struct fib6_node
*fn
;
1290 struct rt6_info
*rt
;
1293 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1297 read_lock_bh(&table
->tb6_lock
);
1299 fn
= fib6_locate(&table
->tb6_root
,
1300 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1301 &cfg
->fc_src
, cfg
->fc_src_len
);
1304 for (rt
= fn
->leaf
; rt
; rt
= rt
->u
.dst
.rt6_next
) {
1305 if (cfg
->fc_ifindex
&&
1306 (rt
->rt6i_dev
== NULL
||
1307 rt
->rt6i_dev
->ifindex
!= cfg
->fc_ifindex
))
1309 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1310 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1312 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1314 dst_hold(&rt
->u
.dst
);
1315 read_unlock_bh(&table
->tb6_lock
);
1317 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1320 read_unlock_bh(&table
->tb6_lock
);
1328 struct ip6rd_flowi
{
1330 struct in6_addr gateway
;
1333 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
1334 struct fib6_table
*table
,
1338 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl
;
1339 struct rt6_info
*rt
;
1340 struct fib6_node
*fn
;
1343 * Get the "current" route for this destination and
1344 * check if the redirect has come from approriate router.
1346 * RFC 2461 specifies that redirects should only be
1347 * accepted if they come from the nexthop to the target.
1348 * Due to the way the routes are chosen, this notion
1349 * is a bit fuzzy and one might need to check all possible
1353 read_lock_bh(&table
->tb6_lock
);
1354 fn
= fib6_lookup(&table
->tb6_root
, &fl
->fl6_dst
, &fl
->fl6_src
);
1356 for (rt
= fn
->leaf
; rt
; rt
= rt
->u
.dst
.rt6_next
) {
1358 * Current route is on-link; redirect is always invalid.
1360 * Seems, previous statement is not true. It could
1361 * be node, which looks for us as on-link (f.e. proxy ndisc)
1362 * But then router serving it might decide, that we should
1363 * know truth 8)8) --ANK (980726).
1365 if (rt6_check_expired(rt
))
1367 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
1369 if (fl
->oif
!= rt
->rt6i_dev
->ifindex
)
1371 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
))
1377 rt
= net
->ipv6
.ip6_null_entry
;
1378 BACKTRACK(net
, &fl
->fl6_src
);
1380 dst_hold(&rt
->u
.dst
);
1382 read_unlock_bh(&table
->tb6_lock
);
1387 static struct rt6_info
*ip6_route_redirect(struct in6_addr
*dest
,
1388 struct in6_addr
*src
,
1389 struct in6_addr
*gateway
,
1390 struct net_device
*dev
)
1392 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1393 struct net
*net
= dev
->nd_net
;
1394 struct ip6rd_flowi rdfl
= {
1396 .oif
= dev
->ifindex
,
1404 .gateway
= *gateway
,
1407 if (rt6_need_strict(dest
))
1408 flags
|= RT6_LOOKUP_F_IFACE
;
1410 return (struct rt6_info
*)fib6_rule_lookup(net
, (struct flowi
*)&rdfl
,
1411 flags
, __ip6_route_redirect
);
1414 void rt6_redirect(struct in6_addr
*dest
, struct in6_addr
*src
,
1415 struct in6_addr
*saddr
,
1416 struct neighbour
*neigh
, u8
*lladdr
, int on_link
)
1418 struct rt6_info
*rt
, *nrt
= NULL
;
1419 struct netevent_redirect netevent
;
1420 struct net
*net
= neigh
->dev
->nd_net
;
1422 rt
= ip6_route_redirect(dest
, src
, saddr
, neigh
->dev
);
1424 if (rt
== net
->ipv6
.ip6_null_entry
) {
1425 if (net_ratelimit())
1426 printk(KERN_DEBUG
"rt6_redirect: source isn't a valid nexthop "
1427 "for redirect target\n");
1432 * We have finally decided to accept it.
1435 neigh_update(neigh
, lladdr
, NUD_STALE
,
1436 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
1437 NEIGH_UPDATE_F_OVERRIDE
|
1438 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
1439 NEIGH_UPDATE_F_ISROUTER
))
1443 * Redirect received -> path was valid.
1444 * Look, redirects are sent only in response to data packets,
1445 * so that this nexthop apparently is reachable. --ANK
1447 dst_confirm(&rt
->u
.dst
);
1449 /* Duplicate redirect: silently ignore. */
1450 if (neigh
== rt
->u
.dst
.neighbour
)
1453 nrt
= ip6_rt_copy(rt
);
1457 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
1459 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
1461 ipv6_addr_copy(&nrt
->rt6i_dst
.addr
, dest
);
1462 nrt
->rt6i_dst
.plen
= 128;
1463 nrt
->u
.dst
.flags
|= DST_HOST
;
1465 ipv6_addr_copy(&nrt
->rt6i_gateway
, (struct in6_addr
*)neigh
->primary_key
);
1466 nrt
->rt6i_nexthop
= neigh_clone(neigh
);
1467 /* Reset pmtu, it may be better */
1468 nrt
->u
.dst
.metrics
[RTAX_MTU
-1] = ipv6_get_mtu(neigh
->dev
);
1469 nrt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] = ipv6_advmss(neigh
->dev
->nd_net
,
1470 dst_mtu(&nrt
->u
.dst
));
1472 if (ip6_ins_rt(nrt
))
1475 netevent
.old
= &rt
->u
.dst
;
1476 netevent
.new = &nrt
->u
.dst
;
1477 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
1479 if (rt
->rt6i_flags
&RTF_CACHE
) {
1485 dst_release(&rt
->u
.dst
);
1490 * Handle ICMP "packet too big" messages
1491 * i.e. Path MTU discovery
1494 void rt6_pmtu_discovery(struct in6_addr
*daddr
, struct in6_addr
*saddr
,
1495 struct net_device
*dev
, u32 pmtu
)
1497 struct rt6_info
*rt
, *nrt
;
1498 struct net
*net
= dev
->nd_net
;
1501 rt
= rt6_lookup(net
, daddr
, saddr
, dev
->ifindex
, 0);
1505 if (pmtu
>= dst_mtu(&rt
->u
.dst
))
1508 if (pmtu
< IPV6_MIN_MTU
) {
1510 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1511 * MTU (1280) and a fragment header should always be included
1512 * after a node receiving Too Big message reporting PMTU is
1513 * less than the IPv6 Minimum Link MTU.
1515 pmtu
= IPV6_MIN_MTU
;
1519 /* New mtu received -> path was valid.
1520 They are sent only in response to data packets,
1521 so that this nexthop apparently is reachable. --ANK
1523 dst_confirm(&rt
->u
.dst
);
1525 /* Host route. If it is static, it would be better
1526 not to override it, but add new one, so that
1527 when cache entry will expire old pmtu
1528 would return automatically.
1530 if (rt
->rt6i_flags
& RTF_CACHE
) {
1531 rt
->u
.dst
.metrics
[RTAX_MTU
-1] = pmtu
;
1533 rt
->u
.dst
.metrics
[RTAX_FEATURES
-1] |= RTAX_FEATURE_ALLFRAG
;
1534 dst_set_expires(&rt
->u
.dst
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1535 rt
->rt6i_flags
|= RTF_MODIFIED
|RTF_EXPIRES
;
1540 Two cases are possible:
1541 1. It is connected route. Action: COW
1542 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1544 if (!rt
->rt6i_nexthop
&& !(rt
->rt6i_flags
& RTF_NONEXTHOP
))
1545 nrt
= rt6_alloc_cow(rt
, daddr
, saddr
);
1547 nrt
= rt6_alloc_clone(rt
, daddr
);
1550 nrt
->u
.dst
.metrics
[RTAX_MTU
-1] = pmtu
;
1552 nrt
->u
.dst
.metrics
[RTAX_FEATURES
-1] |= RTAX_FEATURE_ALLFRAG
;
1554 /* According to RFC 1981, detecting PMTU increase shouldn't be
1555 * happened within 5 mins, the recommended timer is 10 mins.
1556 * Here this route expiration time is set to ip6_rt_mtu_expires
1557 * which is 10 mins. After 10 mins the decreased pmtu is expired
1558 * and detecting PMTU increase will be automatically happened.
1560 dst_set_expires(&nrt
->u
.dst
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1561 nrt
->rt6i_flags
|= RTF_DYNAMIC
|RTF_EXPIRES
;
1566 dst_release(&rt
->u
.dst
);
1570 * Misc support functions
1573 static struct rt6_info
* ip6_rt_copy(struct rt6_info
*ort
)
1575 struct net
*net
= ort
->rt6i_dev
->nd_net
;
1576 struct rt6_info
*rt
= ip6_dst_alloc(net
->ipv6
.ip6_dst_ops
);
1579 rt
->u
.dst
.input
= ort
->u
.dst
.input
;
1580 rt
->u
.dst
.output
= ort
->u
.dst
.output
;
1582 memcpy(rt
->u
.dst
.metrics
, ort
->u
.dst
.metrics
, RTAX_MAX
*sizeof(u32
));
1583 rt
->u
.dst
.error
= ort
->u
.dst
.error
;
1584 rt
->u
.dst
.dev
= ort
->u
.dst
.dev
;
1586 dev_hold(rt
->u
.dst
.dev
);
1587 rt
->rt6i_idev
= ort
->rt6i_idev
;
1589 in6_dev_hold(rt
->rt6i_idev
);
1590 rt
->u
.dst
.lastuse
= jiffies
;
1591 rt
->rt6i_expires
= 0;
1593 ipv6_addr_copy(&rt
->rt6i_gateway
, &ort
->rt6i_gateway
);
1594 rt
->rt6i_flags
= ort
->rt6i_flags
& ~RTF_EXPIRES
;
1595 rt
->rt6i_metric
= 0;
1597 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1598 #ifdef CONFIG_IPV6_SUBTREES
1599 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1601 rt
->rt6i_table
= ort
->rt6i_table
;
1606 #ifdef CONFIG_IPV6_ROUTE_INFO
1607 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
1608 struct in6_addr
*prefix
, int prefixlen
,
1609 struct in6_addr
*gwaddr
, int ifindex
)
1611 struct fib6_node
*fn
;
1612 struct rt6_info
*rt
= NULL
;
1613 struct fib6_table
*table
;
1615 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
1619 write_lock_bh(&table
->tb6_lock
);
1620 fn
= fib6_locate(&table
->tb6_root
, prefix
,prefixlen
, NULL
, 0);
1624 for (rt
= fn
->leaf
; rt
; rt
= rt
->u
.dst
.rt6_next
) {
1625 if (rt
->rt6i_dev
->ifindex
!= ifindex
)
1627 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
1629 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
1631 dst_hold(&rt
->u
.dst
);
1635 write_unlock_bh(&table
->tb6_lock
);
1639 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
1640 struct in6_addr
*prefix
, int prefixlen
,
1641 struct in6_addr
*gwaddr
, int ifindex
,
1644 struct fib6_config cfg
= {
1645 .fc_table
= RT6_TABLE_INFO
,
1646 .fc_metric
= IP6_RT_PRIO_USER
,
1647 .fc_ifindex
= ifindex
,
1648 .fc_dst_len
= prefixlen
,
1649 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
1650 RTF_UP
| RTF_PREF(pref
),
1652 .fc_nlinfo
.nlh
= NULL
,
1653 .fc_nlinfo
.nl_net
= net
,
1656 ipv6_addr_copy(&cfg
.fc_dst
, prefix
);
1657 ipv6_addr_copy(&cfg
.fc_gateway
, gwaddr
);
1659 /* We should treat it as a default route if prefix length is 0. */
1661 cfg
.fc_flags
|= RTF_DEFAULT
;
1663 ip6_route_add(&cfg
);
1665 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
1669 struct rt6_info
*rt6_get_dflt_router(struct in6_addr
*addr
, struct net_device
*dev
)
1671 struct rt6_info
*rt
;
1672 struct fib6_table
*table
;
1674 table
= fib6_get_table(dev
->nd_net
, RT6_TABLE_DFLT
);
1678 write_lock_bh(&table
->tb6_lock
);
1679 for (rt
= table
->tb6_root
.leaf
; rt
; rt
=rt
->u
.dst
.rt6_next
) {
1680 if (dev
== rt
->rt6i_dev
&&
1681 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
1682 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
1686 dst_hold(&rt
->u
.dst
);
1687 write_unlock_bh(&table
->tb6_lock
);
1691 EXPORT_SYMBOL(rt6_get_dflt_router
);
1693 struct rt6_info
*rt6_add_dflt_router(struct in6_addr
*gwaddr
,
1694 struct net_device
*dev
,
1697 struct fib6_config cfg
= {
1698 .fc_table
= RT6_TABLE_DFLT
,
1699 .fc_metric
= IP6_RT_PRIO_USER
,
1700 .fc_ifindex
= dev
->ifindex
,
1701 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
1702 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
1704 .fc_nlinfo
.nlh
= NULL
,
1705 .fc_nlinfo
.nl_net
= dev
->nd_net
,
1708 ipv6_addr_copy(&cfg
.fc_gateway
, gwaddr
);
1710 ip6_route_add(&cfg
);
1712 return rt6_get_dflt_router(gwaddr
, dev
);
1715 void rt6_purge_dflt_routers(struct net
*net
)
1717 struct rt6_info
*rt
;
1718 struct fib6_table
*table
;
1720 /* NOTE: Keep consistent with rt6_get_dflt_router */
1721 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
1726 read_lock_bh(&table
->tb6_lock
);
1727 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->u
.dst
.rt6_next
) {
1728 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) {
1729 dst_hold(&rt
->u
.dst
);
1730 read_unlock_bh(&table
->tb6_lock
);
1735 read_unlock_bh(&table
->tb6_lock
);
1738 static void rtmsg_to_fib6_config(struct net
*net
,
1739 struct in6_rtmsg
*rtmsg
,
1740 struct fib6_config
*cfg
)
1742 memset(cfg
, 0, sizeof(*cfg
));
1744 cfg
->fc_table
= RT6_TABLE_MAIN
;
1745 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
1746 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
1747 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
1748 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
1749 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
1750 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
1752 cfg
->fc_nlinfo
.nl_net
= net
;
1754 ipv6_addr_copy(&cfg
->fc_dst
, &rtmsg
->rtmsg_dst
);
1755 ipv6_addr_copy(&cfg
->fc_src
, &rtmsg
->rtmsg_src
);
1756 ipv6_addr_copy(&cfg
->fc_gateway
, &rtmsg
->rtmsg_gateway
);
1759 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
1761 struct fib6_config cfg
;
1762 struct in6_rtmsg rtmsg
;
1766 case SIOCADDRT
: /* Add a route */
1767 case SIOCDELRT
: /* Delete a route */
1768 if (!capable(CAP_NET_ADMIN
))
1770 err
= copy_from_user(&rtmsg
, arg
,
1771 sizeof(struct in6_rtmsg
));
1775 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
1780 err
= ip6_route_add(&cfg
);
1783 err
= ip6_route_del(&cfg
);
1797 * Drop the packet on the floor
1800 static int ip6_pkt_drop(struct sk_buff
*skb
, int code
, int ipstats_mib_noroutes
)
1803 switch (ipstats_mib_noroutes
) {
1804 case IPSTATS_MIB_INNOROUTES
:
1805 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
1806 if (type
== IPV6_ADDR_ANY
|| type
== IPV6_ADDR_RESERVED
) {
1807 IP6_INC_STATS(ip6_dst_idev(skb
->dst
), IPSTATS_MIB_INADDRERRORS
);
1811 case IPSTATS_MIB_OUTNOROUTES
:
1812 IP6_INC_STATS(ip6_dst_idev(skb
->dst
), ipstats_mib_noroutes
);
1815 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0, skb
->dev
);
1820 static int ip6_pkt_discard(struct sk_buff
*skb
)
1822 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
1825 static int ip6_pkt_discard_out(struct sk_buff
*skb
)
1827 skb
->dev
= skb
->dst
->dev
;
1828 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
1831 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1833 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
1835 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
1838 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
)
1840 skb
->dev
= skb
->dst
->dev
;
1841 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
1847 * Allocate a dst for local (unicast / anycast) address.
1850 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
1851 const struct in6_addr
*addr
,
1854 struct net
*net
= idev
->dev
->nd_net
;
1855 struct rt6_info
*rt
= ip6_dst_alloc(net
->ipv6
.ip6_dst_ops
);
1858 return ERR_PTR(-ENOMEM
);
1860 dev_hold(net
->loopback_dev
);
1863 rt
->u
.dst
.flags
= DST_HOST
;
1864 rt
->u
.dst
.input
= ip6_input
;
1865 rt
->u
.dst
.output
= ip6_output
;
1866 rt
->rt6i_dev
= net
->loopback_dev
;
1867 rt
->rt6i_idev
= idev
;
1868 rt
->u
.dst
.metrics
[RTAX_MTU
-1] = ipv6_get_mtu(rt
->rt6i_dev
);
1869 rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] = ipv6_advmss(net
, dst_mtu(&rt
->u
.dst
));
1870 rt
->u
.dst
.metrics
[RTAX_HOPLIMIT
-1] = -1;
1871 rt
->u
.dst
.obsolete
= -1;
1873 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
1875 rt
->rt6i_flags
|= RTF_ANYCAST
;
1877 rt
->rt6i_flags
|= RTF_LOCAL
;
1878 rt
->rt6i_nexthop
= ndisc_get_neigh(rt
->rt6i_dev
, &rt
->rt6i_gateway
);
1879 if (rt
->rt6i_nexthop
== NULL
) {
1880 dst_free(&rt
->u
.dst
);
1881 return ERR_PTR(-ENOMEM
);
1884 ipv6_addr_copy(&rt
->rt6i_dst
.addr
, addr
);
1885 rt
->rt6i_dst
.plen
= 128;
1886 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
1888 atomic_set(&rt
->u
.dst
.__refcnt
, 1);
1893 struct arg_dev_net
{
1894 struct net_device
*dev
;
1898 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
1900 struct net_device
*dev
= ((struct arg_dev_net
*)arg
)->dev
;
1901 struct net
*net
= ((struct arg_dev_net
*)arg
)->net
;
1903 if (((void *)rt
->rt6i_dev
== dev
|| dev
== NULL
) &&
1904 rt
!= net
->ipv6
.ip6_null_entry
) {
1905 RT6_TRACE("deleted by ifdown %p\n", rt
);
1911 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
1913 struct arg_dev_net adn
= {
1918 fib6_clean_all(net
, fib6_ifdown
, 0, &adn
);
1921 struct rt6_mtu_change_arg
1923 struct net_device
*dev
;
1927 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
1929 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
1930 struct inet6_dev
*idev
;
1931 struct net
*net
= arg
->dev
->nd_net
;
1933 /* In IPv6 pmtu discovery is not optional,
1934 so that RTAX_MTU lock cannot disable it.
1935 We still use this lock to block changes
1936 caused by addrconf/ndisc.
1939 idev
= __in6_dev_get(arg
->dev
);
1943 /* For administrative MTU increase, there is no way to discover
1944 IPv6 PMTU increase, so PMTU increase should be updated here.
1945 Since RFC 1981 doesn't include administrative MTU increase
1946 update PMTU increase is a MUST. (i.e. jumbo frame)
1949 If new MTU is less than route PMTU, this new MTU will be the
1950 lowest MTU in the path, update the route PMTU to reflect PMTU
1951 decreases; if new MTU is greater than route PMTU, and the
1952 old MTU is the lowest MTU in the path, update the route PMTU
1953 to reflect the increase. In this case if the other nodes' MTU
1954 also have the lowest MTU, TOO BIG MESSAGE will be lead to
1957 if (rt
->rt6i_dev
== arg
->dev
&&
1958 !dst_metric_locked(&rt
->u
.dst
, RTAX_MTU
) &&
1959 (dst_mtu(&rt
->u
.dst
) >= arg
->mtu
||
1960 (dst_mtu(&rt
->u
.dst
) < arg
->mtu
&&
1961 dst_mtu(&rt
->u
.dst
) == idev
->cnf
.mtu6
))) {
1962 rt
->u
.dst
.metrics
[RTAX_MTU
-1] = arg
->mtu
;
1963 rt
->u
.dst
.metrics
[RTAX_ADVMSS
-1] = ipv6_advmss(net
, arg
->mtu
);
1968 void rt6_mtu_change(struct net_device
*dev
, unsigned mtu
)
1970 struct rt6_mtu_change_arg arg
= {
1975 fib6_clean_all(dev
->nd_net
, rt6_mtu_change_route
, 0, &arg
);
1978 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
1979 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
1980 [RTA_OIF
] = { .type
= NLA_U32
},
1981 [RTA_IIF
] = { .type
= NLA_U32
},
1982 [RTA_PRIORITY
] = { .type
= NLA_U32
},
1983 [RTA_METRICS
] = { .type
= NLA_NESTED
},
1986 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
1987 struct fib6_config
*cfg
)
1990 struct nlattr
*tb
[RTA_MAX
+1];
1993 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
1998 rtm
= nlmsg_data(nlh
);
1999 memset(cfg
, 0, sizeof(*cfg
));
2001 cfg
->fc_table
= rtm
->rtm_table
;
2002 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2003 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2004 cfg
->fc_flags
= RTF_UP
;
2005 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2007 if (rtm
->rtm_type
== RTN_UNREACHABLE
)
2008 cfg
->fc_flags
|= RTF_REJECT
;
2010 cfg
->fc_nlinfo
.pid
= NETLINK_CB(skb
).pid
;
2011 cfg
->fc_nlinfo
.nlh
= nlh
;
2012 cfg
->fc_nlinfo
.nl_net
= skb
->sk
->sk_net
;
2014 if (tb
[RTA_GATEWAY
]) {
2015 nla_memcpy(&cfg
->fc_gateway
, tb
[RTA_GATEWAY
], 16);
2016 cfg
->fc_flags
|= RTF_GATEWAY
;
2020 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2022 if (nla_len(tb
[RTA_DST
]) < plen
)
2025 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2029 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2031 if (nla_len(tb
[RTA_SRC
]) < plen
)
2034 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2038 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2040 if (tb
[RTA_PRIORITY
])
2041 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2043 if (tb
[RTA_METRICS
]) {
2044 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2045 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2049 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2056 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
2058 struct fib6_config cfg
;
2061 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2065 return ip6_route_del(&cfg
);
2068 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
2070 struct fib6_config cfg
;
2073 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2077 return ip6_route_add(&cfg
);
2080 static inline size_t rt6_nlmsg_size(void)
2082 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2083 + nla_total_size(16) /* RTA_SRC */
2084 + nla_total_size(16) /* RTA_DST */
2085 + nla_total_size(16) /* RTA_GATEWAY */
2086 + nla_total_size(16) /* RTA_PREFSRC */
2087 + nla_total_size(4) /* RTA_TABLE */
2088 + nla_total_size(4) /* RTA_IIF */
2089 + nla_total_size(4) /* RTA_OIF */
2090 + nla_total_size(4) /* RTA_PRIORITY */
2091 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2092 + nla_total_size(sizeof(struct rta_cacheinfo
));
2095 static int rt6_fill_node(struct sk_buff
*skb
, struct rt6_info
*rt
,
2096 struct in6_addr
*dst
, struct in6_addr
*src
,
2097 int iif
, int type
, u32 pid
, u32 seq
,
2098 int prefix
, unsigned int flags
)
2101 struct nlmsghdr
*nlh
;
2105 if (prefix
) { /* user wants prefix routes only */
2106 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2107 /* success since this is not a prefix route */
2112 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*rtm
), flags
);
2116 rtm
= nlmsg_data(nlh
);
2117 rtm
->rtm_family
= AF_INET6
;
2118 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2119 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2122 table
= rt
->rt6i_table
->tb6_id
;
2124 table
= RT6_TABLE_UNSPEC
;
2125 rtm
->rtm_table
= table
;
2126 NLA_PUT_U32(skb
, RTA_TABLE
, table
);
2127 if (rt
->rt6i_flags
&RTF_REJECT
)
2128 rtm
->rtm_type
= RTN_UNREACHABLE
;
2129 else if (rt
->rt6i_dev
&& (rt
->rt6i_dev
->flags
&IFF_LOOPBACK
))
2130 rtm
->rtm_type
= RTN_LOCAL
;
2132 rtm
->rtm_type
= RTN_UNICAST
;
2134 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2135 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2136 if (rt
->rt6i_flags
&RTF_DYNAMIC
)
2137 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2138 else if (rt
->rt6i_flags
& RTF_ADDRCONF
)
2139 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2140 else if (rt
->rt6i_flags
&RTF_DEFAULT
)
2141 rtm
->rtm_protocol
= RTPROT_RA
;
2143 if (rt
->rt6i_flags
&RTF_CACHE
)
2144 rtm
->rtm_flags
|= RTM_F_CLONED
;
2147 NLA_PUT(skb
, RTA_DST
, 16, dst
);
2148 rtm
->rtm_dst_len
= 128;
2149 } else if (rtm
->rtm_dst_len
)
2150 NLA_PUT(skb
, RTA_DST
, 16, &rt
->rt6i_dst
.addr
);
2151 #ifdef CONFIG_IPV6_SUBTREES
2153 NLA_PUT(skb
, RTA_SRC
, 16, src
);
2154 rtm
->rtm_src_len
= 128;
2155 } else if (rtm
->rtm_src_len
)
2156 NLA_PUT(skb
, RTA_SRC
, 16, &rt
->rt6i_src
.addr
);
2159 NLA_PUT_U32(skb
, RTA_IIF
, iif
);
2161 struct in6_addr saddr_buf
;
2162 if (ipv6_dev_get_saddr(ip6_dst_idev(&rt
->u
.dst
)->dev
,
2163 dst
, &saddr_buf
) == 0)
2164 NLA_PUT(skb
, RTA_PREFSRC
, 16, &saddr_buf
);
2167 if (rtnetlink_put_metrics(skb
, rt
->u
.dst
.metrics
) < 0)
2168 goto nla_put_failure
;
2170 if (rt
->u
.dst
.neighbour
)
2171 NLA_PUT(skb
, RTA_GATEWAY
, 16, &rt
->u
.dst
.neighbour
->primary_key
);
2174 NLA_PUT_U32(skb
, RTA_OIF
, rt
->rt6i_dev
->ifindex
);
2176 NLA_PUT_U32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
);
2178 expires
= rt
->rt6i_expires
? rt
->rt6i_expires
- jiffies
: 0;
2179 if (rtnl_put_cacheinfo(skb
, &rt
->u
.dst
, 0, 0, 0,
2180 expires
, rt
->u
.dst
.error
) < 0)
2181 goto nla_put_failure
;
2183 return nlmsg_end(skb
, nlh
);
2186 nlmsg_cancel(skb
, nlh
);
2190 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2192 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2195 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2196 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2197 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2201 return rt6_fill_node(arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2202 NETLINK_CB(arg
->cb
->skb
).pid
, arg
->cb
->nlh
->nlmsg_seq
,
2203 prefix
, NLM_F_MULTI
);
2206 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
* nlh
, void *arg
)
2208 struct net
*net
= in_skb
->sk
->sk_net
;
2209 struct nlattr
*tb
[RTA_MAX
+1];
2210 struct rt6_info
*rt
;
2211 struct sk_buff
*skb
;
2216 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2221 memset(&fl
, 0, sizeof(fl
));
2224 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2227 ipv6_addr_copy(&fl
.fl6_src
, nla_data(tb
[RTA_SRC
]));
2231 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
2234 ipv6_addr_copy(&fl
.fl6_dst
, nla_data(tb
[RTA_DST
]));
2238 iif
= nla_get_u32(tb
[RTA_IIF
]);
2241 fl
.oif
= nla_get_u32(tb
[RTA_OIF
]);
2244 struct net_device
*dev
;
2245 dev
= __dev_get_by_index(net
, iif
);
2252 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2258 /* Reserve room for dummy headers, this skb can pass
2259 through good chunk of routing engine.
2261 skb_reset_mac_header(skb
);
2262 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
2264 rt
= (struct rt6_info
*) ip6_route_output(&init_net
, NULL
, &fl
);
2265 skb
->dst
= &rt
->u
.dst
;
2267 err
= rt6_fill_node(skb
, rt
, &fl
.fl6_dst
, &fl
.fl6_src
, iif
,
2268 RTM_NEWROUTE
, NETLINK_CB(in_skb
).pid
,
2269 nlh
->nlmsg_seq
, 0, 0);
2275 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).pid
);
2280 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
2282 struct sk_buff
*skb
;
2283 struct net
*net
= info
->nl_net
;
2288 seq
= info
->nlh
!= NULL
? info
->nlh
->nlmsg_seq
: 0;
2290 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
2294 err
= rt6_fill_node(skb
, rt
, NULL
, NULL
, 0,
2295 event
, info
->pid
, seq
, 0, 0);
2297 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2298 WARN_ON(err
== -EMSGSIZE
);
2302 err
= rtnl_notify(skb
, net
, info
->pid
, RTNLGRP_IPV6_ROUTE
,
2303 info
->nlh
, gfp_any());
2306 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
2309 static int ip6_route_dev_notify(struct notifier_block
*this,
2310 unsigned long event
, void *data
)
2312 struct net_device
*dev
= (struct net_device
*)data
;
2313 struct net
*net
= dev
->nd_net
;
2315 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
2316 net
->ipv6
.ip6_null_entry
->u
.dst
.dev
= dev
;
2317 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
2318 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2319 net
->ipv6
.ip6_prohibit_entry
->u
.dst
.dev
= dev
;
2320 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
2321 net
->ipv6
.ip6_blk_hole_entry
->u
.dst
.dev
= dev
;
2322 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
2333 #ifdef CONFIG_PROC_FS
2335 #define RT6_INFO_LEN (32 + 4 + 32 + 4 + 32 + 40 + 5 + 1)
2346 static int rt6_info_route(struct rt6_info
*rt
, void *p_arg
)
2348 struct seq_file
*m
= p_arg
;
2350 seq_printf(m
, NIP6_SEQFMT
" %02x ", NIP6(rt
->rt6i_dst
.addr
),
2353 #ifdef CONFIG_IPV6_SUBTREES
2354 seq_printf(m
, NIP6_SEQFMT
" %02x ", NIP6(rt
->rt6i_src
.addr
),
2357 seq_puts(m
, "00000000000000000000000000000000 00 ");
2360 if (rt
->rt6i_nexthop
) {
2361 seq_printf(m
, NIP6_SEQFMT
,
2362 NIP6(*((struct in6_addr
*)rt
->rt6i_nexthop
->primary_key
)));
2364 seq_puts(m
, "00000000000000000000000000000000");
2366 seq_printf(m
, " %08x %08x %08x %08x %8s\n",
2367 rt
->rt6i_metric
, atomic_read(&rt
->u
.dst
.__refcnt
),
2368 rt
->u
.dst
.__use
, rt
->rt6i_flags
,
2369 rt
->rt6i_dev
? rt
->rt6i_dev
->name
: "");
2373 static int ipv6_route_show(struct seq_file
*m
, void *v
)
2375 struct net
*net
= (struct net
*)m
->private;
2376 fib6_clean_all(net
, rt6_info_route
, 0, m
);
2380 static int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2382 struct net
*net
= get_proc_net(inode
);
2385 return single_open(file
, ipv6_route_show
, net
);
2388 static int ipv6_route_release(struct inode
*inode
, struct file
*file
)
2390 struct seq_file
*seq
= file
->private_data
;
2391 struct net
*net
= seq
->private;
2393 return single_release(inode
, file
);
2396 static const struct file_operations ipv6_route_proc_fops
= {
2397 .owner
= THIS_MODULE
,
2398 .open
= ipv6_route_open
,
2400 .llseek
= seq_lseek
,
2401 .release
= ipv6_route_release
,
2404 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
2406 struct net
*net
= (struct net
*)seq
->private;
2407 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
2408 net
->ipv6
.rt6_stats
->fib_nodes
,
2409 net
->ipv6
.rt6_stats
->fib_route_nodes
,
2410 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
2411 net
->ipv6
.rt6_stats
->fib_rt_entries
,
2412 net
->ipv6
.rt6_stats
->fib_rt_cache
,
2413 atomic_read(&net
->ipv6
.ip6_dst_ops
->entries
),
2414 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
2419 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
2421 struct net
*net
= get_proc_net(inode
);
2422 return single_open(file
, rt6_stats_seq_show
, net
);
2425 static int rt6_stats_seq_release(struct inode
*inode
, struct file
*file
)
2427 struct seq_file
*seq
= file
->private_data
;
2428 struct net
*net
= (struct net
*)seq
->private;
2430 return single_release(inode
, file
);
2433 static const struct file_operations rt6_stats_seq_fops
= {
2434 .owner
= THIS_MODULE
,
2435 .open
= rt6_stats_seq_open
,
2437 .llseek
= seq_lseek
,
2438 .release
= rt6_stats_seq_release
,
2440 #endif /* CONFIG_PROC_FS */
2442 #ifdef CONFIG_SYSCTL
2445 int ipv6_sysctl_rtcache_flush(ctl_table
*ctl
, int write
, struct file
* filp
,
2446 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2448 struct net
*net
= current
->nsproxy
->net_ns
;
2449 int delay
= net
->ipv6
.sysctl
.flush_delay
;
2451 proc_dointvec(ctl
, write
, filp
, buffer
, lenp
, ppos
);
2452 fib6_run_gc(delay
<= 0 ? ~0UL : (unsigned long)delay
, net
);
2458 ctl_table ipv6_route_table_template
[] = {
2460 .procname
= "flush",
2461 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
2462 .maxlen
= sizeof(int),
2464 .proc_handler
= &ipv6_sysctl_rtcache_flush
2467 .ctl_name
= NET_IPV6_ROUTE_GC_THRESH
,
2468 .procname
= "gc_thresh",
2469 .data
= &ip6_dst_ops_template
.gc_thresh
,
2470 .maxlen
= sizeof(int),
2472 .proc_handler
= &proc_dointvec
,
2475 .ctl_name
= NET_IPV6_ROUTE_MAX_SIZE
,
2476 .procname
= "max_size",
2477 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
2478 .maxlen
= sizeof(int),
2480 .proc_handler
= &proc_dointvec
,
2483 .ctl_name
= NET_IPV6_ROUTE_GC_MIN_INTERVAL
,
2484 .procname
= "gc_min_interval",
2485 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2486 .maxlen
= sizeof(int),
2488 .proc_handler
= &proc_dointvec_jiffies
,
2489 .strategy
= &sysctl_jiffies
,
2492 .ctl_name
= NET_IPV6_ROUTE_GC_TIMEOUT
,
2493 .procname
= "gc_timeout",
2494 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
2495 .maxlen
= sizeof(int),
2497 .proc_handler
= &proc_dointvec_jiffies
,
2498 .strategy
= &sysctl_jiffies
,
2501 .ctl_name
= NET_IPV6_ROUTE_GC_INTERVAL
,
2502 .procname
= "gc_interval",
2503 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
2504 .maxlen
= sizeof(int),
2506 .proc_handler
= &proc_dointvec_jiffies
,
2507 .strategy
= &sysctl_jiffies
,
2510 .ctl_name
= NET_IPV6_ROUTE_GC_ELASTICITY
,
2511 .procname
= "gc_elasticity",
2512 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
2513 .maxlen
= sizeof(int),
2515 .proc_handler
= &proc_dointvec_jiffies
,
2516 .strategy
= &sysctl_jiffies
,
2519 .ctl_name
= NET_IPV6_ROUTE_MTU_EXPIRES
,
2520 .procname
= "mtu_expires",
2521 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
2522 .maxlen
= sizeof(int),
2524 .proc_handler
= &proc_dointvec_jiffies
,
2525 .strategy
= &sysctl_jiffies
,
2528 .ctl_name
= NET_IPV6_ROUTE_MIN_ADVMSS
,
2529 .procname
= "min_adv_mss",
2530 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
2531 .maxlen
= sizeof(int),
2533 .proc_handler
= &proc_dointvec_jiffies
,
2534 .strategy
= &sysctl_jiffies
,
2537 .ctl_name
= NET_IPV6_ROUTE_GC_MIN_INTERVAL_MS
,
2538 .procname
= "gc_min_interval_ms",
2539 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2540 .maxlen
= sizeof(int),
2542 .proc_handler
= &proc_dointvec_ms_jiffies
,
2543 .strategy
= &sysctl_ms_jiffies
,
2548 struct ctl_table
*ipv6_route_sysctl_init(struct net
*net
)
2550 struct ctl_table
*table
;
2552 table
= kmemdup(ipv6_route_table_template
,
2553 sizeof(ipv6_route_table_template
),
2557 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
2558 table
[1].data
= &net
->ipv6
.ip6_dst_ops
->gc_thresh
;
2559 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
2560 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2561 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2562 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
2563 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2564 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
2565 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2572 static int ip6_route_net_init(struct net
*net
)
2577 net
->ipv6
.ip6_dst_ops
= kmemdup(&ip6_dst_ops_template
,
2578 sizeof(*net
->ipv6
.ip6_dst_ops
),
2580 if (!net
->ipv6
.ip6_dst_ops
)
2582 net
->ipv6
.ip6_dst_ops
->dst_net
= net
;
2584 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
2585 sizeof(*net
->ipv6
.ip6_null_entry
),
2587 if (!net
->ipv6
.ip6_null_entry
)
2588 goto out_ip6_dst_ops
;
2589 net
->ipv6
.ip6_null_entry
->u
.dst
.path
=
2590 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
2591 net
->ipv6
.ip6_null_entry
->u
.dst
.ops
= net
->ipv6
.ip6_dst_ops
;
2593 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2594 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
2595 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
2597 if (!net
->ipv6
.ip6_prohibit_entry
) {
2598 kfree(net
->ipv6
.ip6_null_entry
);
2601 net
->ipv6
.ip6_prohibit_entry
->u
.dst
.path
=
2602 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
2603 net
->ipv6
.ip6_prohibit_entry
->u
.dst
.ops
= net
->ipv6
.ip6_dst_ops
;
2605 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
2606 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
2608 if (!net
->ipv6
.ip6_blk_hole_entry
) {
2609 kfree(net
->ipv6
.ip6_null_entry
);
2610 kfree(net
->ipv6
.ip6_prohibit_entry
);
2613 net
->ipv6
.ip6_blk_hole_entry
->u
.dst
.path
=
2614 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
2615 net
->ipv6
.ip6_blk_hole_entry
->u
.dst
.ops
= net
->ipv6
.ip6_dst_ops
;
2618 #ifdef CONFIG_PROC_FS
2619 proc_net_fops_create(net
, "ipv6_route", 0, &ipv6_route_proc_fops
);
2620 proc_net_fops_create(net
, "rt6_stats", S_IRUGO
, &rt6_stats_seq_fops
);
2622 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
2629 kfree(net
->ipv6
.ip6_dst_ops
);
2633 static void ip6_route_net_exit(struct net
*net
)
2635 #ifdef CONFIG_PROC_FS
2636 proc_net_remove(net
, "ipv6_route");
2637 proc_net_remove(net
, "rt6_stats");
2639 kfree(net
->ipv6
.ip6_null_entry
);
2640 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2641 kfree(net
->ipv6
.ip6_prohibit_entry
);
2642 kfree(net
->ipv6
.ip6_blk_hole_entry
);
2644 kfree(net
->ipv6
.ip6_dst_ops
);
2647 static struct pernet_operations ip6_route_net_ops
= {
2648 .init
= ip6_route_net_init
,
2649 .exit
= ip6_route_net_exit
,
2652 static struct notifier_block ip6_route_dev_notifier
= {
2653 .notifier_call
= ip6_route_dev_notify
,
2657 int __init
ip6_route_init(void)
2662 ip6_dst_ops_template
.kmem_cachep
=
2663 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
2664 SLAB_HWCACHE_ALIGN
, NULL
);
2665 if (!ip6_dst_ops_template
.kmem_cachep
)
2668 ret
= register_pernet_subsys(&ip6_route_net_ops
);
2670 goto out_kmem_cache
;
2672 /* Registering of the loopback is done before this portion of code,
2673 * the loopback reference in rt6_info will not be taken, do it
2674 * manually for init_net */
2675 init_net
.ipv6
.ip6_null_entry
->u
.dst
.dev
= init_net
.loopback_dev
;
2676 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
2677 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2678 init_net
.ipv6
.ip6_prohibit_entry
->u
.dst
.dev
= init_net
.loopback_dev
;
2679 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
2680 init_net
.ipv6
.ip6_blk_hole_entry
->u
.dst
.dev
= init_net
.loopback_dev
;
2681 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
2685 goto out_register_subsys
;
2691 ret
= fib6_rules_init();
2696 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
) ||
2697 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
) ||
2698 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
))
2699 goto fib6_rules_init
;
2701 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
2703 goto fib6_rules_init
;
2709 fib6_rules_cleanup();
2714 out_register_subsys
:
2715 unregister_pernet_subsys(&ip6_route_net_ops
);
2717 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
2721 void ip6_route_cleanup(void)
2723 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
2724 fib6_rules_cleanup();
2727 unregister_pernet_subsys(&ip6_route_net_ops
);
2728 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);