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>
68 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
69 const struct in6_addr
*dest
);
70 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
71 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
72 static unsigned int ip6_mtu(const struct dst_entry
*dst
);
73 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
74 static void ip6_dst_destroy(struct dst_entry
*);
75 static void ip6_dst_ifdown(struct dst_entry
*,
76 struct net_device
*dev
, int how
);
77 static int ip6_dst_gc(struct dst_ops
*ops
);
79 static int ip6_pkt_discard(struct sk_buff
*skb
);
80 static int ip6_pkt_discard_out(struct sk_buff
*skb
);
81 static void ip6_link_failure(struct sk_buff
*skb
);
82 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
83 struct sk_buff
*skb
, u32 mtu
);
84 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
,
87 #ifdef CONFIG_IPV6_ROUTE_INFO
88 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
89 const struct in6_addr
*prefix
, int prefixlen
,
90 const struct in6_addr
*gwaddr
, int ifindex
,
92 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
93 const struct in6_addr
*prefix
, int prefixlen
,
94 const struct in6_addr
*gwaddr
, int ifindex
);
97 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
99 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
100 struct inet_peer
*peer
;
103 if (!(rt
->dst
.flags
& DST_HOST
))
106 peer
= rt6_get_peer_create(rt
);
108 u32
*old_p
= __DST_METRICS_PTR(old
);
109 unsigned long prev
, new;
112 if (inet_metrics_new(peer
))
113 memcpy(p
, old_p
, sizeof(u32
) * RTAX_MAX
);
115 new = (unsigned long) p
;
116 prev
= cmpxchg(&dst
->_metrics
, old
, new);
119 p
= __DST_METRICS_PTR(prev
);
120 if (prev
& DST_METRICS_READ_ONLY
)
127 static inline const void *choose_neigh_daddr(struct rt6_info
*rt
,
131 struct in6_addr
*p
= &rt
->rt6i_gateway
;
133 if (!ipv6_addr_any(p
))
134 return (const void *) p
;
136 return &ipv6_hdr(skb
)->daddr
;
140 static struct neighbour
*ip6_neigh_lookup(const struct dst_entry
*dst
,
144 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
147 daddr
= choose_neigh_daddr(rt
, skb
, daddr
);
148 n
= __ipv6_neigh_lookup(dst
->dev
, daddr
);
151 return neigh_create(&nd_tbl
, daddr
, dst
->dev
);
154 static int rt6_bind_neighbour(struct rt6_info
*rt
, struct net_device
*dev
)
156 struct neighbour
*n
= __ipv6_neigh_lookup(dev
, &rt
->rt6i_gateway
);
158 n
= neigh_create(&nd_tbl
, &rt
->rt6i_gateway
, dev
);
167 static struct dst_ops ip6_dst_ops_template
= {
169 .protocol
= cpu_to_be16(ETH_P_IPV6
),
172 .check
= ip6_dst_check
,
173 .default_advmss
= ip6_default_advmss
,
175 .cow_metrics
= ipv6_cow_metrics
,
176 .destroy
= ip6_dst_destroy
,
177 .ifdown
= ip6_dst_ifdown
,
178 .negative_advice
= ip6_negative_advice
,
179 .link_failure
= ip6_link_failure
,
180 .update_pmtu
= ip6_rt_update_pmtu
,
181 .redirect
= rt6_do_redirect
,
182 .local_out
= __ip6_local_out
,
183 .neigh_lookup
= ip6_neigh_lookup
,
186 static unsigned int ip6_blackhole_mtu(const struct dst_entry
*dst
)
188 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
190 return mtu
? : dst
->dev
->mtu
;
193 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
194 struct sk_buff
*skb
, u32 mtu
)
198 static void ip6_rt_blackhole_redirect(struct dst_entry
*dst
, struct sock
*sk
,
203 static u32
*ip6_rt_blackhole_cow_metrics(struct dst_entry
*dst
,
209 static struct dst_ops ip6_dst_blackhole_ops
= {
211 .protocol
= cpu_to_be16(ETH_P_IPV6
),
212 .destroy
= ip6_dst_destroy
,
213 .check
= ip6_dst_check
,
214 .mtu
= ip6_blackhole_mtu
,
215 .default_advmss
= ip6_default_advmss
,
216 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
217 .redirect
= ip6_rt_blackhole_redirect
,
218 .cow_metrics
= ip6_rt_blackhole_cow_metrics
,
219 .neigh_lookup
= ip6_neigh_lookup
,
222 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
223 [RTAX_HOPLIMIT
- 1] = 0,
226 static const struct rt6_info ip6_null_entry_template
= {
228 .__refcnt
= ATOMIC_INIT(1),
230 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
231 .error
= -ENETUNREACH
,
232 .input
= ip6_pkt_discard
,
233 .output
= ip6_pkt_discard_out
,
235 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
236 .rt6i_protocol
= RTPROT_KERNEL
,
237 .rt6i_metric
= ~(u32
) 0,
238 .rt6i_ref
= ATOMIC_INIT(1),
241 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
243 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
244 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
);
246 static const struct rt6_info ip6_prohibit_entry_template
= {
248 .__refcnt
= ATOMIC_INIT(1),
250 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
252 .input
= ip6_pkt_prohibit
,
253 .output
= ip6_pkt_prohibit_out
,
255 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
256 .rt6i_protocol
= RTPROT_KERNEL
,
257 .rt6i_metric
= ~(u32
) 0,
258 .rt6i_ref
= ATOMIC_INIT(1),
261 static const struct rt6_info ip6_blk_hole_entry_template
= {
263 .__refcnt
= ATOMIC_INIT(1),
265 .obsolete
= DST_OBSOLETE_FORCE_CHK
,
267 .input
= dst_discard
,
268 .output
= dst_discard
,
270 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
271 .rt6i_protocol
= RTPROT_KERNEL
,
272 .rt6i_metric
= ~(u32
) 0,
273 .rt6i_ref
= ATOMIC_INIT(1),
278 /* allocate dst with ip6_dst_ops */
279 static inline struct rt6_info
*ip6_dst_alloc(struct net
*net
,
280 struct net_device
*dev
,
282 struct fib6_table
*table
)
284 struct rt6_info
*rt
= dst_alloc(&net
->ipv6
.ip6_dst_ops
, dev
,
285 0, DST_OBSOLETE_FORCE_CHK
, flags
);
288 struct dst_entry
*dst
= &rt
->dst
;
290 memset(dst
+ 1, 0, sizeof(*rt
) - sizeof(*dst
));
291 rt6_init_peer(rt
, table
? &table
->tb6_peers
: net
->ipv6
.peers
);
292 rt
->rt6i_genid
= rt_genid(net
);
293 INIT_LIST_HEAD(&rt
->rt6i_siblings
);
294 rt
->rt6i_nsiblings
= 0;
299 static void ip6_dst_destroy(struct dst_entry
*dst
)
301 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
302 struct inet6_dev
*idev
= rt
->rt6i_idev
;
305 neigh_release(rt
->n
);
307 if (!(rt
->dst
.flags
& DST_HOST
))
308 dst_destroy_metrics_generic(dst
);
311 rt
->rt6i_idev
= NULL
;
315 if (!(rt
->rt6i_flags
& RTF_EXPIRES
) && dst
->from
)
316 dst_release(dst
->from
);
318 if (rt6_has_peer(rt
)) {
319 struct inet_peer
*peer
= rt6_peer_ptr(rt
);
324 void rt6_bind_peer(struct rt6_info
*rt
, int create
)
326 struct inet_peer_base
*base
;
327 struct inet_peer
*peer
;
329 base
= inetpeer_base_ptr(rt
->_rt6i_peer
);
333 peer
= inet_getpeer_v6(base
, &rt
->rt6i_dst
.addr
, create
);
335 if (!rt6_set_peer(rt
, peer
))
340 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
343 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
344 struct inet6_dev
*idev
= rt
->rt6i_idev
;
345 struct net_device
*loopback_dev
=
346 dev_net(dev
)->loopback_dev
;
348 if (dev
!= loopback_dev
) {
349 if (idev
&& idev
->dev
== dev
) {
350 struct inet6_dev
*loopback_idev
=
351 in6_dev_get(loopback_dev
);
353 rt
->rt6i_idev
= loopback_idev
;
357 if (rt
->n
&& rt
->n
->dev
== dev
) {
358 rt
->n
->dev
= loopback_dev
;
359 dev_hold(loopback_dev
);
365 static bool rt6_check_expired(const struct rt6_info
*rt
)
367 if (rt
->rt6i_flags
& RTF_EXPIRES
) {
368 if (time_after(jiffies
, rt
->dst
.expires
))
370 } else if (rt
->dst
.from
) {
371 return rt6_check_expired((struct rt6_info
*) rt
->dst
.from
);
376 static bool rt6_need_strict(const struct in6_addr
*daddr
)
378 return ipv6_addr_type(daddr
) &
379 (IPV6_ADDR_MULTICAST
| IPV6_ADDR_LINKLOCAL
| IPV6_ADDR_LOOPBACK
);
382 /* Multipath route selection:
383 * Hash based function using packet header and flowlabel.
384 * Adapted from fib_info_hashfn()
386 static int rt6_info_hash_nhsfn(unsigned int candidate_count
,
387 const struct flowi6
*fl6
)
389 unsigned int val
= fl6
->flowi6_proto
;
391 val
^= ipv6_addr_hash(&fl6
->daddr
);
392 val
^= ipv6_addr_hash(&fl6
->saddr
);
394 /* Work only if this not encapsulated */
395 switch (fl6
->flowi6_proto
) {
399 val
^= (__force u16
)fl6
->fl6_sport
;
400 val
^= (__force u16
)fl6
->fl6_dport
;
404 val
^= (__force u16
)fl6
->fl6_icmp_type
;
405 val
^= (__force u16
)fl6
->fl6_icmp_code
;
408 /* RFC6438 recommands to use flowlabel */
409 val
^= (__force u32
)fl6
->flowlabel
;
411 /* Perhaps, we need to tune, this function? */
412 val
= val
^ (val
>> 7) ^ (val
>> 12);
413 return val
% candidate_count
;
416 static struct rt6_info
*rt6_multipath_select(struct rt6_info
*match
,
419 struct rt6_info
*sibling
, *next_sibling
;
422 route_choosen
= rt6_info_hash_nhsfn(match
->rt6i_nsiblings
+ 1, fl6
);
423 /* Don't change the route, if route_choosen == 0
424 * (siblings does not include ourself)
427 list_for_each_entry_safe(sibling
, next_sibling
,
428 &match
->rt6i_siblings
, rt6i_siblings
) {
430 if (route_choosen
== 0) {
439 * Route lookup. Any table->tb6_lock is implied.
442 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
444 const struct in6_addr
*saddr
,
448 struct rt6_info
*local
= NULL
;
449 struct rt6_info
*sprt
;
451 if (!oif
&& ipv6_addr_any(saddr
))
454 for (sprt
= rt
; sprt
; sprt
= sprt
->dst
.rt6_next
) {
455 struct net_device
*dev
= sprt
->dst
.dev
;
458 if (dev
->ifindex
== oif
)
460 if (dev
->flags
& IFF_LOOPBACK
) {
461 if (!sprt
->rt6i_idev
||
462 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
463 if (flags
& RT6_LOOKUP_F_IFACE
&& oif
)
465 if (local
&& (!oif
||
466 local
->rt6i_idev
->dev
->ifindex
== oif
))
472 if (ipv6_chk_addr(net
, saddr
, dev
,
473 flags
& RT6_LOOKUP_F_IFACE
))
482 if (flags
& RT6_LOOKUP_F_IFACE
)
483 return net
->ipv6
.ip6_null_entry
;
489 #ifdef CONFIG_IPV6_ROUTER_PREF
490 static void rt6_probe(struct rt6_info
*rt
)
492 struct neighbour
*neigh
;
494 * Okay, this does not seem to be appropriate
495 * for now, however, we need to check if it
496 * is really so; aka Router Reachability Probing.
498 * Router Reachability Probe MUST be rate-limited
499 * to no more than one per minute.
501 neigh
= rt
? rt
->n
: NULL
;
504 write_lock_bh(&neigh
->lock
);
505 if (neigh
->nud_state
& NUD_VALID
) {
506 write_unlock_bh(&neigh
->lock
);
509 if (!(neigh
->nud_state
& NUD_VALID
) &&
510 time_after(jiffies
, neigh
->updated
+ rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
511 struct in6_addr mcaddr
;
512 struct in6_addr
*target
;
514 neigh
->updated
= jiffies
;
515 write_unlock_bh(&neigh
->lock
);
517 target
= (struct in6_addr
*)&neigh
->primary_key
;
518 addrconf_addr_solict_mult(target
, &mcaddr
);
519 ndisc_send_ns(rt
->dst
.dev
, NULL
, target
, &mcaddr
, NULL
);
521 write_unlock_bh(&neigh
->lock
);
525 static inline void rt6_probe(struct rt6_info
*rt
)
531 * Default Router Selection (RFC 2461 6.3.6)
533 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
535 struct net_device
*dev
= rt
->dst
.dev
;
536 if (!oif
|| dev
->ifindex
== oif
)
538 if ((dev
->flags
& IFF_LOOPBACK
) &&
539 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
544 static inline bool rt6_check_neigh(struct rt6_info
*rt
)
546 struct neighbour
*neigh
;
549 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
550 !(rt
->rt6i_flags
& RTF_GATEWAY
))
554 neigh
= __ipv6_neigh_lookup_noref(rt
->dst
.dev
, &rt
->rt6i_gateway
);
556 read_lock(&neigh
->lock
);
557 if (neigh
->nud_state
& NUD_VALID
)
559 #ifdef CONFIG_IPV6_ROUTER_PREF
560 else if (!(neigh
->nud_state
& NUD_FAILED
))
563 read_unlock(&neigh
->lock
);
565 rcu_read_unlock_bh();
570 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
575 m
= rt6_check_dev(rt
, oif
);
576 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
578 #ifdef CONFIG_IPV6_ROUTER_PREF
579 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
581 if (!rt6_check_neigh(rt
) && (strict
& RT6_LOOKUP_F_REACHABLE
))
586 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
587 int *mpri
, struct rt6_info
*match
)
591 if (rt6_check_expired(rt
))
594 m
= rt6_score_route(rt
, oif
, strict
);
599 if (strict
& RT6_LOOKUP_F_REACHABLE
)
603 } else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
611 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
612 struct rt6_info
*rr_head
,
613 u32 metric
, int oif
, int strict
)
615 struct rt6_info
*rt
, *match
;
619 for (rt
= rr_head
; rt
&& rt
->rt6i_metric
== metric
;
620 rt
= rt
->dst
.rt6_next
)
621 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
622 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
&& rt
->rt6i_metric
== metric
;
623 rt
= rt
->dst
.rt6_next
)
624 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
629 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
631 struct rt6_info
*match
, *rt0
;
636 fn
->rr_ptr
= rt0
= fn
->leaf
;
638 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
);
641 (strict
& RT6_LOOKUP_F_REACHABLE
)) {
642 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
644 /* no entries matched; do round-robin */
645 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
652 net
= dev_net(rt0
->dst
.dev
);
653 return match
? match
: net
->ipv6
.ip6_null_entry
;
656 #ifdef CONFIG_IPV6_ROUTE_INFO
657 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
658 const struct in6_addr
*gwaddr
)
660 struct net
*net
= dev_net(dev
);
661 struct route_info
*rinfo
= (struct route_info
*) opt
;
662 struct in6_addr prefix_buf
, *prefix
;
664 unsigned long lifetime
;
667 if (len
< sizeof(struct route_info
)) {
671 /* Sanity check for prefix_len and length */
672 if (rinfo
->length
> 3) {
674 } else if (rinfo
->prefix_len
> 128) {
676 } else if (rinfo
->prefix_len
> 64) {
677 if (rinfo
->length
< 2) {
680 } else if (rinfo
->prefix_len
> 0) {
681 if (rinfo
->length
< 1) {
686 pref
= rinfo
->route_pref
;
687 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
690 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
692 if (rinfo
->length
== 3)
693 prefix
= (struct in6_addr
*)rinfo
->prefix
;
695 /* this function is safe */
696 ipv6_addr_prefix(&prefix_buf
,
697 (struct in6_addr
*)rinfo
->prefix
,
699 prefix
= &prefix_buf
;
702 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
,
705 if (rt
&& !lifetime
) {
711 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
714 rt
->rt6i_flags
= RTF_ROUTEINFO
|
715 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
718 if (!addrconf_finite_timeout(lifetime
))
719 rt6_clean_expires(rt
);
721 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
729 #define BACKTRACK(__net, saddr) \
731 if (rt == __net->ipv6.ip6_null_entry) { \
732 struct fib6_node *pn; \
734 if (fn->fn_flags & RTN_TL_ROOT) \
737 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
738 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
741 if (fn->fn_flags & RTN_RTINFO) \
747 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
748 struct fib6_table
*table
,
749 struct flowi6
*fl6
, int flags
)
751 struct fib6_node
*fn
;
754 read_lock_bh(&table
->tb6_lock
);
755 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
758 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
, fl6
->flowi6_oif
, flags
);
759 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
760 rt
= rt6_multipath_select(rt
, fl6
);
761 BACKTRACK(net
, &fl6
->saddr
);
763 dst_use(&rt
->dst
, jiffies
);
764 read_unlock_bh(&table
->tb6_lock
);
769 struct dst_entry
* ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
772 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
774 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
776 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
777 const struct in6_addr
*saddr
, int oif
, int strict
)
779 struct flowi6 fl6
= {
783 struct dst_entry
*dst
;
784 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
787 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
788 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
791 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
793 return (struct rt6_info
*) dst
;
800 EXPORT_SYMBOL(rt6_lookup
);
802 /* ip6_ins_rt is called with FREE table->tb6_lock.
803 It takes new route entry, the addition fails by any reason the
804 route is freed. In any case, if caller does not hold it, it may
808 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
)
811 struct fib6_table
*table
;
813 table
= rt
->rt6i_table
;
814 write_lock_bh(&table
->tb6_lock
);
815 err
= fib6_add(&table
->tb6_root
, rt
, info
);
816 write_unlock_bh(&table
->tb6_lock
);
821 int ip6_ins_rt(struct rt6_info
*rt
)
823 struct nl_info info
= {
824 .nl_net
= dev_net(rt
->dst
.dev
),
826 return __ip6_ins_rt(rt
, &info
);
829 static struct rt6_info
*rt6_alloc_cow(struct rt6_info
*ort
,
830 const struct in6_addr
*daddr
,
831 const struct in6_addr
*saddr
)
839 rt
= ip6_rt_copy(ort
, daddr
);
842 int attempts
= !in_softirq();
844 if (!(rt
->rt6i_flags
& RTF_GATEWAY
)) {
845 if (ort
->rt6i_dst
.plen
!= 128 &&
846 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
847 rt
->rt6i_flags
|= RTF_ANYCAST
;
848 rt
->rt6i_gateway
= *daddr
;
851 rt
->rt6i_flags
|= RTF_CACHE
;
853 #ifdef CONFIG_IPV6_SUBTREES
854 if (rt
->rt6i_src
.plen
&& saddr
) {
855 rt
->rt6i_src
.addr
= *saddr
;
856 rt
->rt6i_src
.plen
= 128;
861 if (rt6_bind_neighbour(rt
, rt
->dst
.dev
)) {
862 struct net
*net
= dev_net(rt
->dst
.dev
);
863 int saved_rt_min_interval
=
864 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
865 int saved_rt_elasticity
=
866 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
868 if (attempts
-- > 0) {
869 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 1;
870 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= 0;
872 ip6_dst_gc(&net
->ipv6
.ip6_dst_ops
);
874 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
=
876 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
=
877 saved_rt_min_interval
;
881 net_warn_ratelimited("Neighbour table overflow\n");
890 static struct rt6_info
*rt6_alloc_clone(struct rt6_info
*ort
,
891 const struct in6_addr
*daddr
)
893 struct rt6_info
*rt
= ip6_rt_copy(ort
, daddr
);
896 rt
->rt6i_flags
|= RTF_CACHE
;
897 rt
->n
= neigh_clone(ort
->n
);
902 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
903 struct flowi6
*fl6
, int flags
)
905 struct fib6_node
*fn
;
906 struct rt6_info
*rt
, *nrt
;
910 int reachable
= net
->ipv6
.devconf_all
->forwarding
? 0 : RT6_LOOKUP_F_REACHABLE
;
912 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
915 read_lock_bh(&table
->tb6_lock
);
918 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
921 rt
= rt6_select(fn
, oif
, strict
| reachable
);
922 if (rt
->rt6i_nsiblings
&& oif
== 0)
923 rt
= rt6_multipath_select(rt
, fl6
);
924 BACKTRACK(net
, &fl6
->saddr
);
925 if (rt
== net
->ipv6
.ip6_null_entry
||
926 rt
->rt6i_flags
& RTF_CACHE
)
930 read_unlock_bh(&table
->tb6_lock
);
932 if (!(rt
->rt6i_flags
& (RTF_NONEXTHOP
| RTF_GATEWAY
)))
933 nrt
= rt6_alloc_cow(rt
, &fl6
->daddr
, &fl6
->saddr
);
934 else if (!(rt
->dst
.flags
& DST_HOST
))
935 nrt
= rt6_alloc_clone(rt
, &fl6
->daddr
);
940 rt
= nrt
? : net
->ipv6
.ip6_null_entry
;
944 err
= ip6_ins_rt(nrt
);
953 * Race condition! In the gap, when table->tb6_lock was
954 * released someone could insert this route. Relookup.
965 read_unlock_bh(&table
->tb6_lock
);
967 rt
->dst
.lastuse
= jiffies
;
973 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
974 struct flowi6
*fl6
, int flags
)
976 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
979 static struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
980 struct net_device
*dev
,
981 struct flowi6
*fl6
, int flags
)
983 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
984 flags
|= RT6_LOOKUP_F_IFACE
;
986 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
989 void ip6_route_input(struct sk_buff
*skb
)
991 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
992 struct net
*net
= dev_net(skb
->dev
);
993 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
994 struct flowi6 fl6
= {
995 .flowi6_iif
= skb
->dev
->ifindex
,
998 .flowlabel
= ip6_flowinfo(iph
),
999 .flowi6_mark
= skb
->mark
,
1000 .flowi6_proto
= iph
->nexthdr
,
1003 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
1006 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
1007 struct flowi6
*fl6
, int flags
)
1009 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1012 struct dst_entry
* ip6_route_output(struct net
*net
, const struct sock
*sk
,
1017 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1019 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
))
1020 flags
|= RT6_LOOKUP_F_IFACE
;
1022 if (!ipv6_addr_any(&fl6
->saddr
))
1023 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1025 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1027 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1030 EXPORT_SYMBOL(ip6_route_output
);
1032 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1034 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1035 struct dst_entry
*new = NULL
;
1037 rt
= dst_alloc(&ip6_dst_blackhole_ops
, ort
->dst
.dev
, 1, DST_OBSOLETE_NONE
, 0);
1041 memset(new + 1, 0, sizeof(*rt
) - sizeof(*new));
1042 rt6_init_peer(rt
, net
->ipv6
.peers
);
1045 new->input
= dst_discard
;
1046 new->output
= dst_discard
;
1048 if (dst_metrics_read_only(&ort
->dst
))
1049 new->_metrics
= ort
->dst
._metrics
;
1051 dst_copy_metrics(new, &ort
->dst
);
1052 rt
->rt6i_idev
= ort
->rt6i_idev
;
1054 in6_dev_hold(rt
->rt6i_idev
);
1056 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1057 rt
->rt6i_flags
= ort
->rt6i_flags
;
1058 rt6_clean_expires(rt
);
1059 rt
->rt6i_metric
= 0;
1061 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1062 #ifdef CONFIG_IPV6_SUBTREES
1063 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1069 dst_release(dst_orig
);
1070 return new ? new : ERR_PTR(-ENOMEM
);
1074 * Destination cache support functions
1077 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1079 struct rt6_info
*rt
;
1081 rt
= (struct rt6_info
*) dst
;
1083 /* All IPV6 dsts are created with ->obsolete set to the value
1084 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1085 * into this function always.
1087 if (rt
->rt6i_genid
!= rt_genid(dev_net(rt
->dst
.dev
)))
1090 if (rt
->rt6i_node
&& (rt
->rt6i_node
->fn_sernum
== cookie
))
1096 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1098 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1101 if (rt
->rt6i_flags
& RTF_CACHE
) {
1102 if (rt6_check_expired(rt
)) {
1114 static void ip6_link_failure(struct sk_buff
*skb
)
1116 struct rt6_info
*rt
;
1118 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1120 rt
= (struct rt6_info
*) skb_dst(skb
);
1122 if (rt
->rt6i_flags
& RTF_CACHE
)
1123 rt6_update_expires(rt
, 0);
1124 else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
))
1125 rt
->rt6i_node
->fn_sernum
= -1;
1129 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1130 struct sk_buff
*skb
, u32 mtu
)
1132 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1135 if (mtu
< dst_mtu(dst
) && rt6
->rt6i_dst
.plen
== 128) {
1136 struct net
*net
= dev_net(dst
->dev
);
1138 rt6
->rt6i_flags
|= RTF_MODIFIED
;
1139 if (mtu
< IPV6_MIN_MTU
) {
1140 u32 features
= dst_metric(dst
, RTAX_FEATURES
);
1142 features
|= RTAX_FEATURE_ALLFRAG
;
1143 dst_metric_set(dst
, RTAX_FEATURES
, features
);
1145 dst_metric_set(dst
, RTAX_MTU
, mtu
);
1146 rt6_update_expires(rt6
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1150 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1153 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1154 struct dst_entry
*dst
;
1157 memset(&fl6
, 0, sizeof(fl6
));
1158 fl6
.flowi6_oif
= oif
;
1159 fl6
.flowi6_mark
= mark
;
1160 fl6
.flowi6_flags
= 0;
1161 fl6
.daddr
= iph
->daddr
;
1162 fl6
.saddr
= iph
->saddr
;
1163 fl6
.flowlabel
= ip6_flowinfo(iph
);
1165 dst
= ip6_route_output(net
, NULL
, &fl6
);
1167 ip6_rt_update_pmtu(dst
, NULL
, skb
, ntohl(mtu
));
1170 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1172 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1174 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1175 sk
->sk_bound_dev_if
, sk
->sk_mark
);
1177 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1179 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
)
1181 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1182 struct dst_entry
*dst
;
1185 memset(&fl6
, 0, sizeof(fl6
));
1186 fl6
.flowi6_oif
= oif
;
1187 fl6
.flowi6_mark
= mark
;
1188 fl6
.flowi6_flags
= 0;
1189 fl6
.daddr
= iph
->daddr
;
1190 fl6
.saddr
= iph
->saddr
;
1191 fl6
.flowlabel
= ip6_flowinfo(iph
);
1193 dst
= ip6_route_output(net
, NULL
, &fl6
);
1195 rt6_do_redirect(dst
, NULL
, skb
);
1198 EXPORT_SYMBOL_GPL(ip6_redirect
);
1200 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1202 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
);
1204 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1206 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1208 struct net_device
*dev
= dst
->dev
;
1209 unsigned int mtu
= dst_mtu(dst
);
1210 struct net
*net
= dev_net(dev
);
1212 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1214 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1215 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1218 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1219 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1220 * IPV6_MAXPLEN is also valid and means: "any MSS,
1221 * rely only on pmtu discovery"
1223 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1228 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1230 struct inet6_dev
*idev
;
1231 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1239 idev
= __in6_dev_get(dst
->dev
);
1241 mtu
= idev
->cnf
.mtu6
;
1247 static struct dst_entry
*icmp6_dst_gc_list
;
1248 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1250 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1251 struct neighbour
*neigh
,
1254 struct dst_entry
*dst
;
1255 struct rt6_info
*rt
;
1256 struct inet6_dev
*idev
= in6_dev_get(dev
);
1257 struct net
*net
= dev_net(dev
);
1259 if (unlikely(!idev
))
1260 return ERR_PTR(-ENODEV
);
1262 rt
= ip6_dst_alloc(net
, dev
, 0, NULL
);
1263 if (unlikely(!rt
)) {
1265 dst
= ERR_PTR(-ENOMEM
);
1272 neigh
= ip6_neigh_lookup(&rt
->dst
, NULL
, &fl6
->daddr
);
1273 if (IS_ERR(neigh
)) {
1276 return ERR_CAST(neigh
);
1280 rt
->dst
.flags
|= DST_HOST
;
1281 rt
->dst
.output
= ip6_output
;
1283 atomic_set(&rt
->dst
.__refcnt
, 1);
1284 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1285 rt
->rt6i_dst
.plen
= 128;
1286 rt
->rt6i_idev
= idev
;
1287 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1289 spin_lock_bh(&icmp6_dst_lock
);
1290 rt
->dst
.next
= icmp6_dst_gc_list
;
1291 icmp6_dst_gc_list
= &rt
->dst
;
1292 spin_unlock_bh(&icmp6_dst_lock
);
1294 fib6_force_start_gc(net
);
1296 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1302 int icmp6_dst_gc(void)
1304 struct dst_entry
*dst
, **pprev
;
1307 spin_lock_bh(&icmp6_dst_lock
);
1308 pprev
= &icmp6_dst_gc_list
;
1310 while ((dst
= *pprev
) != NULL
) {
1311 if (!atomic_read(&dst
->__refcnt
)) {
1320 spin_unlock_bh(&icmp6_dst_lock
);
1325 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1328 struct dst_entry
*dst
, **pprev
;
1330 spin_lock_bh(&icmp6_dst_lock
);
1331 pprev
= &icmp6_dst_gc_list
;
1332 while ((dst
= *pprev
) != NULL
) {
1333 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1334 if (func(rt
, arg
)) {
1341 spin_unlock_bh(&icmp6_dst_lock
);
1344 static int ip6_dst_gc(struct dst_ops
*ops
)
1346 unsigned long now
= jiffies
;
1347 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1348 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1349 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1350 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1351 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1352 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1355 entries
= dst_entries_get_fast(ops
);
1356 if (time_after(rt_last_gc
+ rt_min_interval
, now
) &&
1357 entries
<= rt_max_size
)
1360 net
->ipv6
.ip6_rt_gc_expire
++;
1361 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
);
1362 net
->ipv6
.ip6_rt_last_gc
= now
;
1363 entries
= dst_entries_get_slow(ops
);
1364 if (entries
< ops
->gc_thresh
)
1365 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1367 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1368 return entries
> rt_max_size
;
1371 int ip6_dst_hoplimit(struct dst_entry
*dst
)
1373 int hoplimit
= dst_metric_raw(dst
, RTAX_HOPLIMIT
);
1374 if (hoplimit
== 0) {
1375 struct net_device
*dev
= dst
->dev
;
1376 struct inet6_dev
*idev
;
1379 idev
= __in6_dev_get(dev
);
1381 hoplimit
= idev
->cnf
.hop_limit
;
1383 hoplimit
= dev_net(dev
)->ipv6
.devconf_all
->hop_limit
;
1388 EXPORT_SYMBOL(ip6_dst_hoplimit
);
1394 int ip6_route_add(struct fib6_config
*cfg
)
1397 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1398 struct rt6_info
*rt
= NULL
;
1399 struct net_device
*dev
= NULL
;
1400 struct inet6_dev
*idev
= NULL
;
1401 struct fib6_table
*table
;
1404 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1406 #ifndef CONFIG_IPV6_SUBTREES
1407 if (cfg
->fc_src_len
)
1410 if (cfg
->fc_ifindex
) {
1412 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1415 idev
= in6_dev_get(dev
);
1420 if (cfg
->fc_metric
== 0)
1421 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1424 if (cfg
->fc_nlinfo
.nlh
&&
1425 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1426 table
= fib6_get_table(net
, cfg
->fc_table
);
1428 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1429 table
= fib6_new_table(net
, cfg
->fc_table
);
1432 table
= fib6_new_table(net
, cfg
->fc_table
);
1438 rt
= ip6_dst_alloc(net
, NULL
, DST_NOCOUNT
, table
);
1445 if (cfg
->fc_flags
& RTF_EXPIRES
)
1446 rt6_set_expires(rt
, jiffies
+
1447 clock_t_to_jiffies(cfg
->fc_expires
));
1449 rt6_clean_expires(rt
);
1451 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1452 cfg
->fc_protocol
= RTPROT_BOOT
;
1453 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1455 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1457 if (addr_type
& IPV6_ADDR_MULTICAST
)
1458 rt
->dst
.input
= ip6_mc_input
;
1459 else if (cfg
->fc_flags
& RTF_LOCAL
)
1460 rt
->dst
.input
= ip6_input
;
1462 rt
->dst
.input
= ip6_forward
;
1464 rt
->dst
.output
= ip6_output
;
1466 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1467 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1468 if (rt
->rt6i_dst
.plen
== 128)
1469 rt
->dst
.flags
|= DST_HOST
;
1471 if (!(rt
->dst
.flags
& DST_HOST
) && cfg
->fc_mx
) {
1472 u32
*metrics
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_KERNEL
);
1477 dst_init_metrics(&rt
->dst
, metrics
, 0);
1479 #ifdef CONFIG_IPV6_SUBTREES
1480 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1481 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1484 rt
->rt6i_metric
= cfg
->fc_metric
;
1486 /* We cannot add true routes via loopback here,
1487 they would result in kernel looping; promote them to reject routes
1489 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1490 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1491 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1492 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1493 /* hold loopback dev/idev if we haven't done so. */
1494 if (dev
!= net
->loopback_dev
) {
1499 dev
= net
->loopback_dev
;
1501 idev
= in6_dev_get(dev
);
1507 rt
->dst
.output
= ip6_pkt_discard_out
;
1508 rt
->dst
.input
= ip6_pkt_discard
;
1509 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1510 switch (cfg
->fc_type
) {
1512 rt
->dst
.error
= -EINVAL
;
1515 rt
->dst
.error
= -EACCES
;
1518 rt
->dst
.error
= -EAGAIN
;
1521 rt
->dst
.error
= -ENETUNREACH
;
1527 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1528 const struct in6_addr
*gw_addr
;
1531 gw_addr
= &cfg
->fc_gateway
;
1532 rt
->rt6i_gateway
= *gw_addr
;
1533 gwa_type
= ipv6_addr_type(gw_addr
);
1535 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1536 struct rt6_info
*grt
;
1538 /* IPv6 strictly inhibits using not link-local
1539 addresses as nexthop address.
1540 Otherwise, router will not able to send redirects.
1541 It is very good, but in some (rare!) circumstances
1542 (SIT, PtP, NBMA NOARP links) it is handy to allow
1543 some exceptions. --ANK
1546 if (!(gwa_type
& IPV6_ADDR_UNICAST
))
1549 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1551 err
= -EHOSTUNREACH
;
1555 if (dev
!= grt
->dst
.dev
) {
1561 idev
= grt
->rt6i_idev
;
1563 in6_dev_hold(grt
->rt6i_idev
);
1565 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
1573 if (!dev
|| (dev
->flags
& IFF_LOOPBACK
))
1581 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
1582 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
1586 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
1587 rt
->rt6i_prefsrc
.plen
= 128;
1589 rt
->rt6i_prefsrc
.plen
= 0;
1591 if (cfg
->fc_flags
& (RTF_GATEWAY
| RTF_NONEXTHOP
)) {
1592 err
= rt6_bind_neighbour(rt
, dev
);
1597 rt
->rt6i_flags
= cfg
->fc_flags
;
1604 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1605 int type
= nla_type(nla
);
1608 if (type
> RTAX_MAX
) {
1613 dst_metric_set(&rt
->dst
, type
, nla_get_u32(nla
));
1619 rt
->rt6i_idev
= idev
;
1620 rt
->rt6i_table
= table
;
1622 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1624 return __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
);
1636 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1639 struct fib6_table
*table
;
1640 struct net
*net
= dev_net(rt
->dst
.dev
);
1642 if (rt
== net
->ipv6
.ip6_null_entry
) {
1647 table
= rt
->rt6i_table
;
1648 write_lock_bh(&table
->tb6_lock
);
1649 err
= fib6_del(rt
, info
);
1650 write_unlock_bh(&table
->tb6_lock
);
1657 int ip6_del_rt(struct rt6_info
*rt
)
1659 struct nl_info info
= {
1660 .nl_net
= dev_net(rt
->dst
.dev
),
1662 return __ip6_del_rt(rt
, &info
);
1665 static int ip6_route_del(struct fib6_config
*cfg
)
1667 struct fib6_table
*table
;
1668 struct fib6_node
*fn
;
1669 struct rt6_info
*rt
;
1672 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1676 read_lock_bh(&table
->tb6_lock
);
1678 fn
= fib6_locate(&table
->tb6_root
,
1679 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1680 &cfg
->fc_src
, cfg
->fc_src_len
);
1683 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1684 if (cfg
->fc_ifindex
&&
1686 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
1688 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1689 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1691 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1694 read_unlock_bh(&table
->tb6_lock
);
1696 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1699 read_unlock_bh(&table
->tb6_lock
);
1704 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
1706 struct net
*net
= dev_net(skb
->dev
);
1707 struct netevent_redirect netevent
;
1708 struct rt6_info
*rt
, *nrt
= NULL
;
1709 struct ndisc_options ndopts
;
1710 struct neighbour
*old_neigh
;
1711 struct inet6_dev
*in6_dev
;
1712 struct neighbour
*neigh
;
1714 int optlen
, on_link
;
1717 optlen
= skb
->tail
- skb
->transport_header
;
1718 optlen
-= sizeof(*msg
);
1721 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1725 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1727 if (ipv6_addr_is_multicast(&msg
->dest
)) {
1728 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1733 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
1735 } else if (ipv6_addr_type(&msg
->target
) !=
1736 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
1737 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1741 in6_dev
= __in6_dev_get(skb
->dev
);
1744 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
1748 * The IP source address of the Redirect MUST be the same as the current
1749 * first-hop router for the specified ICMP Destination Address.
1752 if (!ndisc_parse_options(msg
->opt
, optlen
, &ndopts
)) {
1753 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1758 if (ndopts
.nd_opts_tgt_lladdr
) {
1759 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
1762 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1767 rt
= (struct rt6_info
*) dst
;
1768 if (rt
== net
->ipv6
.ip6_null_entry
) {
1769 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1773 /* Redirect received -> path was valid.
1774 * Look, redirects are sent only in response to data packets,
1775 * so that this nexthop apparently is reachable. --ANK
1777 dst_confirm(&rt
->dst
);
1779 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
1783 /* Duplicate redirect: silently ignore. */
1785 if (neigh
== old_neigh
)
1789 * We have finally decided to accept it.
1792 neigh_update(neigh
, lladdr
, NUD_STALE
,
1793 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
1794 NEIGH_UPDATE_F_OVERRIDE
|
1795 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
1796 NEIGH_UPDATE_F_ISROUTER
))
1799 nrt
= ip6_rt_copy(rt
, &msg
->dest
);
1803 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
1805 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
1807 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
1808 nrt
->n
= neigh_clone(neigh
);
1810 if (ip6_ins_rt(nrt
))
1813 netevent
.old
= &rt
->dst
;
1814 netevent
.new = &nrt
->dst
;
1815 netevent
.daddr
= &msg
->dest
;
1816 netevent
.neigh
= neigh
;
1817 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
1819 if (rt
->rt6i_flags
& RTF_CACHE
) {
1820 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
1825 neigh_release(neigh
);
1829 * Misc support functions
1832 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
1833 const struct in6_addr
*dest
)
1835 struct net
*net
= dev_net(ort
->dst
.dev
);
1836 struct rt6_info
*rt
= ip6_dst_alloc(net
, ort
->dst
.dev
, 0,
1840 rt
->dst
.input
= ort
->dst
.input
;
1841 rt
->dst
.output
= ort
->dst
.output
;
1842 rt
->dst
.flags
|= DST_HOST
;
1844 rt
->rt6i_dst
.addr
= *dest
;
1845 rt
->rt6i_dst
.plen
= 128;
1846 dst_copy_metrics(&rt
->dst
, &ort
->dst
);
1847 rt
->dst
.error
= ort
->dst
.error
;
1848 rt
->rt6i_idev
= ort
->rt6i_idev
;
1850 in6_dev_hold(rt
->rt6i_idev
);
1851 rt
->dst
.lastuse
= jiffies
;
1853 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1854 rt
->rt6i_flags
= ort
->rt6i_flags
;
1855 if ((ort
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) ==
1856 (RTF_DEFAULT
| RTF_ADDRCONF
))
1857 rt6_set_from(rt
, ort
);
1859 rt6_clean_expires(rt
);
1860 rt
->rt6i_metric
= 0;
1862 #ifdef CONFIG_IPV6_SUBTREES
1863 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1865 memcpy(&rt
->rt6i_prefsrc
, &ort
->rt6i_prefsrc
, sizeof(struct rt6key
));
1866 rt
->rt6i_table
= ort
->rt6i_table
;
1871 #ifdef CONFIG_IPV6_ROUTE_INFO
1872 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
1873 const struct in6_addr
*prefix
, int prefixlen
,
1874 const struct in6_addr
*gwaddr
, int ifindex
)
1876 struct fib6_node
*fn
;
1877 struct rt6_info
*rt
= NULL
;
1878 struct fib6_table
*table
;
1880 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
1884 read_lock_bh(&table
->tb6_lock
);
1885 fn
= fib6_locate(&table
->tb6_root
, prefix
,prefixlen
, NULL
, 0);
1889 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1890 if (rt
->dst
.dev
->ifindex
!= ifindex
)
1892 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
1894 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
1900 read_unlock_bh(&table
->tb6_lock
);
1904 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
1905 const struct in6_addr
*prefix
, int prefixlen
,
1906 const struct in6_addr
*gwaddr
, int ifindex
,
1909 struct fib6_config cfg
= {
1910 .fc_table
= RT6_TABLE_INFO
,
1911 .fc_metric
= IP6_RT_PRIO_USER
,
1912 .fc_ifindex
= ifindex
,
1913 .fc_dst_len
= prefixlen
,
1914 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
1915 RTF_UP
| RTF_PREF(pref
),
1916 .fc_nlinfo
.portid
= 0,
1917 .fc_nlinfo
.nlh
= NULL
,
1918 .fc_nlinfo
.nl_net
= net
,
1921 cfg
.fc_dst
= *prefix
;
1922 cfg
.fc_gateway
= *gwaddr
;
1924 /* We should treat it as a default route if prefix length is 0. */
1926 cfg
.fc_flags
|= RTF_DEFAULT
;
1928 ip6_route_add(&cfg
);
1930 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
1934 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
1936 struct rt6_info
*rt
;
1937 struct fib6_table
*table
;
1939 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
1943 read_lock_bh(&table
->tb6_lock
);
1944 for (rt
= table
->tb6_root
.leaf
; rt
; rt
=rt
->dst
.rt6_next
) {
1945 if (dev
== rt
->dst
.dev
&&
1946 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
1947 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
1952 read_unlock_bh(&table
->tb6_lock
);
1956 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
1957 struct net_device
*dev
,
1960 struct fib6_config cfg
= {
1961 .fc_table
= RT6_TABLE_DFLT
,
1962 .fc_metric
= IP6_RT_PRIO_USER
,
1963 .fc_ifindex
= dev
->ifindex
,
1964 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
1965 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
1966 .fc_nlinfo
.portid
= 0,
1967 .fc_nlinfo
.nlh
= NULL
,
1968 .fc_nlinfo
.nl_net
= dev_net(dev
),
1971 cfg
.fc_gateway
= *gwaddr
;
1973 ip6_route_add(&cfg
);
1975 return rt6_get_dflt_router(gwaddr
, dev
);
1978 void rt6_purge_dflt_routers(struct net
*net
)
1980 struct rt6_info
*rt
;
1981 struct fib6_table
*table
;
1983 /* NOTE: Keep consistent with rt6_get_dflt_router */
1984 table
= fib6_get_table(net
, 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 (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) {
1993 read_unlock_bh(&table
->tb6_lock
);
1998 read_unlock_bh(&table
->tb6_lock
);
2001 static void rtmsg_to_fib6_config(struct net
*net
,
2002 struct in6_rtmsg
*rtmsg
,
2003 struct fib6_config
*cfg
)
2005 memset(cfg
, 0, sizeof(*cfg
));
2007 cfg
->fc_table
= RT6_TABLE_MAIN
;
2008 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
2009 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
2010 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
2011 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
2012 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
2013 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
2015 cfg
->fc_nlinfo
.nl_net
= net
;
2017 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
2018 cfg
->fc_src
= rtmsg
->rtmsg_src
;
2019 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
2022 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
2024 struct fib6_config cfg
;
2025 struct in6_rtmsg rtmsg
;
2029 case SIOCADDRT
: /* Add a route */
2030 case SIOCDELRT
: /* Delete a route */
2031 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2033 err
= copy_from_user(&rtmsg
, arg
,
2034 sizeof(struct in6_rtmsg
));
2038 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
2043 err
= ip6_route_add(&cfg
);
2046 err
= ip6_route_del(&cfg
);
2060 * Drop the packet on the floor
2063 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2066 struct dst_entry
*dst
= skb_dst(skb
);
2067 switch (ipstats_mib_noroutes
) {
2068 case IPSTATS_MIB_INNOROUTES
:
2069 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2070 if (type
== IPV6_ADDR_ANY
) {
2071 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2072 IPSTATS_MIB_INADDRERRORS
);
2076 case IPSTATS_MIB_OUTNOROUTES
:
2077 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2078 ipstats_mib_noroutes
);
2081 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2086 static int ip6_pkt_discard(struct sk_buff
*skb
)
2088 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2091 static int ip6_pkt_discard_out(struct sk_buff
*skb
)
2093 skb
->dev
= skb_dst(skb
)->dev
;
2094 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2097 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2099 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2101 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2104 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
)
2106 skb
->dev
= skb_dst(skb
)->dev
;
2107 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2113 * Allocate a dst for local (unicast / anycast) address.
2116 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2117 const struct in6_addr
*addr
,
2120 struct net
*net
= dev_net(idev
->dev
);
2121 struct rt6_info
*rt
= ip6_dst_alloc(net
, net
->loopback_dev
, 0, NULL
);
2125 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2126 return ERR_PTR(-ENOMEM
);
2131 rt
->dst
.flags
|= DST_HOST
;
2132 rt
->dst
.input
= ip6_input
;
2133 rt
->dst
.output
= ip6_output
;
2134 rt
->rt6i_idev
= idev
;
2136 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2138 rt
->rt6i_flags
|= RTF_ANYCAST
;
2140 rt
->rt6i_flags
|= RTF_LOCAL
;
2141 err
= rt6_bind_neighbour(rt
, rt
->dst
.dev
);
2144 return ERR_PTR(err
);
2147 rt
->rt6i_dst
.addr
= *addr
;
2148 rt
->rt6i_dst
.plen
= 128;
2149 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2151 atomic_set(&rt
->dst
.__refcnt
, 1);
2156 int ip6_route_get_saddr(struct net
*net
,
2157 struct rt6_info
*rt
,
2158 const struct in6_addr
*daddr
,
2160 struct in6_addr
*saddr
)
2162 struct inet6_dev
*idev
= ip6_dst_idev((struct dst_entry
*)rt
);
2164 if (rt
->rt6i_prefsrc
.plen
)
2165 *saddr
= rt
->rt6i_prefsrc
.addr
;
2167 err
= ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2168 daddr
, prefs
, saddr
);
2172 /* remove deleted ip from prefsrc entries */
2173 struct arg_dev_net_ip
{
2174 struct net_device
*dev
;
2176 struct in6_addr
*addr
;
2179 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2181 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2182 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2183 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2185 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2186 rt
!= net
->ipv6
.ip6_null_entry
&&
2187 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2188 /* remove prefsrc entry */
2189 rt
->rt6i_prefsrc
.plen
= 0;
2194 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2196 struct net
*net
= dev_net(ifp
->idev
->dev
);
2197 struct arg_dev_net_ip adni
= {
2198 .dev
= ifp
->idev
->dev
,
2202 fib6_clean_all(net
, fib6_remove_prefsrc
, 0, &adni
);
2205 struct arg_dev_net
{
2206 struct net_device
*dev
;
2210 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2212 const struct arg_dev_net
*adn
= arg
;
2213 const struct net_device
*dev
= adn
->dev
;
2215 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2216 rt
!= adn
->net
->ipv6
.ip6_null_entry
)
2222 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2224 struct arg_dev_net adn
= {
2229 fib6_clean_all(net
, fib6_ifdown
, 0, &adn
);
2230 icmp6_clean_all(fib6_ifdown
, &adn
);
2233 struct rt6_mtu_change_arg
{
2234 struct net_device
*dev
;
2238 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2240 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2241 struct inet6_dev
*idev
;
2243 /* In IPv6 pmtu discovery is not optional,
2244 so that RTAX_MTU lock cannot disable it.
2245 We still use this lock to block changes
2246 caused by addrconf/ndisc.
2249 idev
= __in6_dev_get(arg
->dev
);
2253 /* For administrative MTU increase, there is no way to discover
2254 IPv6 PMTU increase, so PMTU increase should be updated here.
2255 Since RFC 1981 doesn't include administrative MTU increase
2256 update PMTU increase is a MUST. (i.e. jumbo frame)
2259 If new MTU is less than route PMTU, this new MTU will be the
2260 lowest MTU in the path, update the route PMTU to reflect PMTU
2261 decreases; if new MTU is greater than route PMTU, and the
2262 old MTU is the lowest MTU in the path, update the route PMTU
2263 to reflect the increase. In this case if the other nodes' MTU
2264 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2267 if (rt
->dst
.dev
== arg
->dev
&&
2268 !dst_metric_locked(&rt
->dst
, RTAX_MTU
) &&
2269 (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2270 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2271 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
))) {
2272 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2277 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2279 struct rt6_mtu_change_arg arg
= {
2284 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, 0, &arg
);
2287 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2288 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2289 [RTA_OIF
] = { .type
= NLA_U32
},
2290 [RTA_IIF
] = { .type
= NLA_U32
},
2291 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2292 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2293 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2296 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2297 struct fib6_config
*cfg
)
2300 struct nlattr
*tb
[RTA_MAX
+1];
2303 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2308 rtm
= nlmsg_data(nlh
);
2309 memset(cfg
, 0, sizeof(*cfg
));
2311 cfg
->fc_table
= rtm
->rtm_table
;
2312 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2313 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2314 cfg
->fc_flags
= RTF_UP
;
2315 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2316 cfg
->fc_type
= rtm
->rtm_type
;
2318 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2319 rtm
->rtm_type
== RTN_BLACKHOLE
||
2320 rtm
->rtm_type
== RTN_PROHIBIT
||
2321 rtm
->rtm_type
== RTN_THROW
)
2322 cfg
->fc_flags
|= RTF_REJECT
;
2324 if (rtm
->rtm_type
== RTN_LOCAL
)
2325 cfg
->fc_flags
|= RTF_LOCAL
;
2327 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2328 cfg
->fc_nlinfo
.nlh
= nlh
;
2329 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2331 if (tb
[RTA_GATEWAY
]) {
2332 nla_memcpy(&cfg
->fc_gateway
, tb
[RTA_GATEWAY
], 16);
2333 cfg
->fc_flags
|= RTF_GATEWAY
;
2337 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2339 if (nla_len(tb
[RTA_DST
]) < plen
)
2342 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2346 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2348 if (nla_len(tb
[RTA_SRC
]) < plen
)
2351 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2354 if (tb
[RTA_PREFSRC
])
2355 nla_memcpy(&cfg
->fc_prefsrc
, tb
[RTA_PREFSRC
], 16);
2358 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2360 if (tb
[RTA_PRIORITY
])
2361 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2363 if (tb
[RTA_METRICS
]) {
2364 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2365 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2369 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2371 if (tb
[RTA_MULTIPATH
]) {
2372 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2373 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2381 static int ip6_route_multipath(struct fib6_config
*cfg
, int add
)
2383 struct fib6_config r_cfg
;
2384 struct rtnexthop
*rtnh
;
2387 int err
= 0, last_err
= 0;
2390 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
2391 remaining
= cfg
->fc_mp_len
;
2393 /* Parse a Multipath Entry */
2394 while (rtnh_ok(rtnh
, remaining
)) {
2395 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
2396 if (rtnh
->rtnh_ifindex
)
2397 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
2399 attrlen
= rtnh_attrlen(rtnh
);
2401 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
2403 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
2405 nla_memcpy(&r_cfg
.fc_gateway
, nla
, 16);
2406 r_cfg
.fc_flags
|= RTF_GATEWAY
;
2409 err
= add
? ip6_route_add(&r_cfg
) : ip6_route_del(&r_cfg
);
2412 /* If we are trying to remove a route, do not stop the
2413 * loop when ip6_route_del() fails (because next hop is
2414 * already gone), we should try to remove all next hops.
2417 /* If add fails, we should try to delete all
2418 * next hops that have been already added.
2424 /* Because each route is added like a single route we remove
2425 * this flag after the first nexthop (if there is a collision,
2426 * we have already fail to add the first nexthop:
2427 * fib6_add_rt2node() has reject it).
2429 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~NLM_F_EXCL
;
2430 rtnh
= rtnh_next(rtnh
, &remaining
);
2436 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
2438 struct fib6_config cfg
;
2441 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2446 return ip6_route_multipath(&cfg
, 0);
2448 return ip6_route_del(&cfg
);
2451 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
2453 struct fib6_config cfg
;
2456 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2461 return ip6_route_multipath(&cfg
, 1);
2463 return ip6_route_add(&cfg
);
2466 static inline size_t rt6_nlmsg_size(void)
2468 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2469 + nla_total_size(16) /* RTA_SRC */
2470 + nla_total_size(16) /* RTA_DST */
2471 + nla_total_size(16) /* RTA_GATEWAY */
2472 + nla_total_size(16) /* RTA_PREFSRC */
2473 + nla_total_size(4) /* RTA_TABLE */
2474 + nla_total_size(4) /* RTA_IIF */
2475 + nla_total_size(4) /* RTA_OIF */
2476 + nla_total_size(4) /* RTA_PRIORITY */
2477 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2478 + nla_total_size(sizeof(struct rta_cacheinfo
));
2481 static int rt6_fill_node(struct net
*net
,
2482 struct sk_buff
*skb
, struct rt6_info
*rt
,
2483 struct in6_addr
*dst
, struct in6_addr
*src
,
2484 int iif
, int type
, u32 portid
, u32 seq
,
2485 int prefix
, int nowait
, unsigned int flags
)
2488 struct nlmsghdr
*nlh
;
2492 if (prefix
) { /* user wants prefix routes only */
2493 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2494 /* success since this is not a prefix route */
2499 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
2503 rtm
= nlmsg_data(nlh
);
2504 rtm
->rtm_family
= AF_INET6
;
2505 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2506 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2509 table
= rt
->rt6i_table
->tb6_id
;
2511 table
= RT6_TABLE_UNSPEC
;
2512 rtm
->rtm_table
= table
;
2513 if (nla_put_u32(skb
, RTA_TABLE
, table
))
2514 goto nla_put_failure
;
2515 if (rt
->rt6i_flags
& RTF_REJECT
) {
2516 switch (rt
->dst
.error
) {
2518 rtm
->rtm_type
= RTN_BLACKHOLE
;
2521 rtm
->rtm_type
= RTN_PROHIBIT
;
2524 rtm
->rtm_type
= RTN_THROW
;
2527 rtm
->rtm_type
= RTN_UNREACHABLE
;
2531 else if (rt
->rt6i_flags
& RTF_LOCAL
)
2532 rtm
->rtm_type
= RTN_LOCAL
;
2533 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
2534 rtm
->rtm_type
= RTN_LOCAL
;
2536 rtm
->rtm_type
= RTN_UNICAST
;
2538 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2539 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2540 if (rt
->rt6i_flags
& RTF_DYNAMIC
)
2541 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2542 else if (rt
->rt6i_flags
& RTF_ADDRCONF
) {
2543 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ROUTEINFO
))
2544 rtm
->rtm_protocol
= RTPROT_RA
;
2546 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2549 if (rt
->rt6i_flags
& RTF_CACHE
)
2550 rtm
->rtm_flags
|= RTM_F_CLONED
;
2553 if (nla_put(skb
, RTA_DST
, 16, dst
))
2554 goto nla_put_failure
;
2555 rtm
->rtm_dst_len
= 128;
2556 } else if (rtm
->rtm_dst_len
)
2557 if (nla_put(skb
, RTA_DST
, 16, &rt
->rt6i_dst
.addr
))
2558 goto nla_put_failure
;
2559 #ifdef CONFIG_IPV6_SUBTREES
2561 if (nla_put(skb
, RTA_SRC
, 16, src
))
2562 goto nla_put_failure
;
2563 rtm
->rtm_src_len
= 128;
2564 } else if (rtm
->rtm_src_len
&&
2565 nla_put(skb
, RTA_SRC
, 16, &rt
->rt6i_src
.addr
))
2566 goto nla_put_failure
;
2569 #ifdef CONFIG_IPV6_MROUTE
2570 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2571 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2576 goto nla_put_failure
;
2578 if (err
== -EMSGSIZE
)
2579 goto nla_put_failure
;
2584 if (nla_put_u32(skb
, RTA_IIF
, iif
))
2585 goto nla_put_failure
;
2587 struct in6_addr saddr_buf
;
2588 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
2589 nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2590 goto nla_put_failure
;
2593 if (rt
->rt6i_prefsrc
.plen
) {
2594 struct in6_addr saddr_buf
;
2595 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
2596 if (nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2597 goto nla_put_failure
;
2600 if (rtnetlink_put_metrics(skb
, dst_metrics_ptr(&rt
->dst
)) < 0)
2601 goto nla_put_failure
;
2603 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2604 if (nla_put(skb
, RTA_GATEWAY
, 16, &rt
->rt6i_gateway
) < 0)
2605 goto nla_put_failure
;
2609 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2610 goto nla_put_failure
;
2611 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
2612 goto nla_put_failure
;
2614 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
2616 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
2617 goto nla_put_failure
;
2619 return nlmsg_end(skb
, nlh
);
2622 nlmsg_cancel(skb
, nlh
);
2626 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2628 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2631 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2632 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2633 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2637 return rt6_fill_node(arg
->net
,
2638 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2639 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
2640 prefix
, 0, NLM_F_MULTI
);
2643 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
* nlh
, void *arg
)
2645 struct net
*net
= sock_net(in_skb
->sk
);
2646 struct nlattr
*tb
[RTA_MAX
+1];
2647 struct rt6_info
*rt
;
2648 struct sk_buff
*skb
;
2651 int err
, iif
= 0, oif
= 0;
2653 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2658 memset(&fl6
, 0, sizeof(fl6
));
2661 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2664 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
2668 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
2671 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
2675 iif
= nla_get_u32(tb
[RTA_IIF
]);
2678 oif
= nla_get_u32(tb
[RTA_OIF
]);
2681 struct net_device
*dev
;
2684 dev
= __dev_get_by_index(net
, iif
);
2690 fl6
.flowi6_iif
= iif
;
2692 if (!ipv6_addr_any(&fl6
.saddr
))
2693 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
2695 rt
= (struct rt6_info
*)ip6_route_input_lookup(net
, dev
, &fl6
,
2698 fl6
.flowi6_oif
= oif
;
2700 rt
= (struct rt6_info
*)ip6_route_output(net
, NULL
, &fl6
);
2703 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2710 /* Reserve room for dummy headers, this skb can pass
2711 through good chunk of routing engine.
2713 skb_reset_mac_header(skb
);
2714 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
2716 skb_dst_set(skb
, &rt
->dst
);
2718 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
2719 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
2720 nlh
->nlmsg_seq
, 0, 0, 0);
2726 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2731 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
2733 struct sk_buff
*skb
;
2734 struct net
*net
= info
->nl_net
;
2739 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
2741 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
2745 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
2746 event
, info
->portid
, seq
, 0, 0, 0);
2748 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2749 WARN_ON(err
== -EMSGSIZE
);
2753 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
2754 info
->nlh
, gfp_any());
2758 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
2761 static int ip6_route_dev_notify(struct notifier_block
*this,
2762 unsigned long event
, void *data
)
2764 struct net_device
*dev
= (struct net_device
*)data
;
2765 struct net
*net
= dev_net(dev
);
2767 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
2768 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
2769 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
2770 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2771 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
2772 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
2773 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
2774 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
2785 #ifdef CONFIG_PROC_FS
2796 static int rt6_info_route(struct rt6_info
*rt
, void *p_arg
)
2798 struct seq_file
*m
= p_arg
;
2800 seq_printf(m
, "%pi6 %02x ", &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
);
2802 #ifdef CONFIG_IPV6_SUBTREES
2803 seq_printf(m
, "%pi6 %02x ", &rt
->rt6i_src
.addr
, rt
->rt6i_src
.plen
);
2805 seq_puts(m
, "00000000000000000000000000000000 00 ");
2807 if (rt
->rt6i_flags
& RTF_GATEWAY
) {
2808 seq_printf(m
, "%pi6", &rt
->rt6i_gateway
);
2810 seq_puts(m
, "00000000000000000000000000000000");
2812 seq_printf(m
, " %08x %08x %08x %08x %8s\n",
2813 rt
->rt6i_metric
, atomic_read(&rt
->dst
.__refcnt
),
2814 rt
->dst
.__use
, rt
->rt6i_flags
,
2815 rt
->dst
.dev
? rt
->dst
.dev
->name
: "");
2819 static int ipv6_route_show(struct seq_file
*m
, void *v
)
2821 struct net
*net
= (struct net
*)m
->private;
2822 fib6_clean_all_ro(net
, rt6_info_route
, 0, m
);
2826 static int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2828 return single_open_net(inode
, file
, ipv6_route_show
);
2831 static const struct file_operations ipv6_route_proc_fops
= {
2832 .owner
= THIS_MODULE
,
2833 .open
= ipv6_route_open
,
2835 .llseek
= seq_lseek
,
2836 .release
= single_release_net
,
2839 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
2841 struct net
*net
= (struct net
*)seq
->private;
2842 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
2843 net
->ipv6
.rt6_stats
->fib_nodes
,
2844 net
->ipv6
.rt6_stats
->fib_route_nodes
,
2845 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
2846 net
->ipv6
.rt6_stats
->fib_rt_entries
,
2847 net
->ipv6
.rt6_stats
->fib_rt_cache
,
2848 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
2849 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
2854 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
2856 return single_open_net(inode
, file
, rt6_stats_seq_show
);
2859 static const struct file_operations rt6_stats_seq_fops
= {
2860 .owner
= THIS_MODULE
,
2861 .open
= rt6_stats_seq_open
,
2863 .llseek
= seq_lseek
,
2864 .release
= single_release_net
,
2866 #endif /* CONFIG_PROC_FS */
2868 #ifdef CONFIG_SYSCTL
2871 int ipv6_sysctl_rtcache_flush(ctl_table
*ctl
, int write
,
2872 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2879 net
= (struct net
*)ctl
->extra1
;
2880 delay
= net
->ipv6
.sysctl
.flush_delay
;
2881 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
2882 fib6_run_gc(delay
<= 0 ? ~0UL : (unsigned long)delay
, net
);
2886 ctl_table ipv6_route_table_template
[] = {
2888 .procname
= "flush",
2889 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
2890 .maxlen
= sizeof(int),
2892 .proc_handler
= ipv6_sysctl_rtcache_flush
2895 .procname
= "gc_thresh",
2896 .data
= &ip6_dst_ops_template
.gc_thresh
,
2897 .maxlen
= sizeof(int),
2899 .proc_handler
= proc_dointvec
,
2902 .procname
= "max_size",
2903 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
2904 .maxlen
= sizeof(int),
2906 .proc_handler
= proc_dointvec
,
2909 .procname
= "gc_min_interval",
2910 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2911 .maxlen
= sizeof(int),
2913 .proc_handler
= proc_dointvec_jiffies
,
2916 .procname
= "gc_timeout",
2917 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
2918 .maxlen
= sizeof(int),
2920 .proc_handler
= proc_dointvec_jiffies
,
2923 .procname
= "gc_interval",
2924 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
2925 .maxlen
= sizeof(int),
2927 .proc_handler
= proc_dointvec_jiffies
,
2930 .procname
= "gc_elasticity",
2931 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
2932 .maxlen
= sizeof(int),
2934 .proc_handler
= proc_dointvec
,
2937 .procname
= "mtu_expires",
2938 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
2939 .maxlen
= sizeof(int),
2941 .proc_handler
= proc_dointvec_jiffies
,
2944 .procname
= "min_adv_mss",
2945 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
2946 .maxlen
= sizeof(int),
2948 .proc_handler
= proc_dointvec
,
2951 .procname
= "gc_min_interval_ms",
2952 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2953 .maxlen
= sizeof(int),
2955 .proc_handler
= proc_dointvec_ms_jiffies
,
2960 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
2962 struct ctl_table
*table
;
2964 table
= kmemdup(ipv6_route_table_template
,
2965 sizeof(ipv6_route_table_template
),
2969 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
2970 table
[0].extra1
= net
;
2971 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
2972 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
2973 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2974 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2975 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
2976 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2977 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
2978 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2979 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2981 /* Don't export sysctls to unprivileged users */
2982 if (net
->user_ns
!= &init_user_ns
)
2983 table
[0].procname
= NULL
;
2990 static int __net_init
ip6_route_net_init(struct net
*net
)
2994 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
2995 sizeof(net
->ipv6
.ip6_dst_ops
));
2997 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
2998 goto out_ip6_dst_ops
;
3000 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
3001 sizeof(*net
->ipv6
.ip6_null_entry
),
3003 if (!net
->ipv6
.ip6_null_entry
)
3004 goto out_ip6_dst_entries
;
3005 net
->ipv6
.ip6_null_entry
->dst
.path
=
3006 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
3007 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3008 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
3009 ip6_template_metrics
, true);
3011 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3012 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
3013 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
3015 if (!net
->ipv6
.ip6_prohibit_entry
)
3016 goto out_ip6_null_entry
;
3017 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
3018 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
3019 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3020 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
3021 ip6_template_metrics
, true);
3023 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
3024 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
3026 if (!net
->ipv6
.ip6_blk_hole_entry
)
3027 goto out_ip6_prohibit_entry
;
3028 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
3029 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
3030 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3031 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
3032 ip6_template_metrics
, true);
3035 net
->ipv6
.sysctl
.flush_delay
= 0;
3036 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
3037 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
3038 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
3039 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
3040 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
3041 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
3042 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
3044 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
3050 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3051 out_ip6_prohibit_entry
:
3052 kfree(net
->ipv6
.ip6_prohibit_entry
);
3054 kfree(net
->ipv6
.ip6_null_entry
);
3056 out_ip6_dst_entries
:
3057 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3062 static void __net_exit
ip6_route_net_exit(struct net
*net
)
3064 kfree(net
->ipv6
.ip6_null_entry
);
3065 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3066 kfree(net
->ipv6
.ip6_prohibit_entry
);
3067 kfree(net
->ipv6
.ip6_blk_hole_entry
);
3069 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3072 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3074 #ifdef CONFIG_PROC_FS
3075 proc_net_fops_create(net
, "ipv6_route", 0, &ipv6_route_proc_fops
);
3076 proc_net_fops_create(net
, "rt6_stats", S_IRUGO
, &rt6_stats_seq_fops
);
3081 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
3083 #ifdef CONFIG_PROC_FS
3084 proc_net_remove(net
, "ipv6_route");
3085 proc_net_remove(net
, "rt6_stats");
3089 static struct pernet_operations ip6_route_net_ops
= {
3090 .init
= ip6_route_net_init
,
3091 .exit
= ip6_route_net_exit
,
3094 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
3096 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
3100 inet_peer_base_init(bp
);
3101 net
->ipv6
.peers
= bp
;
3105 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
3107 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
3109 net
->ipv6
.peers
= NULL
;
3110 inetpeer_invalidate_tree(bp
);
3114 static struct pernet_operations ipv6_inetpeer_ops
= {
3115 .init
= ipv6_inetpeer_init
,
3116 .exit
= ipv6_inetpeer_exit
,
3119 static struct pernet_operations ip6_route_net_late_ops
= {
3120 .init
= ip6_route_net_init_late
,
3121 .exit
= ip6_route_net_exit_late
,
3124 static struct notifier_block ip6_route_dev_notifier
= {
3125 .notifier_call
= ip6_route_dev_notify
,
3129 int __init
ip6_route_init(void)
3134 ip6_dst_ops_template
.kmem_cachep
=
3135 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
3136 SLAB_HWCACHE_ALIGN
, NULL
);
3137 if (!ip6_dst_ops_template
.kmem_cachep
)
3140 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
3142 goto out_kmem_cache
;
3144 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
3146 goto out_dst_entries
;
3148 ret
= register_pernet_subsys(&ip6_route_net_ops
);
3150 goto out_register_inetpeer
;
3152 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
3154 /* Registering of the loopback is done before this portion of code,
3155 * the loopback reference in rt6_info will not be taken, do it
3156 * manually for init_net */
3157 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
3158 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3159 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3160 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
3161 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3162 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
3163 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3167 goto out_register_subsys
;
3173 ret
= fib6_rules_init();
3177 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
3179 goto fib6_rules_init
;
3182 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
3183 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
3184 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
3185 goto out_register_late_subsys
;
3187 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
3189 goto out_register_late_subsys
;
3194 out_register_late_subsys
:
3195 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3197 fib6_rules_cleanup();
3202 out_register_subsys
:
3203 unregister_pernet_subsys(&ip6_route_net_ops
);
3204 out_register_inetpeer
:
3205 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3207 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3209 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
3213 void ip6_route_cleanup(void)
3215 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
3216 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3217 fib6_rules_cleanup();
3220 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3221 unregister_pernet_subsys(&ip6_route_net_ops
);
3222 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3223 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);