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(&nd_tbl
, 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(&nd_tbl
, 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
;
502 if (!neigh
|| (neigh
->nud_state
& NUD_VALID
))
504 read_lock_bh(&neigh
->lock
);
505 if (!(neigh
->nud_state
& NUD_VALID
) &&
506 time_after(jiffies
, neigh
->updated
+ rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
507 struct in6_addr mcaddr
;
508 struct in6_addr
*target
;
510 neigh
->updated
= jiffies
;
511 read_unlock_bh(&neigh
->lock
);
513 target
= (struct in6_addr
*)&neigh
->primary_key
;
514 addrconf_addr_solict_mult(target
, &mcaddr
);
515 ndisc_send_ns(rt
->dst
.dev
, NULL
, target
, &mcaddr
, NULL
);
517 read_unlock_bh(&neigh
->lock
);
521 static inline void rt6_probe(struct rt6_info
*rt
)
527 * Default Router Selection (RFC 2461 6.3.6)
529 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
531 struct net_device
*dev
= rt
->dst
.dev
;
532 if (!oif
|| dev
->ifindex
== oif
)
534 if ((dev
->flags
& IFF_LOOPBACK
) &&
535 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
540 static inline bool rt6_check_neigh(struct rt6_info
*rt
)
542 struct neighbour
*neigh
;
546 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
547 !(rt
->rt6i_flags
& RTF_GATEWAY
))
550 read_lock_bh(&neigh
->lock
);
551 if (neigh
->nud_state
& NUD_VALID
)
553 #ifdef CONFIG_IPV6_ROUTER_PREF
554 else if (!(neigh
->nud_state
& NUD_FAILED
))
557 read_unlock_bh(&neigh
->lock
);
562 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
567 m
= rt6_check_dev(rt
, oif
);
568 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
570 #ifdef CONFIG_IPV6_ROUTER_PREF
571 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
573 if (!rt6_check_neigh(rt
) && (strict
& RT6_LOOKUP_F_REACHABLE
))
578 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
579 int *mpri
, struct rt6_info
*match
)
583 if (rt6_check_expired(rt
))
586 m
= rt6_score_route(rt
, oif
, strict
);
591 if (strict
& RT6_LOOKUP_F_REACHABLE
)
595 } else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
603 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
604 struct rt6_info
*rr_head
,
605 u32 metric
, int oif
, int strict
)
607 struct rt6_info
*rt
, *match
;
611 for (rt
= rr_head
; rt
&& rt
->rt6i_metric
== metric
;
612 rt
= rt
->dst
.rt6_next
)
613 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
614 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
&& rt
->rt6i_metric
== metric
;
615 rt
= rt
->dst
.rt6_next
)
616 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
621 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
623 struct rt6_info
*match
, *rt0
;
628 fn
->rr_ptr
= rt0
= fn
->leaf
;
630 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
);
633 (strict
& RT6_LOOKUP_F_REACHABLE
)) {
634 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
636 /* no entries matched; do round-robin */
637 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
644 net
= dev_net(rt0
->dst
.dev
);
645 return match
? match
: net
->ipv6
.ip6_null_entry
;
648 #ifdef CONFIG_IPV6_ROUTE_INFO
649 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
650 const struct in6_addr
*gwaddr
)
652 struct net
*net
= dev_net(dev
);
653 struct route_info
*rinfo
= (struct route_info
*) opt
;
654 struct in6_addr prefix_buf
, *prefix
;
656 unsigned long lifetime
;
659 if (len
< sizeof(struct route_info
)) {
663 /* Sanity check for prefix_len and length */
664 if (rinfo
->length
> 3) {
666 } else if (rinfo
->prefix_len
> 128) {
668 } else if (rinfo
->prefix_len
> 64) {
669 if (rinfo
->length
< 2) {
672 } else if (rinfo
->prefix_len
> 0) {
673 if (rinfo
->length
< 1) {
678 pref
= rinfo
->route_pref
;
679 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
682 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
684 if (rinfo
->length
== 3)
685 prefix
= (struct in6_addr
*)rinfo
->prefix
;
687 /* this function is safe */
688 ipv6_addr_prefix(&prefix_buf
,
689 (struct in6_addr
*)rinfo
->prefix
,
691 prefix
= &prefix_buf
;
694 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
,
697 if (rt
&& !lifetime
) {
703 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
706 rt
->rt6i_flags
= RTF_ROUTEINFO
|
707 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
710 if (!addrconf_finite_timeout(lifetime
))
711 rt6_clean_expires(rt
);
713 rt6_set_expires(rt
, jiffies
+ HZ
* lifetime
);
721 #define BACKTRACK(__net, saddr) \
723 if (rt == __net->ipv6.ip6_null_entry) { \
724 struct fib6_node *pn; \
726 if (fn->fn_flags & RTN_TL_ROOT) \
729 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
730 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
733 if (fn->fn_flags & RTN_RTINFO) \
739 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
740 struct fib6_table
*table
,
741 struct flowi6
*fl6
, int flags
)
743 struct fib6_node
*fn
;
746 read_lock_bh(&table
->tb6_lock
);
747 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
750 rt
= rt6_device_match(net
, rt
, &fl6
->saddr
, fl6
->flowi6_oif
, flags
);
751 if (rt
->rt6i_nsiblings
&& fl6
->flowi6_oif
== 0)
752 rt
= rt6_multipath_select(rt
, fl6
);
753 BACKTRACK(net
, &fl6
->saddr
);
755 dst_use(&rt
->dst
, jiffies
);
756 read_unlock_bh(&table
->tb6_lock
);
761 struct dst_entry
* ip6_route_lookup(struct net
*net
, struct flowi6
*fl6
,
764 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_lookup
);
766 EXPORT_SYMBOL_GPL(ip6_route_lookup
);
768 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
769 const struct in6_addr
*saddr
, int oif
, int strict
)
771 struct flowi6 fl6
= {
775 struct dst_entry
*dst
;
776 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
779 memcpy(&fl6
.saddr
, saddr
, sizeof(*saddr
));
780 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
783 dst
= fib6_rule_lookup(net
, &fl6
, flags
, ip6_pol_route_lookup
);
785 return (struct rt6_info
*) dst
;
792 EXPORT_SYMBOL(rt6_lookup
);
794 /* ip6_ins_rt is called with FREE table->tb6_lock.
795 It takes new route entry, the addition fails by any reason the
796 route is freed. In any case, if caller does not hold it, it may
800 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
)
803 struct fib6_table
*table
;
805 table
= rt
->rt6i_table
;
806 write_lock_bh(&table
->tb6_lock
);
807 err
= fib6_add(&table
->tb6_root
, rt
, info
);
808 write_unlock_bh(&table
->tb6_lock
);
813 int ip6_ins_rt(struct rt6_info
*rt
)
815 struct nl_info info
= {
816 .nl_net
= dev_net(rt
->dst
.dev
),
818 return __ip6_ins_rt(rt
, &info
);
821 static struct rt6_info
*rt6_alloc_cow(struct rt6_info
*ort
,
822 const struct in6_addr
*daddr
,
823 const struct in6_addr
*saddr
)
831 rt
= ip6_rt_copy(ort
, daddr
);
834 int attempts
= !in_softirq();
836 if (!(rt
->rt6i_flags
& RTF_GATEWAY
)) {
837 if (ort
->rt6i_dst
.plen
!= 128 &&
838 ipv6_addr_equal(&ort
->rt6i_dst
.addr
, daddr
))
839 rt
->rt6i_flags
|= RTF_ANYCAST
;
840 rt
->rt6i_gateway
= *daddr
;
843 rt
->rt6i_flags
|= RTF_CACHE
;
845 #ifdef CONFIG_IPV6_SUBTREES
846 if (rt
->rt6i_src
.plen
&& saddr
) {
847 rt
->rt6i_src
.addr
= *saddr
;
848 rt
->rt6i_src
.plen
= 128;
853 if (rt6_bind_neighbour(rt
, rt
->dst
.dev
)) {
854 struct net
*net
= dev_net(rt
->dst
.dev
);
855 int saved_rt_min_interval
=
856 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
857 int saved_rt_elasticity
=
858 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
860 if (attempts
-- > 0) {
861 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 1;
862 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= 0;
864 ip6_dst_gc(&net
->ipv6
.ip6_dst_ops
);
866 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
=
868 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
=
869 saved_rt_min_interval
;
873 net_warn_ratelimited("Neighbour table overflow\n");
882 static struct rt6_info
*rt6_alloc_clone(struct rt6_info
*ort
,
883 const struct in6_addr
*daddr
)
885 struct rt6_info
*rt
= ip6_rt_copy(ort
, daddr
);
888 rt
->rt6i_flags
|= RTF_CACHE
;
889 rt
->n
= neigh_clone(ort
->n
);
894 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
895 struct flowi6
*fl6
, int flags
)
897 struct fib6_node
*fn
;
898 struct rt6_info
*rt
, *nrt
;
902 int reachable
= net
->ipv6
.devconf_all
->forwarding
? 0 : RT6_LOOKUP_F_REACHABLE
;
904 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
907 read_lock_bh(&table
->tb6_lock
);
910 fn
= fib6_lookup(&table
->tb6_root
, &fl6
->daddr
, &fl6
->saddr
);
913 rt
= rt6_select(fn
, oif
, strict
| reachable
);
914 if (rt
->rt6i_nsiblings
&& oif
== 0)
915 rt
= rt6_multipath_select(rt
, fl6
);
916 BACKTRACK(net
, &fl6
->saddr
);
917 if (rt
== net
->ipv6
.ip6_null_entry
||
918 rt
->rt6i_flags
& RTF_CACHE
)
922 read_unlock_bh(&table
->tb6_lock
);
924 if (!rt
->n
&& !(rt
->rt6i_flags
& RTF_NONEXTHOP
))
925 nrt
= rt6_alloc_cow(rt
, &fl6
->daddr
, &fl6
->saddr
);
926 else if (!(rt
->dst
.flags
& DST_HOST
))
927 nrt
= rt6_alloc_clone(rt
, &fl6
->daddr
);
932 rt
= nrt
? : net
->ipv6
.ip6_null_entry
;
936 err
= ip6_ins_rt(nrt
);
945 * Race condition! In the gap, when table->tb6_lock was
946 * released someone could insert this route. Relookup.
957 read_unlock_bh(&table
->tb6_lock
);
959 rt
->dst
.lastuse
= jiffies
;
965 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
966 struct flowi6
*fl6
, int flags
)
968 return ip6_pol_route(net
, table
, fl6
->flowi6_iif
, fl6
, flags
);
971 static struct dst_entry
*ip6_route_input_lookup(struct net
*net
,
972 struct net_device
*dev
,
973 struct flowi6
*fl6
, int flags
)
975 if (rt6_need_strict(&fl6
->daddr
) && dev
->type
!= ARPHRD_PIMREG
)
976 flags
|= RT6_LOOKUP_F_IFACE
;
978 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_input
);
981 void ip6_route_input(struct sk_buff
*skb
)
983 const struct ipv6hdr
*iph
= ipv6_hdr(skb
);
984 struct net
*net
= dev_net(skb
->dev
);
985 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
986 struct flowi6 fl6
= {
987 .flowi6_iif
= skb
->dev
->ifindex
,
990 .flowlabel
= ip6_flowinfo(iph
),
991 .flowi6_mark
= skb
->mark
,
992 .flowi6_proto
= iph
->nexthdr
,
995 skb_dst_set(skb
, ip6_route_input_lookup(net
, skb
->dev
, &fl6
, flags
));
998 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
999 struct flowi6
*fl6
, int flags
)
1001 return ip6_pol_route(net
, table
, fl6
->flowi6_oif
, fl6
, flags
);
1004 struct dst_entry
* ip6_route_output(struct net
*net
, const struct sock
*sk
,
1009 fl6
->flowi6_iif
= LOOPBACK_IFINDEX
;
1011 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl6
->daddr
))
1012 flags
|= RT6_LOOKUP_F_IFACE
;
1014 if (!ipv6_addr_any(&fl6
->saddr
))
1015 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
1017 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
1019 return fib6_rule_lookup(net
, fl6
, flags
, ip6_pol_route_output
);
1022 EXPORT_SYMBOL(ip6_route_output
);
1024 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
1026 struct rt6_info
*rt
, *ort
= (struct rt6_info
*) dst_orig
;
1027 struct dst_entry
*new = NULL
;
1029 rt
= dst_alloc(&ip6_dst_blackhole_ops
, ort
->dst
.dev
, 1, DST_OBSOLETE_NONE
, 0);
1033 memset(new + 1, 0, sizeof(*rt
) - sizeof(*new));
1034 rt6_init_peer(rt
, net
->ipv6
.peers
);
1037 new->input
= dst_discard
;
1038 new->output
= dst_discard
;
1040 if (dst_metrics_read_only(&ort
->dst
))
1041 new->_metrics
= ort
->dst
._metrics
;
1043 dst_copy_metrics(new, &ort
->dst
);
1044 rt
->rt6i_idev
= ort
->rt6i_idev
;
1046 in6_dev_hold(rt
->rt6i_idev
);
1048 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1049 rt
->rt6i_flags
= ort
->rt6i_flags
;
1050 rt6_clean_expires(rt
);
1051 rt
->rt6i_metric
= 0;
1053 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1054 #ifdef CONFIG_IPV6_SUBTREES
1055 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1061 dst_release(dst_orig
);
1062 return new ? new : ERR_PTR(-ENOMEM
);
1066 * Destination cache support functions
1069 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
1071 struct rt6_info
*rt
;
1073 rt
= (struct rt6_info
*) dst
;
1075 /* All IPV6 dsts are created with ->obsolete set to the value
1076 * DST_OBSOLETE_FORCE_CHK which forces validation calls down
1077 * into this function always.
1079 if (rt
->rt6i_genid
!= rt_genid(dev_net(rt
->dst
.dev
)))
1082 if (rt
->rt6i_node
&& (rt
->rt6i_node
->fn_sernum
== cookie
))
1088 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
1090 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1093 if (rt
->rt6i_flags
& RTF_CACHE
) {
1094 if (rt6_check_expired(rt
)) {
1106 static void ip6_link_failure(struct sk_buff
*skb
)
1108 struct rt6_info
*rt
;
1110 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
1112 rt
= (struct rt6_info
*) skb_dst(skb
);
1114 if (rt
->rt6i_flags
& RTF_CACHE
)
1115 rt6_update_expires(rt
, 0);
1116 else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
))
1117 rt
->rt6i_node
->fn_sernum
= -1;
1121 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, struct sock
*sk
,
1122 struct sk_buff
*skb
, u32 mtu
)
1124 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
1127 if (mtu
< dst_mtu(dst
) && rt6
->rt6i_dst
.plen
== 128) {
1128 struct net
*net
= dev_net(dst
->dev
);
1130 rt6
->rt6i_flags
|= RTF_MODIFIED
;
1131 if (mtu
< IPV6_MIN_MTU
) {
1132 u32 features
= dst_metric(dst
, RTAX_FEATURES
);
1134 features
|= RTAX_FEATURE_ALLFRAG
;
1135 dst_metric_set(dst
, RTAX_FEATURES
, features
);
1137 dst_metric_set(dst
, RTAX_MTU
, mtu
);
1138 rt6_update_expires(rt6
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1142 void ip6_update_pmtu(struct sk_buff
*skb
, struct net
*net
, __be32 mtu
,
1145 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1146 struct dst_entry
*dst
;
1149 memset(&fl6
, 0, sizeof(fl6
));
1150 fl6
.flowi6_oif
= oif
;
1151 fl6
.flowi6_mark
= mark
;
1152 fl6
.flowi6_flags
= 0;
1153 fl6
.daddr
= iph
->daddr
;
1154 fl6
.saddr
= iph
->saddr
;
1155 fl6
.flowlabel
= ip6_flowinfo(iph
);
1157 dst
= ip6_route_output(net
, NULL
, &fl6
);
1159 ip6_rt_update_pmtu(dst
, NULL
, skb
, ntohl(mtu
));
1162 EXPORT_SYMBOL_GPL(ip6_update_pmtu
);
1164 void ip6_sk_update_pmtu(struct sk_buff
*skb
, struct sock
*sk
, __be32 mtu
)
1166 ip6_update_pmtu(skb
, sock_net(sk
), mtu
,
1167 sk
->sk_bound_dev_if
, sk
->sk_mark
);
1169 EXPORT_SYMBOL_GPL(ip6_sk_update_pmtu
);
1171 void ip6_redirect(struct sk_buff
*skb
, struct net
*net
, int oif
, u32 mark
)
1173 const struct ipv6hdr
*iph
= (struct ipv6hdr
*) skb
->data
;
1174 struct dst_entry
*dst
;
1177 memset(&fl6
, 0, sizeof(fl6
));
1178 fl6
.flowi6_oif
= oif
;
1179 fl6
.flowi6_mark
= mark
;
1180 fl6
.flowi6_flags
= 0;
1181 fl6
.daddr
= iph
->daddr
;
1182 fl6
.saddr
= iph
->saddr
;
1183 fl6
.flowlabel
= ip6_flowinfo(iph
);
1185 dst
= ip6_route_output(net
, NULL
, &fl6
);
1187 rt6_do_redirect(dst
, NULL
, skb
);
1190 EXPORT_SYMBOL_GPL(ip6_redirect
);
1192 void ip6_sk_redirect(struct sk_buff
*skb
, struct sock
*sk
)
1194 ip6_redirect(skb
, sock_net(sk
), sk
->sk_bound_dev_if
, sk
->sk_mark
);
1196 EXPORT_SYMBOL_GPL(ip6_sk_redirect
);
1198 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
1200 struct net_device
*dev
= dst
->dev
;
1201 unsigned int mtu
= dst_mtu(dst
);
1202 struct net
*net
= dev_net(dev
);
1204 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
1206 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
1207 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
1210 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
1211 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
1212 * IPV6_MAXPLEN is also valid and means: "any MSS,
1213 * rely only on pmtu discovery"
1215 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1220 static unsigned int ip6_mtu(const struct dst_entry
*dst
)
1222 struct inet6_dev
*idev
;
1223 unsigned int mtu
= dst_metric_raw(dst
, RTAX_MTU
);
1231 idev
= __in6_dev_get(dst
->dev
);
1233 mtu
= idev
->cnf
.mtu6
;
1239 static struct dst_entry
*icmp6_dst_gc_list
;
1240 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1242 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1243 struct neighbour
*neigh
,
1246 struct dst_entry
*dst
;
1247 struct rt6_info
*rt
;
1248 struct inet6_dev
*idev
= in6_dev_get(dev
);
1249 struct net
*net
= dev_net(dev
);
1251 if (unlikely(!idev
))
1252 return ERR_PTR(-ENODEV
);
1254 rt
= ip6_dst_alloc(net
, dev
, 0, NULL
);
1255 if (unlikely(!rt
)) {
1257 dst
= ERR_PTR(-ENOMEM
);
1264 neigh
= ip6_neigh_lookup(&rt
->dst
, NULL
, &fl6
->daddr
);
1265 if (IS_ERR(neigh
)) {
1268 return ERR_CAST(neigh
);
1272 rt
->dst
.flags
|= DST_HOST
;
1273 rt
->dst
.output
= ip6_output
;
1275 atomic_set(&rt
->dst
.__refcnt
, 1);
1276 rt
->rt6i_dst
.addr
= fl6
->daddr
;
1277 rt
->rt6i_dst
.plen
= 128;
1278 rt
->rt6i_idev
= idev
;
1279 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 0);
1281 spin_lock_bh(&icmp6_dst_lock
);
1282 rt
->dst
.next
= icmp6_dst_gc_list
;
1283 icmp6_dst_gc_list
= &rt
->dst
;
1284 spin_unlock_bh(&icmp6_dst_lock
);
1286 fib6_force_start_gc(net
);
1288 dst
= xfrm_lookup(net
, &rt
->dst
, flowi6_to_flowi(fl6
), NULL
, 0);
1294 int icmp6_dst_gc(void)
1296 struct dst_entry
*dst
, **pprev
;
1299 spin_lock_bh(&icmp6_dst_lock
);
1300 pprev
= &icmp6_dst_gc_list
;
1302 while ((dst
= *pprev
) != NULL
) {
1303 if (!atomic_read(&dst
->__refcnt
)) {
1312 spin_unlock_bh(&icmp6_dst_lock
);
1317 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1320 struct dst_entry
*dst
, **pprev
;
1322 spin_lock_bh(&icmp6_dst_lock
);
1323 pprev
= &icmp6_dst_gc_list
;
1324 while ((dst
= *pprev
) != NULL
) {
1325 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1326 if (func(rt
, arg
)) {
1333 spin_unlock_bh(&icmp6_dst_lock
);
1336 static int ip6_dst_gc(struct dst_ops
*ops
)
1338 unsigned long now
= jiffies
;
1339 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1340 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1341 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1342 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1343 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1344 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1347 entries
= dst_entries_get_fast(ops
);
1348 if (time_after(rt_last_gc
+ rt_min_interval
, now
) &&
1349 entries
<= rt_max_size
)
1352 net
->ipv6
.ip6_rt_gc_expire
++;
1353 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
);
1354 net
->ipv6
.ip6_rt_last_gc
= now
;
1355 entries
= dst_entries_get_slow(ops
);
1356 if (entries
< ops
->gc_thresh
)
1357 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1359 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1360 return entries
> rt_max_size
;
1363 int ip6_dst_hoplimit(struct dst_entry
*dst
)
1365 int hoplimit
= dst_metric_raw(dst
, RTAX_HOPLIMIT
);
1366 if (hoplimit
== 0) {
1367 struct net_device
*dev
= dst
->dev
;
1368 struct inet6_dev
*idev
;
1371 idev
= __in6_dev_get(dev
);
1373 hoplimit
= idev
->cnf
.hop_limit
;
1375 hoplimit
= dev_net(dev
)->ipv6
.devconf_all
->hop_limit
;
1380 EXPORT_SYMBOL(ip6_dst_hoplimit
);
1386 int ip6_route_add(struct fib6_config
*cfg
)
1389 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1390 struct rt6_info
*rt
= NULL
;
1391 struct net_device
*dev
= NULL
;
1392 struct inet6_dev
*idev
= NULL
;
1393 struct fib6_table
*table
;
1396 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1398 #ifndef CONFIG_IPV6_SUBTREES
1399 if (cfg
->fc_src_len
)
1402 if (cfg
->fc_ifindex
) {
1404 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1407 idev
= in6_dev_get(dev
);
1412 if (cfg
->fc_metric
== 0)
1413 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1416 if (cfg
->fc_nlinfo
.nlh
&&
1417 !(cfg
->fc_nlinfo
.nlh
->nlmsg_flags
& NLM_F_CREATE
)) {
1418 table
= fib6_get_table(net
, cfg
->fc_table
);
1420 pr_warn("NLM_F_CREATE should be specified when creating new route\n");
1421 table
= fib6_new_table(net
, cfg
->fc_table
);
1424 table
= fib6_new_table(net
, cfg
->fc_table
);
1430 rt
= ip6_dst_alloc(net
, NULL
, DST_NOCOUNT
, table
);
1437 if (cfg
->fc_flags
& RTF_EXPIRES
)
1438 rt6_set_expires(rt
, jiffies
+
1439 clock_t_to_jiffies(cfg
->fc_expires
));
1441 rt6_clean_expires(rt
);
1443 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1444 cfg
->fc_protocol
= RTPROT_BOOT
;
1445 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1447 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1449 if (addr_type
& IPV6_ADDR_MULTICAST
)
1450 rt
->dst
.input
= ip6_mc_input
;
1451 else if (cfg
->fc_flags
& RTF_LOCAL
)
1452 rt
->dst
.input
= ip6_input
;
1454 rt
->dst
.input
= ip6_forward
;
1456 rt
->dst
.output
= ip6_output
;
1458 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1459 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1460 if (rt
->rt6i_dst
.plen
== 128)
1461 rt
->dst
.flags
|= DST_HOST
;
1463 if (!(rt
->dst
.flags
& DST_HOST
) && cfg
->fc_mx
) {
1464 u32
*metrics
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_KERNEL
);
1469 dst_init_metrics(&rt
->dst
, metrics
, 0);
1471 #ifdef CONFIG_IPV6_SUBTREES
1472 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1473 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1476 rt
->rt6i_metric
= cfg
->fc_metric
;
1478 /* We cannot add true routes via loopback here,
1479 they would result in kernel looping; promote them to reject routes
1481 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1482 (dev
&& (dev
->flags
& IFF_LOOPBACK
) &&
1483 !(addr_type
& IPV6_ADDR_LOOPBACK
) &&
1484 !(cfg
->fc_flags
& RTF_LOCAL
))) {
1485 /* hold loopback dev/idev if we haven't done so. */
1486 if (dev
!= net
->loopback_dev
) {
1491 dev
= net
->loopback_dev
;
1493 idev
= in6_dev_get(dev
);
1499 rt
->dst
.output
= ip6_pkt_discard_out
;
1500 rt
->dst
.input
= ip6_pkt_discard
;
1501 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1502 switch (cfg
->fc_type
) {
1504 rt
->dst
.error
= -EINVAL
;
1507 rt
->dst
.error
= -EACCES
;
1510 rt
->dst
.error
= -EAGAIN
;
1513 rt
->dst
.error
= -ENETUNREACH
;
1519 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1520 const struct in6_addr
*gw_addr
;
1523 gw_addr
= &cfg
->fc_gateway
;
1524 rt
->rt6i_gateway
= *gw_addr
;
1525 gwa_type
= ipv6_addr_type(gw_addr
);
1527 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1528 struct rt6_info
*grt
;
1530 /* IPv6 strictly inhibits using not link-local
1531 addresses as nexthop address.
1532 Otherwise, router will not able to send redirects.
1533 It is very good, but in some (rare!) circumstances
1534 (SIT, PtP, NBMA NOARP links) it is handy to allow
1535 some exceptions. --ANK
1538 if (!(gwa_type
& IPV6_ADDR_UNICAST
))
1541 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1543 err
= -EHOSTUNREACH
;
1547 if (dev
!= grt
->dst
.dev
) {
1553 idev
= grt
->rt6i_idev
;
1555 in6_dev_hold(grt
->rt6i_idev
);
1557 if (!(grt
->rt6i_flags
& RTF_GATEWAY
))
1565 if (!dev
|| (dev
->flags
& IFF_LOOPBACK
))
1573 if (!ipv6_addr_any(&cfg
->fc_prefsrc
)) {
1574 if (!ipv6_chk_addr(net
, &cfg
->fc_prefsrc
, dev
, 0)) {
1578 rt
->rt6i_prefsrc
.addr
= cfg
->fc_prefsrc
;
1579 rt
->rt6i_prefsrc
.plen
= 128;
1581 rt
->rt6i_prefsrc
.plen
= 0;
1583 if (cfg
->fc_flags
& (RTF_GATEWAY
| RTF_NONEXTHOP
)) {
1584 err
= rt6_bind_neighbour(rt
, dev
);
1589 rt
->rt6i_flags
= cfg
->fc_flags
;
1596 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1597 int type
= nla_type(nla
);
1600 if (type
> RTAX_MAX
) {
1605 dst_metric_set(&rt
->dst
, type
, nla_get_u32(nla
));
1611 rt
->rt6i_idev
= idev
;
1612 rt
->rt6i_table
= table
;
1614 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1616 return __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
);
1628 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1631 struct fib6_table
*table
;
1632 struct net
*net
= dev_net(rt
->dst
.dev
);
1634 if (rt
== net
->ipv6
.ip6_null_entry
) {
1639 table
= rt
->rt6i_table
;
1640 write_lock_bh(&table
->tb6_lock
);
1641 err
= fib6_del(rt
, info
);
1642 write_unlock_bh(&table
->tb6_lock
);
1649 int ip6_del_rt(struct rt6_info
*rt
)
1651 struct nl_info info
= {
1652 .nl_net
= dev_net(rt
->dst
.dev
),
1654 return __ip6_del_rt(rt
, &info
);
1657 static int ip6_route_del(struct fib6_config
*cfg
)
1659 struct fib6_table
*table
;
1660 struct fib6_node
*fn
;
1661 struct rt6_info
*rt
;
1664 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1668 read_lock_bh(&table
->tb6_lock
);
1670 fn
= fib6_locate(&table
->tb6_root
,
1671 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1672 &cfg
->fc_src
, cfg
->fc_src_len
);
1675 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1676 if (cfg
->fc_ifindex
&&
1678 rt
->dst
.dev
->ifindex
!= cfg
->fc_ifindex
))
1680 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1681 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1683 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1686 read_unlock_bh(&table
->tb6_lock
);
1688 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1691 read_unlock_bh(&table
->tb6_lock
);
1696 static void rt6_do_redirect(struct dst_entry
*dst
, struct sock
*sk
, struct sk_buff
*skb
)
1698 struct net
*net
= dev_net(skb
->dev
);
1699 struct netevent_redirect netevent
;
1700 struct rt6_info
*rt
, *nrt
= NULL
;
1701 struct ndisc_options ndopts
;
1702 struct neighbour
*old_neigh
;
1703 struct inet6_dev
*in6_dev
;
1704 struct neighbour
*neigh
;
1706 int optlen
, on_link
;
1709 optlen
= skb
->tail
- skb
->transport_header
;
1710 optlen
-= sizeof(*msg
);
1713 net_dbg_ratelimited("rt6_do_redirect: packet too short\n");
1717 msg
= (struct rd_msg
*)icmp6_hdr(skb
);
1719 if (ipv6_addr_is_multicast(&msg
->dest
)) {
1720 net_dbg_ratelimited("rt6_do_redirect: destination address is multicast\n");
1725 if (ipv6_addr_equal(&msg
->dest
, &msg
->target
)) {
1727 } else if (ipv6_addr_type(&msg
->target
) !=
1728 (IPV6_ADDR_UNICAST
|IPV6_ADDR_LINKLOCAL
)) {
1729 net_dbg_ratelimited("rt6_do_redirect: target address is not link-local unicast\n");
1733 in6_dev
= __in6_dev_get(skb
->dev
);
1736 if (in6_dev
->cnf
.forwarding
|| !in6_dev
->cnf
.accept_redirects
)
1740 * The IP source address of the Redirect MUST be the same as the current
1741 * first-hop router for the specified ICMP Destination Address.
1744 if (!ndisc_parse_options(msg
->opt
, optlen
, &ndopts
)) {
1745 net_dbg_ratelimited("rt6_redirect: invalid ND options\n");
1750 if (ndopts
.nd_opts_tgt_lladdr
) {
1751 lladdr
= ndisc_opt_addr_data(ndopts
.nd_opts_tgt_lladdr
,
1754 net_dbg_ratelimited("rt6_redirect: invalid link-layer address length\n");
1759 rt
= (struct rt6_info
*) dst
;
1760 if (rt
== net
->ipv6
.ip6_null_entry
) {
1761 net_dbg_ratelimited("rt6_redirect: source isn't a valid nexthop for redirect target\n");
1765 /* Redirect received -> path was valid.
1766 * Look, redirects are sent only in response to data packets,
1767 * so that this nexthop apparently is reachable. --ANK
1769 dst_confirm(&rt
->dst
);
1771 neigh
= __neigh_lookup(&nd_tbl
, &msg
->target
, skb
->dev
, 1);
1775 /* Duplicate redirect: silently ignore. */
1777 if (neigh
== old_neigh
)
1781 * We have finally decided to accept it.
1784 neigh_update(neigh
, lladdr
, NUD_STALE
,
1785 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
1786 NEIGH_UPDATE_F_OVERRIDE
|
1787 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
1788 NEIGH_UPDATE_F_ISROUTER
))
1791 nrt
= ip6_rt_copy(rt
, &msg
->dest
);
1795 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
1797 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
1799 nrt
->rt6i_gateway
= *(struct in6_addr
*)neigh
->primary_key
;
1800 nrt
->n
= neigh_clone(neigh
);
1802 if (ip6_ins_rt(nrt
))
1805 netevent
.old
= &rt
->dst
;
1806 netevent
.new = &nrt
->dst
;
1807 netevent
.daddr
= &msg
->dest
;
1808 netevent
.neigh
= neigh
;
1809 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
1811 if (rt
->rt6i_flags
& RTF_CACHE
) {
1812 rt
= (struct rt6_info
*) dst_clone(&rt
->dst
);
1817 neigh_release(neigh
);
1821 * Misc support functions
1824 static struct rt6_info
*ip6_rt_copy(struct rt6_info
*ort
,
1825 const struct in6_addr
*dest
)
1827 struct net
*net
= dev_net(ort
->dst
.dev
);
1828 struct rt6_info
*rt
= ip6_dst_alloc(net
, ort
->dst
.dev
, 0,
1832 rt
->dst
.input
= ort
->dst
.input
;
1833 rt
->dst
.output
= ort
->dst
.output
;
1834 rt
->dst
.flags
|= DST_HOST
;
1836 rt
->rt6i_dst
.addr
= *dest
;
1837 rt
->rt6i_dst
.plen
= 128;
1838 dst_copy_metrics(&rt
->dst
, &ort
->dst
);
1839 rt
->dst
.error
= ort
->dst
.error
;
1840 rt
->rt6i_idev
= ort
->rt6i_idev
;
1842 in6_dev_hold(rt
->rt6i_idev
);
1843 rt
->dst
.lastuse
= jiffies
;
1845 rt
->rt6i_gateway
= ort
->rt6i_gateway
;
1846 rt
->rt6i_flags
= ort
->rt6i_flags
;
1847 if ((ort
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) ==
1848 (RTF_DEFAULT
| RTF_ADDRCONF
))
1849 rt6_set_from(rt
, ort
);
1851 rt6_clean_expires(rt
);
1852 rt
->rt6i_metric
= 0;
1854 #ifdef CONFIG_IPV6_SUBTREES
1855 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1857 memcpy(&rt
->rt6i_prefsrc
, &ort
->rt6i_prefsrc
, sizeof(struct rt6key
));
1858 rt
->rt6i_table
= ort
->rt6i_table
;
1863 #ifdef CONFIG_IPV6_ROUTE_INFO
1864 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
1865 const struct in6_addr
*prefix
, int prefixlen
,
1866 const struct in6_addr
*gwaddr
, int ifindex
)
1868 struct fib6_node
*fn
;
1869 struct rt6_info
*rt
= NULL
;
1870 struct fib6_table
*table
;
1872 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
1876 read_lock_bh(&table
->tb6_lock
);
1877 fn
= fib6_locate(&table
->tb6_root
, prefix
,prefixlen
, NULL
, 0);
1881 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1882 if (rt
->dst
.dev
->ifindex
!= ifindex
)
1884 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
1886 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
1892 read_unlock_bh(&table
->tb6_lock
);
1896 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
1897 const struct in6_addr
*prefix
, int prefixlen
,
1898 const struct in6_addr
*gwaddr
, int ifindex
,
1901 struct fib6_config cfg
= {
1902 .fc_table
= RT6_TABLE_INFO
,
1903 .fc_metric
= IP6_RT_PRIO_USER
,
1904 .fc_ifindex
= ifindex
,
1905 .fc_dst_len
= prefixlen
,
1906 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
1907 RTF_UP
| RTF_PREF(pref
),
1908 .fc_nlinfo
.portid
= 0,
1909 .fc_nlinfo
.nlh
= NULL
,
1910 .fc_nlinfo
.nl_net
= net
,
1913 cfg
.fc_dst
= *prefix
;
1914 cfg
.fc_gateway
= *gwaddr
;
1916 /* We should treat it as a default route if prefix length is 0. */
1918 cfg
.fc_flags
|= RTF_DEFAULT
;
1920 ip6_route_add(&cfg
);
1922 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
1926 struct rt6_info
*rt6_get_dflt_router(const struct in6_addr
*addr
, struct net_device
*dev
)
1928 struct rt6_info
*rt
;
1929 struct fib6_table
*table
;
1931 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
1935 read_lock_bh(&table
->tb6_lock
);
1936 for (rt
= table
->tb6_root
.leaf
; rt
; rt
=rt
->dst
.rt6_next
) {
1937 if (dev
== rt
->dst
.dev
&&
1938 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
1939 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
1944 read_unlock_bh(&table
->tb6_lock
);
1948 struct rt6_info
*rt6_add_dflt_router(const struct in6_addr
*gwaddr
,
1949 struct net_device
*dev
,
1952 struct fib6_config cfg
= {
1953 .fc_table
= RT6_TABLE_DFLT
,
1954 .fc_metric
= IP6_RT_PRIO_USER
,
1955 .fc_ifindex
= dev
->ifindex
,
1956 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
1957 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
1958 .fc_nlinfo
.portid
= 0,
1959 .fc_nlinfo
.nlh
= NULL
,
1960 .fc_nlinfo
.nl_net
= dev_net(dev
),
1963 cfg
.fc_gateway
= *gwaddr
;
1965 ip6_route_add(&cfg
);
1967 return rt6_get_dflt_router(gwaddr
, dev
);
1970 void rt6_purge_dflt_routers(struct net
*net
)
1972 struct rt6_info
*rt
;
1973 struct fib6_table
*table
;
1975 /* NOTE: Keep consistent with rt6_get_dflt_router */
1976 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
1981 read_lock_bh(&table
->tb6_lock
);
1982 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1983 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) {
1985 read_unlock_bh(&table
->tb6_lock
);
1990 read_unlock_bh(&table
->tb6_lock
);
1993 static void rtmsg_to_fib6_config(struct net
*net
,
1994 struct in6_rtmsg
*rtmsg
,
1995 struct fib6_config
*cfg
)
1997 memset(cfg
, 0, sizeof(*cfg
));
1999 cfg
->fc_table
= RT6_TABLE_MAIN
;
2000 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
2001 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
2002 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
2003 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
2004 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
2005 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
2007 cfg
->fc_nlinfo
.nl_net
= net
;
2009 cfg
->fc_dst
= rtmsg
->rtmsg_dst
;
2010 cfg
->fc_src
= rtmsg
->rtmsg_src
;
2011 cfg
->fc_gateway
= rtmsg
->rtmsg_gateway
;
2014 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
2016 struct fib6_config cfg
;
2017 struct in6_rtmsg rtmsg
;
2021 case SIOCADDRT
: /* Add a route */
2022 case SIOCDELRT
: /* Delete a route */
2023 if (!ns_capable(net
->user_ns
, CAP_NET_ADMIN
))
2025 err
= copy_from_user(&rtmsg
, arg
,
2026 sizeof(struct in6_rtmsg
));
2030 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
2035 err
= ip6_route_add(&cfg
);
2038 err
= ip6_route_del(&cfg
);
2052 * Drop the packet on the floor
2055 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
2058 struct dst_entry
*dst
= skb_dst(skb
);
2059 switch (ipstats_mib_noroutes
) {
2060 case IPSTATS_MIB_INNOROUTES
:
2061 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
2062 if (type
== IPV6_ADDR_ANY
) {
2063 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2064 IPSTATS_MIB_INADDRERRORS
);
2068 case IPSTATS_MIB_OUTNOROUTES
:
2069 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
2070 ipstats_mib_noroutes
);
2073 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
2078 static int ip6_pkt_discard(struct sk_buff
*skb
)
2080 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
2083 static int ip6_pkt_discard_out(struct sk_buff
*skb
)
2085 skb
->dev
= skb_dst(skb
)->dev
;
2086 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
2089 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2091 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
2093 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
2096 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
)
2098 skb
->dev
= skb_dst(skb
)->dev
;
2099 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
2105 * Allocate a dst for local (unicast / anycast) address.
2108 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
2109 const struct in6_addr
*addr
,
2112 struct net
*net
= dev_net(idev
->dev
);
2113 struct rt6_info
*rt
= ip6_dst_alloc(net
, net
->loopback_dev
, 0, NULL
);
2117 net_warn_ratelimited("Maximum number of routes reached, consider increasing route/max_size\n");
2118 return ERR_PTR(-ENOMEM
);
2123 rt
->dst
.flags
|= DST_HOST
;
2124 rt
->dst
.input
= ip6_input
;
2125 rt
->dst
.output
= ip6_output
;
2126 rt
->rt6i_idev
= idev
;
2128 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2130 rt
->rt6i_flags
|= RTF_ANYCAST
;
2132 rt
->rt6i_flags
|= RTF_LOCAL
;
2133 err
= rt6_bind_neighbour(rt
, rt
->dst
.dev
);
2136 return ERR_PTR(err
);
2139 rt
->rt6i_dst
.addr
= *addr
;
2140 rt
->rt6i_dst
.plen
= 128;
2141 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2143 atomic_set(&rt
->dst
.__refcnt
, 1);
2148 int ip6_route_get_saddr(struct net
*net
,
2149 struct rt6_info
*rt
,
2150 const struct in6_addr
*daddr
,
2152 struct in6_addr
*saddr
)
2154 struct inet6_dev
*idev
= ip6_dst_idev((struct dst_entry
*)rt
);
2156 if (rt
->rt6i_prefsrc
.plen
)
2157 *saddr
= rt
->rt6i_prefsrc
.addr
;
2159 err
= ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2160 daddr
, prefs
, saddr
);
2164 /* remove deleted ip from prefsrc entries */
2165 struct arg_dev_net_ip
{
2166 struct net_device
*dev
;
2168 struct in6_addr
*addr
;
2171 static int fib6_remove_prefsrc(struct rt6_info
*rt
, void *arg
)
2173 struct net_device
*dev
= ((struct arg_dev_net_ip
*)arg
)->dev
;
2174 struct net
*net
= ((struct arg_dev_net_ip
*)arg
)->net
;
2175 struct in6_addr
*addr
= ((struct arg_dev_net_ip
*)arg
)->addr
;
2177 if (((void *)rt
->dst
.dev
== dev
|| !dev
) &&
2178 rt
!= net
->ipv6
.ip6_null_entry
&&
2179 ipv6_addr_equal(addr
, &rt
->rt6i_prefsrc
.addr
)) {
2180 /* remove prefsrc entry */
2181 rt
->rt6i_prefsrc
.plen
= 0;
2186 void rt6_remove_prefsrc(struct inet6_ifaddr
*ifp
)
2188 struct net
*net
= dev_net(ifp
->idev
->dev
);
2189 struct arg_dev_net_ip adni
= {
2190 .dev
= ifp
->idev
->dev
,
2194 fib6_clean_all(net
, fib6_remove_prefsrc
, 0, &adni
);
2197 struct arg_dev_net
{
2198 struct net_device
*dev
;
2202 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2204 const struct arg_dev_net
*adn
= arg
;
2205 const struct net_device
*dev
= adn
->dev
;
2207 if ((rt
->dst
.dev
== dev
|| !dev
) &&
2208 rt
!= adn
->net
->ipv6
.ip6_null_entry
)
2214 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2216 struct arg_dev_net adn
= {
2221 fib6_clean_all(net
, fib6_ifdown
, 0, &adn
);
2222 icmp6_clean_all(fib6_ifdown
, &adn
);
2225 struct rt6_mtu_change_arg
{
2226 struct net_device
*dev
;
2230 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2232 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2233 struct inet6_dev
*idev
;
2235 /* In IPv6 pmtu discovery is not optional,
2236 so that RTAX_MTU lock cannot disable it.
2237 We still use this lock to block changes
2238 caused by addrconf/ndisc.
2241 idev
= __in6_dev_get(arg
->dev
);
2245 /* For administrative MTU increase, there is no way to discover
2246 IPv6 PMTU increase, so PMTU increase should be updated here.
2247 Since RFC 1981 doesn't include administrative MTU increase
2248 update PMTU increase is a MUST. (i.e. jumbo frame)
2251 If new MTU is less than route PMTU, this new MTU will be the
2252 lowest MTU in the path, update the route PMTU to reflect PMTU
2253 decreases; if new MTU is greater than route PMTU, and the
2254 old MTU is the lowest MTU in the path, update the route PMTU
2255 to reflect the increase. In this case if the other nodes' MTU
2256 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2259 if (rt
->dst
.dev
== arg
->dev
&&
2260 !dst_metric_locked(&rt
->dst
, RTAX_MTU
) &&
2261 (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2262 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2263 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
))) {
2264 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2269 void rt6_mtu_change(struct net_device
*dev
, unsigned int mtu
)
2271 struct rt6_mtu_change_arg arg
= {
2276 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, 0, &arg
);
2279 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2280 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2281 [RTA_OIF
] = { .type
= NLA_U32
},
2282 [RTA_IIF
] = { .type
= NLA_U32
},
2283 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2284 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2285 [RTA_MULTIPATH
] = { .len
= sizeof(struct rtnexthop
) },
2288 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2289 struct fib6_config
*cfg
)
2292 struct nlattr
*tb
[RTA_MAX
+1];
2295 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2300 rtm
= nlmsg_data(nlh
);
2301 memset(cfg
, 0, sizeof(*cfg
));
2303 cfg
->fc_table
= rtm
->rtm_table
;
2304 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2305 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2306 cfg
->fc_flags
= RTF_UP
;
2307 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2308 cfg
->fc_type
= rtm
->rtm_type
;
2310 if (rtm
->rtm_type
== RTN_UNREACHABLE
||
2311 rtm
->rtm_type
== RTN_BLACKHOLE
||
2312 rtm
->rtm_type
== RTN_PROHIBIT
||
2313 rtm
->rtm_type
== RTN_THROW
)
2314 cfg
->fc_flags
|= RTF_REJECT
;
2316 if (rtm
->rtm_type
== RTN_LOCAL
)
2317 cfg
->fc_flags
|= RTF_LOCAL
;
2319 cfg
->fc_nlinfo
.portid
= NETLINK_CB(skb
).portid
;
2320 cfg
->fc_nlinfo
.nlh
= nlh
;
2321 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2323 if (tb
[RTA_GATEWAY
]) {
2324 nla_memcpy(&cfg
->fc_gateway
, tb
[RTA_GATEWAY
], 16);
2325 cfg
->fc_flags
|= RTF_GATEWAY
;
2329 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2331 if (nla_len(tb
[RTA_DST
]) < plen
)
2334 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2338 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2340 if (nla_len(tb
[RTA_SRC
]) < plen
)
2343 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2346 if (tb
[RTA_PREFSRC
])
2347 nla_memcpy(&cfg
->fc_prefsrc
, tb
[RTA_PREFSRC
], 16);
2350 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2352 if (tb
[RTA_PRIORITY
])
2353 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2355 if (tb
[RTA_METRICS
]) {
2356 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2357 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2361 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2363 if (tb
[RTA_MULTIPATH
]) {
2364 cfg
->fc_mp
= nla_data(tb
[RTA_MULTIPATH
]);
2365 cfg
->fc_mp_len
= nla_len(tb
[RTA_MULTIPATH
]);
2373 static int ip6_route_multipath(struct fib6_config
*cfg
, int add
)
2375 struct fib6_config r_cfg
;
2376 struct rtnexthop
*rtnh
;
2379 int err
= 0, last_err
= 0;
2382 rtnh
= (struct rtnexthop
*)cfg
->fc_mp
;
2383 remaining
= cfg
->fc_mp_len
;
2385 /* Parse a Multipath Entry */
2386 while (rtnh_ok(rtnh
, remaining
)) {
2387 memcpy(&r_cfg
, cfg
, sizeof(*cfg
));
2388 if (rtnh
->rtnh_ifindex
)
2389 r_cfg
.fc_ifindex
= rtnh
->rtnh_ifindex
;
2391 attrlen
= rtnh_attrlen(rtnh
);
2393 struct nlattr
*nla
, *attrs
= rtnh_attrs(rtnh
);
2395 nla
= nla_find(attrs
, attrlen
, RTA_GATEWAY
);
2397 nla_memcpy(&r_cfg
.fc_gateway
, nla
, 16);
2398 r_cfg
.fc_flags
|= RTF_GATEWAY
;
2401 err
= add
? ip6_route_add(&r_cfg
) : ip6_route_del(&r_cfg
);
2404 /* If we are trying to remove a route, do not stop the
2405 * loop when ip6_route_del() fails (because next hop is
2406 * already gone), we should try to remove all next hops.
2409 /* If add fails, we should try to delete all
2410 * next hops that have been already added.
2416 /* Because each route is added like a single route we remove
2417 * this flag after the first nexthop (if there is a collision,
2418 * we have already fail to add the first nexthop:
2419 * fib6_add_rt2node() has reject it).
2421 cfg
->fc_nlinfo
.nlh
->nlmsg_flags
&= ~NLM_F_EXCL
;
2422 rtnh
= rtnh_next(rtnh
, &remaining
);
2428 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
2430 struct fib6_config cfg
;
2433 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2438 return ip6_route_multipath(&cfg
, 0);
2440 return ip6_route_del(&cfg
);
2443 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
2445 struct fib6_config cfg
;
2448 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2453 return ip6_route_multipath(&cfg
, 1);
2455 return ip6_route_add(&cfg
);
2458 static inline size_t rt6_nlmsg_size(void)
2460 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2461 + nla_total_size(16) /* RTA_SRC */
2462 + nla_total_size(16) /* RTA_DST */
2463 + nla_total_size(16) /* RTA_GATEWAY */
2464 + nla_total_size(16) /* RTA_PREFSRC */
2465 + nla_total_size(4) /* RTA_TABLE */
2466 + nla_total_size(4) /* RTA_IIF */
2467 + nla_total_size(4) /* RTA_OIF */
2468 + nla_total_size(4) /* RTA_PRIORITY */
2469 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2470 + nla_total_size(sizeof(struct rta_cacheinfo
));
2473 static int rt6_fill_node(struct net
*net
,
2474 struct sk_buff
*skb
, struct rt6_info
*rt
,
2475 struct in6_addr
*dst
, struct in6_addr
*src
,
2476 int iif
, int type
, u32 portid
, u32 seq
,
2477 int prefix
, int nowait
, unsigned int flags
)
2480 struct nlmsghdr
*nlh
;
2483 struct neighbour
*n
;
2485 if (prefix
) { /* user wants prefix routes only */
2486 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2487 /* success since this is not a prefix route */
2492 nlh
= nlmsg_put(skb
, portid
, seq
, type
, sizeof(*rtm
), flags
);
2496 rtm
= nlmsg_data(nlh
);
2497 rtm
->rtm_family
= AF_INET6
;
2498 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2499 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2502 table
= rt
->rt6i_table
->tb6_id
;
2504 table
= RT6_TABLE_UNSPEC
;
2505 rtm
->rtm_table
= table
;
2506 if (nla_put_u32(skb
, RTA_TABLE
, table
))
2507 goto nla_put_failure
;
2508 if (rt
->rt6i_flags
& RTF_REJECT
) {
2509 switch (rt
->dst
.error
) {
2511 rtm
->rtm_type
= RTN_BLACKHOLE
;
2514 rtm
->rtm_type
= RTN_PROHIBIT
;
2517 rtm
->rtm_type
= RTN_THROW
;
2520 rtm
->rtm_type
= RTN_UNREACHABLE
;
2524 else if (rt
->rt6i_flags
& RTF_LOCAL
)
2525 rtm
->rtm_type
= RTN_LOCAL
;
2526 else if (rt
->dst
.dev
&& (rt
->dst
.dev
->flags
& IFF_LOOPBACK
))
2527 rtm
->rtm_type
= RTN_LOCAL
;
2529 rtm
->rtm_type
= RTN_UNICAST
;
2531 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2532 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2533 if (rt
->rt6i_flags
& RTF_DYNAMIC
)
2534 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2535 else if (rt
->rt6i_flags
& RTF_ADDRCONF
) {
2536 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ROUTEINFO
))
2537 rtm
->rtm_protocol
= RTPROT_RA
;
2539 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2542 if (rt
->rt6i_flags
& RTF_CACHE
)
2543 rtm
->rtm_flags
|= RTM_F_CLONED
;
2546 if (nla_put(skb
, RTA_DST
, 16, dst
))
2547 goto nla_put_failure
;
2548 rtm
->rtm_dst_len
= 128;
2549 } else if (rtm
->rtm_dst_len
)
2550 if (nla_put(skb
, RTA_DST
, 16, &rt
->rt6i_dst
.addr
))
2551 goto nla_put_failure
;
2552 #ifdef CONFIG_IPV6_SUBTREES
2554 if (nla_put(skb
, RTA_SRC
, 16, src
))
2555 goto nla_put_failure
;
2556 rtm
->rtm_src_len
= 128;
2557 } else if (rtm
->rtm_src_len
&&
2558 nla_put(skb
, RTA_SRC
, 16, &rt
->rt6i_src
.addr
))
2559 goto nla_put_failure
;
2562 #ifdef CONFIG_IPV6_MROUTE
2563 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2564 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2569 goto nla_put_failure
;
2571 if (err
== -EMSGSIZE
)
2572 goto nla_put_failure
;
2577 if (nla_put_u32(skb
, RTA_IIF
, iif
))
2578 goto nla_put_failure
;
2580 struct in6_addr saddr_buf
;
2581 if (ip6_route_get_saddr(net
, rt
, dst
, 0, &saddr_buf
) == 0 &&
2582 nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2583 goto nla_put_failure
;
2586 if (rt
->rt6i_prefsrc
.plen
) {
2587 struct in6_addr saddr_buf
;
2588 saddr_buf
= rt
->rt6i_prefsrc
.addr
;
2589 if (nla_put(skb
, RTA_PREFSRC
, 16, &saddr_buf
))
2590 goto nla_put_failure
;
2593 if (rtnetlink_put_metrics(skb
, dst_metrics_ptr(&rt
->dst
)) < 0)
2594 goto nla_put_failure
;
2598 if (nla_put(skb
, RTA_GATEWAY
, 16, &n
->primary_key
) < 0)
2599 goto nla_put_failure
;
2603 nla_put_u32(skb
, RTA_OIF
, rt
->dst
.dev
->ifindex
))
2604 goto nla_put_failure
;
2605 if (nla_put_u32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
))
2606 goto nla_put_failure
;
2608 expires
= (rt
->rt6i_flags
& RTF_EXPIRES
) ? rt
->dst
.expires
- jiffies
: 0;
2610 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, expires
, rt
->dst
.error
) < 0)
2611 goto nla_put_failure
;
2613 return nlmsg_end(skb
, nlh
);
2616 nlmsg_cancel(skb
, nlh
);
2620 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2622 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2625 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2626 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2627 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2631 return rt6_fill_node(arg
->net
,
2632 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2633 NETLINK_CB(arg
->cb
->skb
).portid
, arg
->cb
->nlh
->nlmsg_seq
,
2634 prefix
, 0, NLM_F_MULTI
);
2637 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
* nlh
, void *arg
)
2639 struct net
*net
= sock_net(in_skb
->sk
);
2640 struct nlattr
*tb
[RTA_MAX
+1];
2641 struct rt6_info
*rt
;
2642 struct sk_buff
*skb
;
2645 int err
, iif
= 0, oif
= 0;
2647 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2652 memset(&fl6
, 0, sizeof(fl6
));
2655 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2658 fl6
.saddr
= *(struct in6_addr
*)nla_data(tb
[RTA_SRC
]);
2662 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
2665 fl6
.daddr
= *(struct in6_addr
*)nla_data(tb
[RTA_DST
]);
2669 iif
= nla_get_u32(tb
[RTA_IIF
]);
2672 oif
= nla_get_u32(tb
[RTA_OIF
]);
2675 struct net_device
*dev
;
2678 dev
= __dev_get_by_index(net
, iif
);
2684 fl6
.flowi6_iif
= iif
;
2686 if (!ipv6_addr_any(&fl6
.saddr
))
2687 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
2689 rt
= (struct rt6_info
*)ip6_route_input_lookup(net
, dev
, &fl6
,
2692 fl6
.flowi6_oif
= oif
;
2694 rt
= (struct rt6_info
*)ip6_route_output(net
, NULL
, &fl6
);
2697 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2704 /* Reserve room for dummy headers, this skb can pass
2705 through good chunk of routing engine.
2707 skb_reset_mac_header(skb
);
2708 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
2710 skb_dst_set(skb
, &rt
->dst
);
2712 err
= rt6_fill_node(net
, skb
, rt
, &fl6
.daddr
, &fl6
.saddr
, iif
,
2713 RTM_NEWROUTE
, NETLINK_CB(in_skb
).portid
,
2714 nlh
->nlmsg_seq
, 0, 0, 0);
2720 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).portid
);
2725 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
2727 struct sk_buff
*skb
;
2728 struct net
*net
= info
->nl_net
;
2733 seq
= info
->nlh
? info
->nlh
->nlmsg_seq
: 0;
2735 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
2739 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
2740 event
, info
->portid
, seq
, 0, 0, 0);
2742 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2743 WARN_ON(err
== -EMSGSIZE
);
2747 rtnl_notify(skb
, net
, info
->portid
, RTNLGRP_IPV6_ROUTE
,
2748 info
->nlh
, gfp_any());
2752 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
2755 static int ip6_route_dev_notify(struct notifier_block
*this,
2756 unsigned long event
, void *data
)
2758 struct net_device
*dev
= (struct net_device
*)data
;
2759 struct net
*net
= dev_net(dev
);
2761 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
2762 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
2763 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
2764 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2765 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
2766 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
2767 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
2768 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
2779 #ifdef CONFIG_PROC_FS
2790 static int rt6_info_route(struct rt6_info
*rt
, void *p_arg
)
2792 struct seq_file
*m
= p_arg
;
2793 struct neighbour
*n
;
2795 seq_printf(m
, "%pi6 %02x ", &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
);
2797 #ifdef CONFIG_IPV6_SUBTREES
2798 seq_printf(m
, "%pi6 %02x ", &rt
->rt6i_src
.addr
, rt
->rt6i_src
.plen
);
2800 seq_puts(m
, "00000000000000000000000000000000 00 ");
2804 seq_printf(m
, "%pi6", n
->primary_key
);
2806 seq_puts(m
, "00000000000000000000000000000000");
2808 seq_printf(m
, " %08x %08x %08x %08x %8s\n",
2809 rt
->rt6i_metric
, atomic_read(&rt
->dst
.__refcnt
),
2810 rt
->dst
.__use
, rt
->rt6i_flags
,
2811 rt
->dst
.dev
? rt
->dst
.dev
->name
: "");
2815 static int ipv6_route_show(struct seq_file
*m
, void *v
)
2817 struct net
*net
= (struct net
*)m
->private;
2818 fib6_clean_all_ro(net
, rt6_info_route
, 0, m
);
2822 static int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2824 return single_open_net(inode
, file
, ipv6_route_show
);
2827 static const struct file_operations ipv6_route_proc_fops
= {
2828 .owner
= THIS_MODULE
,
2829 .open
= ipv6_route_open
,
2831 .llseek
= seq_lseek
,
2832 .release
= single_release_net
,
2835 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
2837 struct net
*net
= (struct net
*)seq
->private;
2838 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
2839 net
->ipv6
.rt6_stats
->fib_nodes
,
2840 net
->ipv6
.rt6_stats
->fib_route_nodes
,
2841 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
2842 net
->ipv6
.rt6_stats
->fib_rt_entries
,
2843 net
->ipv6
.rt6_stats
->fib_rt_cache
,
2844 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
2845 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
2850 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
2852 return single_open_net(inode
, file
, rt6_stats_seq_show
);
2855 static const struct file_operations rt6_stats_seq_fops
= {
2856 .owner
= THIS_MODULE
,
2857 .open
= rt6_stats_seq_open
,
2859 .llseek
= seq_lseek
,
2860 .release
= single_release_net
,
2862 #endif /* CONFIG_PROC_FS */
2864 #ifdef CONFIG_SYSCTL
2867 int ipv6_sysctl_rtcache_flush(ctl_table
*ctl
, int write
,
2868 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2875 net
= (struct net
*)ctl
->extra1
;
2876 delay
= net
->ipv6
.sysctl
.flush_delay
;
2877 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
2878 fib6_run_gc(delay
<= 0 ? ~0UL : (unsigned long)delay
, net
);
2882 ctl_table ipv6_route_table_template
[] = {
2884 .procname
= "flush",
2885 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
2886 .maxlen
= sizeof(int),
2888 .proc_handler
= ipv6_sysctl_rtcache_flush
2891 .procname
= "gc_thresh",
2892 .data
= &ip6_dst_ops_template
.gc_thresh
,
2893 .maxlen
= sizeof(int),
2895 .proc_handler
= proc_dointvec
,
2898 .procname
= "max_size",
2899 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
2900 .maxlen
= sizeof(int),
2902 .proc_handler
= proc_dointvec
,
2905 .procname
= "gc_min_interval",
2906 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2907 .maxlen
= sizeof(int),
2909 .proc_handler
= proc_dointvec_jiffies
,
2912 .procname
= "gc_timeout",
2913 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
2914 .maxlen
= sizeof(int),
2916 .proc_handler
= proc_dointvec_jiffies
,
2919 .procname
= "gc_interval",
2920 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
2921 .maxlen
= sizeof(int),
2923 .proc_handler
= proc_dointvec_jiffies
,
2926 .procname
= "gc_elasticity",
2927 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
2928 .maxlen
= sizeof(int),
2930 .proc_handler
= proc_dointvec
,
2933 .procname
= "mtu_expires",
2934 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
2935 .maxlen
= sizeof(int),
2937 .proc_handler
= proc_dointvec_jiffies
,
2940 .procname
= "min_adv_mss",
2941 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
2942 .maxlen
= sizeof(int),
2944 .proc_handler
= proc_dointvec
,
2947 .procname
= "gc_min_interval_ms",
2948 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2949 .maxlen
= sizeof(int),
2951 .proc_handler
= proc_dointvec_ms_jiffies
,
2956 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
2958 struct ctl_table
*table
;
2960 table
= kmemdup(ipv6_route_table_template
,
2961 sizeof(ipv6_route_table_template
),
2965 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
2966 table
[0].extra1
= net
;
2967 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
2968 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
2969 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2970 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2971 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
2972 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2973 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
2974 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2975 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2977 /* Don't export sysctls to unprivileged users */
2978 if (net
->user_ns
!= &init_user_ns
)
2979 table
[0].procname
= NULL
;
2986 static int __net_init
ip6_route_net_init(struct net
*net
)
2990 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
2991 sizeof(net
->ipv6
.ip6_dst_ops
));
2993 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
2994 goto out_ip6_dst_ops
;
2996 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
2997 sizeof(*net
->ipv6
.ip6_null_entry
),
2999 if (!net
->ipv6
.ip6_null_entry
)
3000 goto out_ip6_dst_entries
;
3001 net
->ipv6
.ip6_null_entry
->dst
.path
=
3002 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
3003 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3004 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
3005 ip6_template_metrics
, true);
3007 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3008 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
3009 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
3011 if (!net
->ipv6
.ip6_prohibit_entry
)
3012 goto out_ip6_null_entry
;
3013 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
3014 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
3015 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3016 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
3017 ip6_template_metrics
, true);
3019 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
3020 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
3022 if (!net
->ipv6
.ip6_blk_hole_entry
)
3023 goto out_ip6_prohibit_entry
;
3024 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
3025 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
3026 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
3027 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
3028 ip6_template_metrics
, true);
3031 net
->ipv6
.sysctl
.flush_delay
= 0;
3032 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
3033 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
3034 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
3035 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
3036 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
3037 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
3038 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
3040 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
3046 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3047 out_ip6_prohibit_entry
:
3048 kfree(net
->ipv6
.ip6_prohibit_entry
);
3050 kfree(net
->ipv6
.ip6_null_entry
);
3052 out_ip6_dst_entries
:
3053 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3058 static void __net_exit
ip6_route_net_exit(struct net
*net
)
3060 kfree(net
->ipv6
.ip6_null_entry
);
3061 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3062 kfree(net
->ipv6
.ip6_prohibit_entry
);
3063 kfree(net
->ipv6
.ip6_blk_hole_entry
);
3065 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
3068 static int __net_init
ip6_route_net_init_late(struct net
*net
)
3070 #ifdef CONFIG_PROC_FS
3071 proc_net_fops_create(net
, "ipv6_route", 0, &ipv6_route_proc_fops
);
3072 proc_net_fops_create(net
, "rt6_stats", S_IRUGO
, &rt6_stats_seq_fops
);
3077 static void __net_exit
ip6_route_net_exit_late(struct net
*net
)
3079 #ifdef CONFIG_PROC_FS
3080 proc_net_remove(net
, "ipv6_route");
3081 proc_net_remove(net
, "rt6_stats");
3085 static struct pernet_operations ip6_route_net_ops
= {
3086 .init
= ip6_route_net_init
,
3087 .exit
= ip6_route_net_exit
,
3090 static int __net_init
ipv6_inetpeer_init(struct net
*net
)
3092 struct inet_peer_base
*bp
= kmalloc(sizeof(*bp
), GFP_KERNEL
);
3096 inet_peer_base_init(bp
);
3097 net
->ipv6
.peers
= bp
;
3101 static void __net_exit
ipv6_inetpeer_exit(struct net
*net
)
3103 struct inet_peer_base
*bp
= net
->ipv6
.peers
;
3105 net
->ipv6
.peers
= NULL
;
3106 inetpeer_invalidate_tree(bp
);
3110 static struct pernet_operations ipv6_inetpeer_ops
= {
3111 .init
= ipv6_inetpeer_init
,
3112 .exit
= ipv6_inetpeer_exit
,
3115 static struct pernet_operations ip6_route_net_late_ops
= {
3116 .init
= ip6_route_net_init_late
,
3117 .exit
= ip6_route_net_exit_late
,
3120 static struct notifier_block ip6_route_dev_notifier
= {
3121 .notifier_call
= ip6_route_dev_notify
,
3125 int __init
ip6_route_init(void)
3130 ip6_dst_ops_template
.kmem_cachep
=
3131 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
3132 SLAB_HWCACHE_ALIGN
, NULL
);
3133 if (!ip6_dst_ops_template
.kmem_cachep
)
3136 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
3138 goto out_kmem_cache
;
3140 ret
= register_pernet_subsys(&ipv6_inetpeer_ops
);
3142 goto out_dst_entries
;
3144 ret
= register_pernet_subsys(&ip6_route_net_ops
);
3146 goto out_register_inetpeer
;
3148 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
3150 /* Registering of the loopback is done before this portion of code,
3151 * the loopback reference in rt6_info will not be taken, do it
3152 * manually for init_net */
3153 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
3154 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3155 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
3156 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
3157 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3158 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
3159 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
3163 goto out_register_subsys
;
3169 ret
= fib6_rules_init();
3173 ret
= register_pernet_subsys(&ip6_route_net_late_ops
);
3175 goto fib6_rules_init
;
3178 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
, NULL
) ||
3179 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
, NULL
) ||
3180 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
, NULL
))
3181 goto out_register_late_subsys
;
3183 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
3185 goto out_register_late_subsys
;
3190 out_register_late_subsys
:
3191 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3193 fib6_rules_cleanup();
3198 out_register_subsys
:
3199 unregister_pernet_subsys(&ip6_route_net_ops
);
3200 out_register_inetpeer
:
3201 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3203 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3205 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
3209 void ip6_route_cleanup(void)
3211 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
3212 unregister_pernet_subsys(&ip6_route_net_late_ops
);
3213 fib6_rules_cleanup();
3216 unregister_pernet_subsys(&ipv6_inetpeer_ops
);
3217 unregister_pernet_subsys(&ip6_route_net_ops
);
3218 dst_entries_destroy(&ip6_dst_blackhole_ops
);
3219 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);