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 #include <linux/capability.h>
28 #include <linux/errno.h>
29 #include <linux/types.h>
30 #include <linux/times.h>
31 #include <linux/socket.h>
32 #include <linux/sockios.h>
33 #include <linux/net.h>
34 #include <linux/route.h>
35 #include <linux/netdevice.h>
36 #include <linux/in6.h>
37 #include <linux/mroute6.h>
38 #include <linux/init.h>
39 #include <linux/if_arp.h>
40 #include <linux/proc_fs.h>
41 #include <linux/seq_file.h>
42 #include <linux/nsproxy.h>
43 #include <linux/slab.h>
44 #include <net/net_namespace.h>
47 #include <net/ip6_fib.h>
48 #include <net/ip6_route.h>
49 #include <net/ndisc.h>
50 #include <net/addrconf.h>
52 #include <linux/rtnetlink.h>
55 #include <net/netevent.h>
56 #include <net/netlink.h>
58 #include <asm/uaccess.h>
61 #include <linux/sysctl.h>
64 /* Set to 3 to get tracing. */
68 #define RDBG(x) printk x
69 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
72 #define RT6_TRACE(x...) do { ; } while (0)
75 static struct rt6_info
* ip6_rt_copy(struct rt6_info
*ort
);
76 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
);
77 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
);
78 static unsigned int ip6_default_mtu(const struct dst_entry
*dst
);
79 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*);
80 static void ip6_dst_destroy(struct dst_entry
*);
81 static void ip6_dst_ifdown(struct dst_entry
*,
82 struct net_device
*dev
, int how
);
83 static int ip6_dst_gc(struct dst_ops
*ops
);
85 static int ip6_pkt_discard(struct sk_buff
*skb
);
86 static int ip6_pkt_discard_out(struct sk_buff
*skb
);
87 static void ip6_link_failure(struct sk_buff
*skb
);
88 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, u32 mtu
);
90 #ifdef CONFIG_IPV6_ROUTE_INFO
91 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
92 struct in6_addr
*prefix
, int prefixlen
,
93 struct in6_addr
*gwaddr
, int ifindex
,
95 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
96 struct in6_addr
*prefix
, int prefixlen
,
97 struct in6_addr
*gwaddr
, int ifindex
);
100 static u32
*ipv6_cow_metrics(struct dst_entry
*dst
, unsigned long old
)
102 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
103 struct inet_peer
*peer
;
107 rt6_bind_peer(rt
, 1);
109 peer
= rt
->rt6i_peer
;
111 u32
*old_p
= __DST_METRICS_PTR(old
);
112 unsigned long prev
, new;
115 if (inet_metrics_new(peer
))
116 memcpy(p
, old_p
, sizeof(u32
) * RTAX_MAX
);
118 new = (unsigned long) p
;
119 prev
= cmpxchg(&dst
->_metrics
, old
, new);
122 p
= __DST_METRICS_PTR(prev
);
123 if (prev
& DST_METRICS_READ_ONLY
)
130 static struct dst_ops ip6_dst_ops_template
= {
132 .protocol
= cpu_to_be16(ETH_P_IPV6
),
135 .check
= ip6_dst_check
,
136 .default_advmss
= ip6_default_advmss
,
137 .default_mtu
= ip6_default_mtu
,
138 .cow_metrics
= ipv6_cow_metrics
,
139 .destroy
= ip6_dst_destroy
,
140 .ifdown
= ip6_dst_ifdown
,
141 .negative_advice
= ip6_negative_advice
,
142 .link_failure
= ip6_link_failure
,
143 .update_pmtu
= ip6_rt_update_pmtu
,
144 .local_out
= __ip6_local_out
,
147 static unsigned int ip6_blackhole_default_mtu(const struct dst_entry
*dst
)
152 static void ip6_rt_blackhole_update_pmtu(struct dst_entry
*dst
, u32 mtu
)
156 static struct dst_ops ip6_dst_blackhole_ops
= {
158 .protocol
= cpu_to_be16(ETH_P_IPV6
),
159 .destroy
= ip6_dst_destroy
,
160 .check
= ip6_dst_check
,
161 .default_mtu
= ip6_blackhole_default_mtu
,
162 .default_advmss
= ip6_default_advmss
,
163 .update_pmtu
= ip6_rt_blackhole_update_pmtu
,
166 static const u32 ip6_template_metrics
[RTAX_MAX
] = {
167 [RTAX_HOPLIMIT
- 1] = 255,
170 static struct rt6_info ip6_null_entry_template
= {
172 .__refcnt
= ATOMIC_INIT(1),
175 .error
= -ENETUNREACH
,
176 .input
= ip6_pkt_discard
,
177 .output
= ip6_pkt_discard_out
,
179 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
180 .rt6i_protocol
= RTPROT_KERNEL
,
181 .rt6i_metric
= ~(u32
) 0,
182 .rt6i_ref
= ATOMIC_INIT(1),
185 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
187 static int ip6_pkt_prohibit(struct sk_buff
*skb
);
188 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
);
190 static struct rt6_info ip6_prohibit_entry_template
= {
192 .__refcnt
= ATOMIC_INIT(1),
196 .input
= ip6_pkt_prohibit
,
197 .output
= ip6_pkt_prohibit_out
,
199 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
200 .rt6i_protocol
= RTPROT_KERNEL
,
201 .rt6i_metric
= ~(u32
) 0,
202 .rt6i_ref
= ATOMIC_INIT(1),
205 static struct rt6_info ip6_blk_hole_entry_template
= {
207 .__refcnt
= ATOMIC_INIT(1),
211 .input
= dst_discard
,
212 .output
= dst_discard
,
214 .rt6i_flags
= (RTF_REJECT
| RTF_NONEXTHOP
),
215 .rt6i_protocol
= RTPROT_KERNEL
,
216 .rt6i_metric
= ~(u32
) 0,
217 .rt6i_ref
= ATOMIC_INIT(1),
222 /* allocate dst with ip6_dst_ops */
223 static inline struct rt6_info
*ip6_dst_alloc(struct dst_ops
*ops
)
225 return (struct rt6_info
*)dst_alloc(ops
, 0);
228 static void ip6_dst_destroy(struct dst_entry
*dst
)
230 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
231 struct inet6_dev
*idev
= rt
->rt6i_idev
;
232 struct inet_peer
*peer
= rt
->rt6i_peer
;
235 rt
->rt6i_idev
= NULL
;
239 rt
->rt6i_peer
= NULL
;
244 static atomic_t __rt6_peer_genid
= ATOMIC_INIT(0);
246 static u32
rt6_peer_genid(void)
248 return atomic_read(&__rt6_peer_genid
);
251 void rt6_bind_peer(struct rt6_info
*rt
, int create
)
253 struct inet_peer
*peer
;
255 peer
= inet_getpeer_v6(&rt
->rt6i_dst
.addr
, create
);
256 if (peer
&& cmpxchg(&rt
->rt6i_peer
, NULL
, peer
) != NULL
)
259 rt
->rt6i_peer_genid
= rt6_peer_genid();
262 static void ip6_dst_ifdown(struct dst_entry
*dst
, struct net_device
*dev
,
265 struct rt6_info
*rt
= (struct rt6_info
*)dst
;
266 struct inet6_dev
*idev
= rt
->rt6i_idev
;
267 struct net_device
*loopback_dev
=
268 dev_net(dev
)->loopback_dev
;
270 if (dev
!= loopback_dev
&& idev
!= NULL
&& idev
->dev
== dev
) {
271 struct inet6_dev
*loopback_idev
=
272 in6_dev_get(loopback_dev
);
273 if (loopback_idev
!= NULL
) {
274 rt
->rt6i_idev
= loopback_idev
;
280 static __inline__
int rt6_check_expired(const struct rt6_info
*rt
)
282 return (rt
->rt6i_flags
& RTF_EXPIRES
) &&
283 time_after(jiffies
, rt
->rt6i_expires
);
286 static inline int rt6_need_strict(struct in6_addr
*daddr
)
288 return ipv6_addr_type(daddr
) &
289 (IPV6_ADDR_MULTICAST
| IPV6_ADDR_LINKLOCAL
| IPV6_ADDR_LOOPBACK
);
293 * Route lookup. Any table->tb6_lock is implied.
296 static inline struct rt6_info
*rt6_device_match(struct net
*net
,
298 struct in6_addr
*saddr
,
302 struct rt6_info
*local
= NULL
;
303 struct rt6_info
*sprt
;
305 if (!oif
&& ipv6_addr_any(saddr
))
308 for (sprt
= rt
; sprt
; sprt
= sprt
->dst
.rt6_next
) {
309 struct net_device
*dev
= sprt
->rt6i_dev
;
312 if (dev
->ifindex
== oif
)
314 if (dev
->flags
& IFF_LOOPBACK
) {
315 if (sprt
->rt6i_idev
== NULL
||
316 sprt
->rt6i_idev
->dev
->ifindex
!= oif
) {
317 if (flags
& RT6_LOOKUP_F_IFACE
&& oif
)
319 if (local
&& (!oif
||
320 local
->rt6i_idev
->dev
->ifindex
== oif
))
326 if (ipv6_chk_addr(net
, saddr
, dev
,
327 flags
& RT6_LOOKUP_F_IFACE
))
336 if (flags
& RT6_LOOKUP_F_IFACE
)
337 return net
->ipv6
.ip6_null_entry
;
343 #ifdef CONFIG_IPV6_ROUTER_PREF
344 static void rt6_probe(struct rt6_info
*rt
)
346 struct neighbour
*neigh
= rt
? rt
->rt6i_nexthop
: NULL
;
348 * Okay, this does not seem to be appropriate
349 * for now, however, we need to check if it
350 * is really so; aka Router Reachability Probing.
352 * Router Reachability Probe MUST be rate-limited
353 * to no more than one per minute.
355 if (!neigh
|| (neigh
->nud_state
& NUD_VALID
))
357 read_lock_bh(&neigh
->lock
);
358 if (!(neigh
->nud_state
& NUD_VALID
) &&
359 time_after(jiffies
, neigh
->updated
+ rt
->rt6i_idev
->cnf
.rtr_probe_interval
)) {
360 struct in6_addr mcaddr
;
361 struct in6_addr
*target
;
363 neigh
->updated
= jiffies
;
364 read_unlock_bh(&neigh
->lock
);
366 target
= (struct in6_addr
*)&neigh
->primary_key
;
367 addrconf_addr_solict_mult(target
, &mcaddr
);
368 ndisc_send_ns(rt
->rt6i_dev
, NULL
, target
, &mcaddr
, NULL
);
370 read_unlock_bh(&neigh
->lock
);
373 static inline void rt6_probe(struct rt6_info
*rt
)
379 * Default Router Selection (RFC 2461 6.3.6)
381 static inline int rt6_check_dev(struct rt6_info
*rt
, int oif
)
383 struct net_device
*dev
= rt
->rt6i_dev
;
384 if (!oif
|| dev
->ifindex
== oif
)
386 if ((dev
->flags
& IFF_LOOPBACK
) &&
387 rt
->rt6i_idev
&& rt
->rt6i_idev
->dev
->ifindex
== oif
)
392 static inline int rt6_check_neigh(struct rt6_info
*rt
)
394 struct neighbour
*neigh
= rt
->rt6i_nexthop
;
396 if (rt
->rt6i_flags
& RTF_NONEXTHOP
||
397 !(rt
->rt6i_flags
& RTF_GATEWAY
))
400 read_lock_bh(&neigh
->lock
);
401 if (neigh
->nud_state
& NUD_VALID
)
403 #ifdef CONFIG_IPV6_ROUTER_PREF
404 else if (neigh
->nud_state
& NUD_FAILED
)
409 read_unlock_bh(&neigh
->lock
);
415 static int rt6_score_route(struct rt6_info
*rt
, int oif
,
420 m
= rt6_check_dev(rt
, oif
);
421 if (!m
&& (strict
& RT6_LOOKUP_F_IFACE
))
423 #ifdef CONFIG_IPV6_ROUTER_PREF
424 m
|= IPV6_DECODE_PREF(IPV6_EXTRACT_PREF(rt
->rt6i_flags
)) << 2;
426 n
= rt6_check_neigh(rt
);
427 if (!n
&& (strict
& RT6_LOOKUP_F_REACHABLE
))
432 static struct rt6_info
*find_match(struct rt6_info
*rt
, int oif
, int strict
,
433 int *mpri
, struct rt6_info
*match
)
437 if (rt6_check_expired(rt
))
440 m
= rt6_score_route(rt
, oif
, strict
);
445 if (strict
& RT6_LOOKUP_F_REACHABLE
)
449 } else if (strict
& RT6_LOOKUP_F_REACHABLE
) {
457 static struct rt6_info
*find_rr_leaf(struct fib6_node
*fn
,
458 struct rt6_info
*rr_head
,
459 u32 metric
, int oif
, int strict
)
461 struct rt6_info
*rt
, *match
;
465 for (rt
= rr_head
; rt
&& rt
->rt6i_metric
== metric
;
466 rt
= rt
->dst
.rt6_next
)
467 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
468 for (rt
= fn
->leaf
; rt
&& rt
!= rr_head
&& rt
->rt6i_metric
== metric
;
469 rt
= rt
->dst
.rt6_next
)
470 match
= find_match(rt
, oif
, strict
, &mpri
, match
);
475 static struct rt6_info
*rt6_select(struct fib6_node
*fn
, int oif
, int strict
)
477 struct rt6_info
*match
, *rt0
;
480 RT6_TRACE("%s(fn->leaf=%p, oif=%d)\n",
481 __func__
, fn
->leaf
, oif
);
485 fn
->rr_ptr
= rt0
= fn
->leaf
;
487 match
= find_rr_leaf(fn
, rt0
, rt0
->rt6i_metric
, oif
, strict
);
490 (strict
& RT6_LOOKUP_F_REACHABLE
)) {
491 struct rt6_info
*next
= rt0
->dst
.rt6_next
;
493 /* no entries matched; do round-robin */
494 if (!next
|| next
->rt6i_metric
!= rt0
->rt6i_metric
)
501 RT6_TRACE("%s() => %p\n",
504 net
= dev_net(rt0
->rt6i_dev
);
505 return match
? match
: net
->ipv6
.ip6_null_entry
;
508 #ifdef CONFIG_IPV6_ROUTE_INFO
509 int rt6_route_rcv(struct net_device
*dev
, u8
*opt
, int len
,
510 struct in6_addr
*gwaddr
)
512 struct net
*net
= dev_net(dev
);
513 struct route_info
*rinfo
= (struct route_info
*) opt
;
514 struct in6_addr prefix_buf
, *prefix
;
516 unsigned long lifetime
;
519 if (len
< sizeof(struct route_info
)) {
523 /* Sanity check for prefix_len and length */
524 if (rinfo
->length
> 3) {
526 } else if (rinfo
->prefix_len
> 128) {
528 } else if (rinfo
->prefix_len
> 64) {
529 if (rinfo
->length
< 2) {
532 } else if (rinfo
->prefix_len
> 0) {
533 if (rinfo
->length
< 1) {
538 pref
= rinfo
->route_pref
;
539 if (pref
== ICMPV6_ROUTER_PREF_INVALID
)
542 lifetime
= addrconf_timeout_fixup(ntohl(rinfo
->lifetime
), HZ
);
544 if (rinfo
->length
== 3)
545 prefix
= (struct in6_addr
*)rinfo
->prefix
;
547 /* this function is safe */
548 ipv6_addr_prefix(&prefix_buf
,
549 (struct in6_addr
*)rinfo
->prefix
,
551 prefix
= &prefix_buf
;
554 rt
= rt6_get_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
,
557 if (rt
&& !lifetime
) {
563 rt
= rt6_add_route_info(net
, prefix
, rinfo
->prefix_len
, gwaddr
, dev
->ifindex
,
566 rt
->rt6i_flags
= RTF_ROUTEINFO
|
567 (rt
->rt6i_flags
& ~RTF_PREF_MASK
) | RTF_PREF(pref
);
570 if (!addrconf_finite_timeout(lifetime
)) {
571 rt
->rt6i_flags
&= ~RTF_EXPIRES
;
573 rt
->rt6i_expires
= jiffies
+ HZ
* lifetime
;
574 rt
->rt6i_flags
|= RTF_EXPIRES
;
576 dst_release(&rt
->dst
);
582 #define BACKTRACK(__net, saddr) \
584 if (rt == __net->ipv6.ip6_null_entry) { \
585 struct fib6_node *pn; \
587 if (fn->fn_flags & RTN_TL_ROOT) \
590 if (FIB6_SUBTREE(pn) && FIB6_SUBTREE(pn) != fn) \
591 fn = fib6_lookup(FIB6_SUBTREE(pn), NULL, saddr); \
594 if (fn->fn_flags & RTN_RTINFO) \
600 static struct rt6_info
*ip6_pol_route_lookup(struct net
*net
,
601 struct fib6_table
*table
,
602 struct flowi
*fl
, int flags
)
604 struct fib6_node
*fn
;
607 read_lock_bh(&table
->tb6_lock
);
608 fn
= fib6_lookup(&table
->tb6_root
, &fl
->fl6_dst
, &fl
->fl6_src
);
611 rt
= rt6_device_match(net
, rt
, &fl
->fl6_src
, fl
->oif
, flags
);
612 BACKTRACK(net
, &fl
->fl6_src
);
614 dst_use(&rt
->dst
, jiffies
);
615 read_unlock_bh(&table
->tb6_lock
);
620 struct rt6_info
*rt6_lookup(struct net
*net
, const struct in6_addr
*daddr
,
621 const struct in6_addr
*saddr
, int oif
, int strict
)
627 struct dst_entry
*dst
;
628 int flags
= strict
? RT6_LOOKUP_F_IFACE
: 0;
631 memcpy(&fl
.fl6_src
, saddr
, sizeof(*saddr
));
632 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
635 dst
= fib6_rule_lookup(net
, &fl
, flags
, ip6_pol_route_lookup
);
637 return (struct rt6_info
*) dst
;
644 EXPORT_SYMBOL(rt6_lookup
);
646 /* ip6_ins_rt is called with FREE table->tb6_lock.
647 It takes new route entry, the addition fails by any reason the
648 route is freed. In any case, if caller does not hold it, it may
652 static int __ip6_ins_rt(struct rt6_info
*rt
, struct nl_info
*info
)
655 struct fib6_table
*table
;
657 table
= rt
->rt6i_table
;
658 write_lock_bh(&table
->tb6_lock
);
659 err
= fib6_add(&table
->tb6_root
, rt
, info
);
660 write_unlock_bh(&table
->tb6_lock
);
665 int ip6_ins_rt(struct rt6_info
*rt
)
667 struct nl_info info
= {
668 .nl_net
= dev_net(rt
->rt6i_dev
),
670 return __ip6_ins_rt(rt
, &info
);
673 static struct rt6_info
*rt6_alloc_cow(struct rt6_info
*ort
, struct in6_addr
*daddr
,
674 struct in6_addr
*saddr
)
682 rt
= ip6_rt_copy(ort
);
685 struct neighbour
*neigh
;
686 int attempts
= !in_softirq();
688 if (!(rt
->rt6i_flags
&RTF_GATEWAY
)) {
689 if (rt
->rt6i_dst
.plen
!= 128 &&
690 ipv6_addr_equal(&rt
->rt6i_dst
.addr
, daddr
))
691 rt
->rt6i_flags
|= RTF_ANYCAST
;
692 ipv6_addr_copy(&rt
->rt6i_gateway
, daddr
);
695 ipv6_addr_copy(&rt
->rt6i_dst
.addr
, daddr
);
696 rt
->rt6i_dst
.plen
= 128;
697 rt
->rt6i_flags
|= RTF_CACHE
;
698 rt
->dst
.flags
|= DST_HOST
;
700 #ifdef CONFIG_IPV6_SUBTREES
701 if (rt
->rt6i_src
.plen
&& saddr
) {
702 ipv6_addr_copy(&rt
->rt6i_src
.addr
, saddr
);
703 rt
->rt6i_src
.plen
= 128;
708 neigh
= ndisc_get_neigh(rt
->rt6i_dev
, &rt
->rt6i_gateway
);
710 struct net
*net
= dev_net(rt
->rt6i_dev
);
711 int saved_rt_min_interval
=
712 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
713 int saved_rt_elasticity
=
714 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
716 if (attempts
-- > 0) {
717 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 1;
718 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= 0;
720 ip6_dst_gc(&net
->ipv6
.ip6_dst_ops
);
722 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
=
724 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
=
725 saved_rt_min_interval
;
731 "ipv6: Neighbour table overflow.\n");
735 rt
->rt6i_nexthop
= neigh
;
742 static struct rt6_info
*rt6_alloc_clone(struct rt6_info
*ort
, struct in6_addr
*daddr
)
744 struct rt6_info
*rt
= ip6_rt_copy(ort
);
746 ipv6_addr_copy(&rt
->rt6i_dst
.addr
, daddr
);
747 rt
->rt6i_dst
.plen
= 128;
748 rt
->rt6i_flags
|= RTF_CACHE
;
749 rt
->dst
.flags
|= DST_HOST
;
750 rt
->rt6i_nexthop
= neigh_clone(ort
->rt6i_nexthop
);
755 static struct rt6_info
*ip6_pol_route(struct net
*net
, struct fib6_table
*table
, int oif
,
756 struct flowi
*fl
, int flags
)
758 struct fib6_node
*fn
;
759 struct rt6_info
*rt
, *nrt
;
763 int reachable
= net
->ipv6
.devconf_all
->forwarding
? 0 : RT6_LOOKUP_F_REACHABLE
;
765 strict
|= flags
& RT6_LOOKUP_F_IFACE
;
768 read_lock_bh(&table
->tb6_lock
);
771 fn
= fib6_lookup(&table
->tb6_root
, &fl
->fl6_dst
, &fl
->fl6_src
);
774 rt
= rt6_select(fn
, oif
, strict
| reachable
);
776 BACKTRACK(net
, &fl
->fl6_src
);
777 if (rt
== net
->ipv6
.ip6_null_entry
||
778 rt
->rt6i_flags
& RTF_CACHE
)
782 read_unlock_bh(&table
->tb6_lock
);
784 if (!rt
->rt6i_nexthop
&& !(rt
->rt6i_flags
& RTF_NONEXTHOP
))
785 nrt
= rt6_alloc_cow(rt
, &fl
->fl6_dst
, &fl
->fl6_src
);
787 nrt
= rt6_alloc_clone(rt
, &fl
->fl6_dst
);
789 dst_release(&rt
->dst
);
790 rt
= nrt
? : net
->ipv6
.ip6_null_entry
;
794 err
= ip6_ins_rt(nrt
);
803 * Race condition! In the gap, when table->tb6_lock was
804 * released someone could insert this route. Relookup.
806 dst_release(&rt
->dst
);
815 read_unlock_bh(&table
->tb6_lock
);
817 rt
->dst
.lastuse
= jiffies
;
823 static struct rt6_info
*ip6_pol_route_input(struct net
*net
, struct fib6_table
*table
,
824 struct flowi
*fl
, int flags
)
826 return ip6_pol_route(net
, table
, fl
->iif
, fl
, flags
);
829 void ip6_route_input(struct sk_buff
*skb
)
831 struct ipv6hdr
*iph
= ipv6_hdr(skb
);
832 struct net
*net
= dev_net(skb
->dev
);
833 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
835 .iif
= skb
->dev
->ifindex
,
836 .fl6_dst
= iph
->daddr
,
837 .fl6_src
= iph
->saddr
,
838 .fl6_flowlabel
= (* (__be32
*) iph
)&IPV6_FLOWINFO_MASK
,
840 .proto
= iph
->nexthdr
,
843 if (rt6_need_strict(&iph
->daddr
) && skb
->dev
->type
!= ARPHRD_PIMREG
)
844 flags
|= RT6_LOOKUP_F_IFACE
;
846 skb_dst_set(skb
, fib6_rule_lookup(net
, &fl
, flags
, ip6_pol_route_input
));
849 static struct rt6_info
*ip6_pol_route_output(struct net
*net
, struct fib6_table
*table
,
850 struct flowi
*fl
, int flags
)
852 return ip6_pol_route(net
, table
, fl
->oif
, fl
, flags
);
855 struct dst_entry
* ip6_route_output(struct net
*net
, struct sock
*sk
,
860 if ((sk
&& sk
->sk_bound_dev_if
) || rt6_need_strict(&fl
->fl6_dst
))
861 flags
|= RT6_LOOKUP_F_IFACE
;
863 if (!ipv6_addr_any(&fl
->fl6_src
))
864 flags
|= RT6_LOOKUP_F_HAS_SADDR
;
866 flags
|= rt6_srcprefs2flags(inet6_sk(sk
)->srcprefs
);
868 return fib6_rule_lookup(net
, fl
, flags
, ip6_pol_route_output
);
871 EXPORT_SYMBOL(ip6_route_output
);
873 struct dst_entry
*ip6_blackhole_route(struct net
*net
, struct dst_entry
*dst_orig
)
875 struct rt6_info
*rt
= dst_alloc(&ip6_dst_blackhole_ops
, 1);
876 struct rt6_info
*ort
= (struct rt6_info
*) dst_orig
;
877 struct dst_entry
*new = NULL
;
883 new->input
= dst_discard
;
884 new->output
= dst_discard
;
886 dst_copy_metrics(new, &ort
->dst
);
887 new->dev
= ort
->dst
.dev
;
890 rt
->rt6i_idev
= ort
->rt6i_idev
;
892 in6_dev_hold(rt
->rt6i_idev
);
893 rt
->rt6i_expires
= 0;
895 ipv6_addr_copy(&rt
->rt6i_gateway
, &ort
->rt6i_gateway
);
896 rt
->rt6i_flags
= ort
->rt6i_flags
& ~RTF_EXPIRES
;
899 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
900 #ifdef CONFIG_IPV6_SUBTREES
901 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
907 dst_release(dst_orig
);
908 return new ? new : ERR_PTR(-ENOMEM
);
912 * Destination cache support functions
915 static struct dst_entry
*ip6_dst_check(struct dst_entry
*dst
, u32 cookie
)
919 rt
= (struct rt6_info
*) dst
;
921 if (rt
->rt6i_node
&& (rt
->rt6i_node
->fn_sernum
== cookie
)) {
922 if (rt
->rt6i_peer_genid
!= rt6_peer_genid()) {
924 rt6_bind_peer(rt
, 0);
925 rt
->rt6i_peer_genid
= rt6_peer_genid();
932 static struct dst_entry
*ip6_negative_advice(struct dst_entry
*dst
)
934 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
937 if (rt
->rt6i_flags
& RTF_CACHE
) {
938 if (rt6_check_expired(rt
)) {
950 static void ip6_link_failure(struct sk_buff
*skb
)
954 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, ICMPV6_ADDR_UNREACH
, 0);
956 rt
= (struct rt6_info
*) skb_dst(skb
);
958 if (rt
->rt6i_flags
&RTF_CACHE
) {
959 dst_set_expires(&rt
->dst
, 0);
960 rt
->rt6i_flags
|= RTF_EXPIRES
;
961 } else if (rt
->rt6i_node
&& (rt
->rt6i_flags
& RTF_DEFAULT
))
962 rt
->rt6i_node
->fn_sernum
= -1;
966 static void ip6_rt_update_pmtu(struct dst_entry
*dst
, u32 mtu
)
968 struct rt6_info
*rt6
= (struct rt6_info
*)dst
;
970 if (mtu
< dst_mtu(dst
) && rt6
->rt6i_dst
.plen
== 128) {
971 rt6
->rt6i_flags
|= RTF_MODIFIED
;
972 if (mtu
< IPV6_MIN_MTU
) {
973 u32 features
= dst_metric(dst
, RTAX_FEATURES
);
975 features
|= RTAX_FEATURE_ALLFRAG
;
976 dst_metric_set(dst
, RTAX_FEATURES
, features
);
978 dst_metric_set(dst
, RTAX_MTU
, mtu
);
982 static unsigned int ip6_default_advmss(const struct dst_entry
*dst
)
984 struct net_device
*dev
= dst
->dev
;
985 unsigned int mtu
= dst_mtu(dst
);
986 struct net
*net
= dev_net(dev
);
988 mtu
-= sizeof(struct ipv6hdr
) + sizeof(struct tcphdr
);
990 if (mtu
< net
->ipv6
.sysctl
.ip6_rt_min_advmss
)
991 mtu
= net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
994 * Maximal non-jumbo IPv6 payload is IPV6_MAXPLEN and
995 * corresponding MSS is IPV6_MAXPLEN - tcp_header_size.
996 * IPV6_MAXPLEN is also valid and means: "any MSS,
997 * rely only on pmtu discovery"
999 if (mtu
> IPV6_MAXPLEN
- sizeof(struct tcphdr
))
1004 static unsigned int ip6_default_mtu(const struct dst_entry
*dst
)
1006 unsigned int mtu
= IPV6_MIN_MTU
;
1007 struct inet6_dev
*idev
;
1010 idev
= __in6_dev_get(dst
->dev
);
1012 mtu
= idev
->cnf
.mtu6
;
1018 static struct dst_entry
*icmp6_dst_gc_list
;
1019 static DEFINE_SPINLOCK(icmp6_dst_lock
);
1021 struct dst_entry
*icmp6_dst_alloc(struct net_device
*dev
,
1022 struct neighbour
*neigh
,
1023 const struct in6_addr
*addr
)
1025 struct rt6_info
*rt
;
1026 struct inet6_dev
*idev
= in6_dev_get(dev
);
1027 struct net
*net
= dev_net(dev
);
1029 if (unlikely(idev
== NULL
))
1032 rt
= ip6_dst_alloc(&net
->ipv6
.ip6_dst_ops
);
1033 if (unlikely(rt
== NULL
)) {
1042 neigh
= ndisc_get_neigh(dev
, addr
);
1048 rt
->rt6i_idev
= idev
;
1049 rt
->rt6i_nexthop
= neigh
;
1050 atomic_set(&rt
->dst
.__refcnt
, 1);
1051 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, 255);
1052 rt
->dst
.output
= ip6_output
;
1054 #if 0 /* there's no chance to use these for ndisc */
1055 rt
->dst
.flags
= ipv6_addr_type(addr
) & IPV6_ADDR_UNICAST
1058 ipv6_addr_copy(&rt
->rt6i_dst
.addr
, addr
);
1059 rt
->rt6i_dst
.plen
= 128;
1062 spin_lock_bh(&icmp6_dst_lock
);
1063 rt
->dst
.next
= icmp6_dst_gc_list
;
1064 icmp6_dst_gc_list
= &rt
->dst
;
1065 spin_unlock_bh(&icmp6_dst_lock
);
1067 fib6_force_start_gc(net
);
1073 int icmp6_dst_gc(void)
1075 struct dst_entry
*dst
, **pprev
;
1078 spin_lock_bh(&icmp6_dst_lock
);
1079 pprev
= &icmp6_dst_gc_list
;
1081 while ((dst
= *pprev
) != NULL
) {
1082 if (!atomic_read(&dst
->__refcnt
)) {
1091 spin_unlock_bh(&icmp6_dst_lock
);
1096 static void icmp6_clean_all(int (*func
)(struct rt6_info
*rt
, void *arg
),
1099 struct dst_entry
*dst
, **pprev
;
1101 spin_lock_bh(&icmp6_dst_lock
);
1102 pprev
= &icmp6_dst_gc_list
;
1103 while ((dst
= *pprev
) != NULL
) {
1104 struct rt6_info
*rt
= (struct rt6_info
*) dst
;
1105 if (func(rt
, arg
)) {
1112 spin_unlock_bh(&icmp6_dst_lock
);
1115 static int ip6_dst_gc(struct dst_ops
*ops
)
1117 unsigned long now
= jiffies
;
1118 struct net
*net
= container_of(ops
, struct net
, ipv6
.ip6_dst_ops
);
1119 int rt_min_interval
= net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
1120 int rt_max_size
= net
->ipv6
.sysctl
.ip6_rt_max_size
;
1121 int rt_elasticity
= net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
1122 int rt_gc_timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
1123 unsigned long rt_last_gc
= net
->ipv6
.ip6_rt_last_gc
;
1126 entries
= dst_entries_get_fast(ops
);
1127 if (time_after(rt_last_gc
+ rt_min_interval
, now
) &&
1128 entries
<= rt_max_size
)
1131 net
->ipv6
.ip6_rt_gc_expire
++;
1132 fib6_run_gc(net
->ipv6
.ip6_rt_gc_expire
, net
);
1133 net
->ipv6
.ip6_rt_last_gc
= now
;
1134 entries
= dst_entries_get_slow(ops
);
1135 if (entries
< ops
->gc_thresh
)
1136 net
->ipv6
.ip6_rt_gc_expire
= rt_gc_timeout
>>1;
1138 net
->ipv6
.ip6_rt_gc_expire
-= net
->ipv6
.ip6_rt_gc_expire
>>rt_elasticity
;
1139 return entries
> rt_max_size
;
1142 /* Clean host part of a prefix. Not necessary in radix tree,
1143 but results in cleaner routing tables.
1145 Remove it only when all the things will work!
1148 int ip6_dst_hoplimit(struct dst_entry
*dst
)
1150 int hoplimit
= dst_metric_raw(dst
, RTAX_HOPLIMIT
);
1151 if (hoplimit
== 0) {
1152 struct net_device
*dev
= dst
->dev
;
1153 struct inet6_dev
*idev
;
1156 idev
= __in6_dev_get(dev
);
1158 hoplimit
= idev
->cnf
.hop_limit
;
1160 hoplimit
= dev_net(dev
)->ipv6
.devconf_all
->hop_limit
;
1165 EXPORT_SYMBOL(ip6_dst_hoplimit
);
1171 int ip6_route_add(struct fib6_config
*cfg
)
1174 struct net
*net
= cfg
->fc_nlinfo
.nl_net
;
1175 struct rt6_info
*rt
= NULL
;
1176 struct net_device
*dev
= NULL
;
1177 struct inet6_dev
*idev
= NULL
;
1178 struct fib6_table
*table
;
1181 if (cfg
->fc_dst_len
> 128 || cfg
->fc_src_len
> 128)
1183 #ifndef CONFIG_IPV6_SUBTREES
1184 if (cfg
->fc_src_len
)
1187 if (cfg
->fc_ifindex
) {
1189 dev
= dev_get_by_index(net
, cfg
->fc_ifindex
);
1192 idev
= in6_dev_get(dev
);
1197 if (cfg
->fc_metric
== 0)
1198 cfg
->fc_metric
= IP6_RT_PRIO_USER
;
1200 table
= fib6_new_table(net
, cfg
->fc_table
);
1201 if (table
== NULL
) {
1206 rt
= ip6_dst_alloc(&net
->ipv6
.ip6_dst_ops
);
1213 rt
->dst
.obsolete
= -1;
1214 rt
->rt6i_expires
= (cfg
->fc_flags
& RTF_EXPIRES
) ?
1215 jiffies
+ clock_t_to_jiffies(cfg
->fc_expires
) :
1218 if (cfg
->fc_protocol
== RTPROT_UNSPEC
)
1219 cfg
->fc_protocol
= RTPROT_BOOT
;
1220 rt
->rt6i_protocol
= cfg
->fc_protocol
;
1222 addr_type
= ipv6_addr_type(&cfg
->fc_dst
);
1224 if (addr_type
& IPV6_ADDR_MULTICAST
)
1225 rt
->dst
.input
= ip6_mc_input
;
1226 else if (cfg
->fc_flags
& RTF_LOCAL
)
1227 rt
->dst
.input
= ip6_input
;
1229 rt
->dst
.input
= ip6_forward
;
1231 rt
->dst
.output
= ip6_output
;
1233 ipv6_addr_prefix(&rt
->rt6i_dst
.addr
, &cfg
->fc_dst
, cfg
->fc_dst_len
);
1234 rt
->rt6i_dst
.plen
= cfg
->fc_dst_len
;
1235 if (rt
->rt6i_dst
.plen
== 128)
1236 rt
->dst
.flags
= DST_HOST
;
1238 #ifdef CONFIG_IPV6_SUBTREES
1239 ipv6_addr_prefix(&rt
->rt6i_src
.addr
, &cfg
->fc_src
, cfg
->fc_src_len
);
1240 rt
->rt6i_src
.plen
= cfg
->fc_src_len
;
1243 rt
->rt6i_metric
= cfg
->fc_metric
;
1245 /* We cannot add true routes via loopback here,
1246 they would result in kernel looping; promote them to reject routes
1248 if ((cfg
->fc_flags
& RTF_REJECT
) ||
1249 (dev
&& (dev
->flags
&IFF_LOOPBACK
) && !(addr_type
&IPV6_ADDR_LOOPBACK
)
1250 && !(cfg
->fc_flags
&RTF_LOCAL
))) {
1251 /* hold loopback dev/idev if we haven't done so. */
1252 if (dev
!= net
->loopback_dev
) {
1257 dev
= net
->loopback_dev
;
1259 idev
= in6_dev_get(dev
);
1265 rt
->dst
.output
= ip6_pkt_discard_out
;
1266 rt
->dst
.input
= ip6_pkt_discard
;
1267 rt
->dst
.error
= -ENETUNREACH
;
1268 rt
->rt6i_flags
= RTF_REJECT
|RTF_NONEXTHOP
;
1272 if (cfg
->fc_flags
& RTF_GATEWAY
) {
1273 struct in6_addr
*gw_addr
;
1276 gw_addr
= &cfg
->fc_gateway
;
1277 ipv6_addr_copy(&rt
->rt6i_gateway
, gw_addr
);
1278 gwa_type
= ipv6_addr_type(gw_addr
);
1280 if (gwa_type
!= (IPV6_ADDR_LINKLOCAL
|IPV6_ADDR_UNICAST
)) {
1281 struct rt6_info
*grt
;
1283 /* IPv6 strictly inhibits using not link-local
1284 addresses as nexthop address.
1285 Otherwise, router will not able to send redirects.
1286 It is very good, but in some (rare!) circumstances
1287 (SIT, PtP, NBMA NOARP links) it is handy to allow
1288 some exceptions. --ANK
1291 if (!(gwa_type
&IPV6_ADDR_UNICAST
))
1294 grt
= rt6_lookup(net
, gw_addr
, NULL
, cfg
->fc_ifindex
, 1);
1296 err
= -EHOSTUNREACH
;
1300 if (dev
!= grt
->rt6i_dev
) {
1301 dst_release(&grt
->dst
);
1305 dev
= grt
->rt6i_dev
;
1306 idev
= grt
->rt6i_idev
;
1308 in6_dev_hold(grt
->rt6i_idev
);
1310 if (!(grt
->rt6i_flags
&RTF_GATEWAY
))
1312 dst_release(&grt
->dst
);
1318 if (dev
== NULL
|| (dev
->flags
&IFF_LOOPBACK
))
1326 if (cfg
->fc_flags
& (RTF_GATEWAY
| RTF_NONEXTHOP
)) {
1327 rt
->rt6i_nexthop
= __neigh_lookup_errno(&nd_tbl
, &rt
->rt6i_gateway
, dev
);
1328 if (IS_ERR(rt
->rt6i_nexthop
)) {
1329 err
= PTR_ERR(rt
->rt6i_nexthop
);
1330 rt
->rt6i_nexthop
= NULL
;
1335 rt
->rt6i_flags
= cfg
->fc_flags
;
1342 nla_for_each_attr(nla
, cfg
->fc_mx
, cfg
->fc_mx_len
, remaining
) {
1343 int type
= nla_type(nla
);
1346 if (type
> RTAX_MAX
) {
1351 dst_metric_set(&rt
->dst
, type
, nla_get_u32(nla
));
1357 rt
->rt6i_idev
= idev
;
1358 rt
->rt6i_table
= table
;
1360 cfg
->fc_nlinfo
.nl_net
= dev_net(dev
);
1362 return __ip6_ins_rt(rt
, &cfg
->fc_nlinfo
);
1374 static int __ip6_del_rt(struct rt6_info
*rt
, struct nl_info
*info
)
1377 struct fib6_table
*table
;
1378 struct net
*net
= dev_net(rt
->rt6i_dev
);
1380 if (rt
== net
->ipv6
.ip6_null_entry
)
1383 table
= rt
->rt6i_table
;
1384 write_lock_bh(&table
->tb6_lock
);
1386 err
= fib6_del(rt
, info
);
1387 dst_release(&rt
->dst
);
1389 write_unlock_bh(&table
->tb6_lock
);
1394 int ip6_del_rt(struct rt6_info
*rt
)
1396 struct nl_info info
= {
1397 .nl_net
= dev_net(rt
->rt6i_dev
),
1399 return __ip6_del_rt(rt
, &info
);
1402 static int ip6_route_del(struct fib6_config
*cfg
)
1404 struct fib6_table
*table
;
1405 struct fib6_node
*fn
;
1406 struct rt6_info
*rt
;
1409 table
= fib6_get_table(cfg
->fc_nlinfo
.nl_net
, cfg
->fc_table
);
1413 read_lock_bh(&table
->tb6_lock
);
1415 fn
= fib6_locate(&table
->tb6_root
,
1416 &cfg
->fc_dst
, cfg
->fc_dst_len
,
1417 &cfg
->fc_src
, cfg
->fc_src_len
);
1420 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1421 if (cfg
->fc_ifindex
&&
1422 (rt
->rt6i_dev
== NULL
||
1423 rt
->rt6i_dev
->ifindex
!= cfg
->fc_ifindex
))
1425 if (cfg
->fc_flags
& RTF_GATEWAY
&&
1426 !ipv6_addr_equal(&cfg
->fc_gateway
, &rt
->rt6i_gateway
))
1428 if (cfg
->fc_metric
&& cfg
->fc_metric
!= rt
->rt6i_metric
)
1431 read_unlock_bh(&table
->tb6_lock
);
1433 return __ip6_del_rt(rt
, &cfg
->fc_nlinfo
);
1436 read_unlock_bh(&table
->tb6_lock
);
1444 struct ip6rd_flowi
{
1446 struct in6_addr gateway
;
1449 static struct rt6_info
*__ip6_route_redirect(struct net
*net
,
1450 struct fib6_table
*table
,
1454 struct ip6rd_flowi
*rdfl
= (struct ip6rd_flowi
*)fl
;
1455 struct rt6_info
*rt
;
1456 struct fib6_node
*fn
;
1459 * Get the "current" route for this destination and
1460 * check if the redirect has come from approriate router.
1462 * RFC 2461 specifies that redirects should only be
1463 * accepted if they come from the nexthop to the target.
1464 * Due to the way the routes are chosen, this notion
1465 * is a bit fuzzy and one might need to check all possible
1469 read_lock_bh(&table
->tb6_lock
);
1470 fn
= fib6_lookup(&table
->tb6_root
, &fl
->fl6_dst
, &fl
->fl6_src
);
1472 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1474 * Current route is on-link; redirect is always invalid.
1476 * Seems, previous statement is not true. It could
1477 * be node, which looks for us as on-link (f.e. proxy ndisc)
1478 * But then router serving it might decide, that we should
1479 * know truth 8)8) --ANK (980726).
1481 if (rt6_check_expired(rt
))
1483 if (!(rt
->rt6i_flags
& RTF_GATEWAY
))
1485 if (fl
->oif
!= rt
->rt6i_dev
->ifindex
)
1487 if (!ipv6_addr_equal(&rdfl
->gateway
, &rt
->rt6i_gateway
))
1493 rt
= net
->ipv6
.ip6_null_entry
;
1494 BACKTRACK(net
, &fl
->fl6_src
);
1498 read_unlock_bh(&table
->tb6_lock
);
1503 static struct rt6_info
*ip6_route_redirect(struct in6_addr
*dest
,
1504 struct in6_addr
*src
,
1505 struct in6_addr
*gateway
,
1506 struct net_device
*dev
)
1508 int flags
= RT6_LOOKUP_F_HAS_SADDR
;
1509 struct net
*net
= dev_net(dev
);
1510 struct ip6rd_flowi rdfl
= {
1512 .oif
= dev
->ifindex
,
1518 ipv6_addr_copy(&rdfl
.gateway
, gateway
);
1520 if (rt6_need_strict(dest
))
1521 flags
|= RT6_LOOKUP_F_IFACE
;
1523 return (struct rt6_info
*)fib6_rule_lookup(net
, (struct flowi
*)&rdfl
,
1524 flags
, __ip6_route_redirect
);
1527 void rt6_redirect(struct in6_addr
*dest
, struct in6_addr
*src
,
1528 struct in6_addr
*saddr
,
1529 struct neighbour
*neigh
, u8
*lladdr
, int on_link
)
1531 struct rt6_info
*rt
, *nrt
= NULL
;
1532 struct netevent_redirect netevent
;
1533 struct net
*net
= dev_net(neigh
->dev
);
1535 rt
= ip6_route_redirect(dest
, src
, saddr
, neigh
->dev
);
1537 if (rt
== net
->ipv6
.ip6_null_entry
) {
1538 if (net_ratelimit())
1539 printk(KERN_DEBUG
"rt6_redirect: source isn't a valid nexthop "
1540 "for redirect target\n");
1545 * We have finally decided to accept it.
1548 neigh_update(neigh
, lladdr
, NUD_STALE
,
1549 NEIGH_UPDATE_F_WEAK_OVERRIDE
|
1550 NEIGH_UPDATE_F_OVERRIDE
|
1551 (on_link
? 0 : (NEIGH_UPDATE_F_OVERRIDE_ISROUTER
|
1552 NEIGH_UPDATE_F_ISROUTER
))
1556 * Redirect received -> path was valid.
1557 * Look, redirects are sent only in response to data packets,
1558 * so that this nexthop apparently is reachable. --ANK
1560 dst_confirm(&rt
->dst
);
1562 /* Duplicate redirect: silently ignore. */
1563 if (neigh
== rt
->dst
.neighbour
)
1566 nrt
= ip6_rt_copy(rt
);
1570 nrt
->rt6i_flags
= RTF_GATEWAY
|RTF_UP
|RTF_DYNAMIC
|RTF_CACHE
;
1572 nrt
->rt6i_flags
&= ~RTF_GATEWAY
;
1574 ipv6_addr_copy(&nrt
->rt6i_dst
.addr
, dest
);
1575 nrt
->rt6i_dst
.plen
= 128;
1576 nrt
->dst
.flags
|= DST_HOST
;
1578 ipv6_addr_copy(&nrt
->rt6i_gateway
, (struct in6_addr
*)neigh
->primary_key
);
1579 nrt
->rt6i_nexthop
= neigh_clone(neigh
);
1581 if (ip6_ins_rt(nrt
))
1584 netevent
.old
= &rt
->dst
;
1585 netevent
.new = &nrt
->dst
;
1586 call_netevent_notifiers(NETEVENT_REDIRECT
, &netevent
);
1588 if (rt
->rt6i_flags
&RTF_CACHE
) {
1594 dst_release(&rt
->dst
);
1598 * Handle ICMP "packet too big" messages
1599 * i.e. Path MTU discovery
1602 static void rt6_do_pmtu_disc(struct in6_addr
*daddr
, struct in6_addr
*saddr
,
1603 struct net
*net
, u32 pmtu
, int ifindex
)
1605 struct rt6_info
*rt
, *nrt
;
1608 rt
= rt6_lookup(net
, daddr
, saddr
, ifindex
, 0);
1612 if (rt6_check_expired(rt
)) {
1617 if (pmtu
>= dst_mtu(&rt
->dst
))
1620 if (pmtu
< IPV6_MIN_MTU
) {
1622 * According to RFC2460, PMTU is set to the IPv6 Minimum Link
1623 * MTU (1280) and a fragment header should always be included
1624 * after a node receiving Too Big message reporting PMTU is
1625 * less than the IPv6 Minimum Link MTU.
1627 pmtu
= IPV6_MIN_MTU
;
1631 /* New mtu received -> path was valid.
1632 They are sent only in response to data packets,
1633 so that this nexthop apparently is reachable. --ANK
1635 dst_confirm(&rt
->dst
);
1637 /* Host route. If it is static, it would be better
1638 not to override it, but add new one, so that
1639 when cache entry will expire old pmtu
1640 would return automatically.
1642 if (rt
->rt6i_flags
& RTF_CACHE
) {
1643 dst_metric_set(&rt
->dst
, RTAX_MTU
, pmtu
);
1645 u32 features
= dst_metric(&rt
->dst
, RTAX_FEATURES
);
1646 features
|= RTAX_FEATURE_ALLFRAG
;
1647 dst_metric_set(&rt
->dst
, RTAX_FEATURES
, features
);
1649 dst_set_expires(&rt
->dst
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1650 rt
->rt6i_flags
|= RTF_MODIFIED
|RTF_EXPIRES
;
1655 Two cases are possible:
1656 1. It is connected route. Action: COW
1657 2. It is gatewayed route or NONEXTHOP route. Action: clone it.
1659 if (!rt
->rt6i_nexthop
&& !(rt
->rt6i_flags
& RTF_NONEXTHOP
))
1660 nrt
= rt6_alloc_cow(rt
, daddr
, saddr
);
1662 nrt
= rt6_alloc_clone(rt
, daddr
);
1665 dst_metric_set(&nrt
->dst
, RTAX_MTU
, pmtu
);
1667 u32 features
= dst_metric(&nrt
->dst
, RTAX_FEATURES
);
1668 features
|= RTAX_FEATURE_ALLFRAG
;
1669 dst_metric_set(&nrt
->dst
, RTAX_FEATURES
, features
);
1672 /* According to RFC 1981, detecting PMTU increase shouldn't be
1673 * happened within 5 mins, the recommended timer is 10 mins.
1674 * Here this route expiration time is set to ip6_rt_mtu_expires
1675 * which is 10 mins. After 10 mins the decreased pmtu is expired
1676 * and detecting PMTU increase will be automatically happened.
1678 dst_set_expires(&nrt
->dst
, net
->ipv6
.sysctl
.ip6_rt_mtu_expires
);
1679 nrt
->rt6i_flags
|= RTF_DYNAMIC
|RTF_EXPIRES
;
1684 dst_release(&rt
->dst
);
1687 void rt6_pmtu_discovery(struct in6_addr
*daddr
, struct in6_addr
*saddr
,
1688 struct net_device
*dev
, u32 pmtu
)
1690 struct net
*net
= dev_net(dev
);
1693 * RFC 1981 states that a node "MUST reduce the size of the packets it
1694 * is sending along the path" that caused the Packet Too Big message.
1695 * Since it's not possible in the general case to determine which
1696 * interface was used to send the original packet, we update the MTU
1697 * on the interface that will be used to send future packets. We also
1698 * update the MTU on the interface that received the Packet Too Big in
1699 * case the original packet was forced out that interface with
1700 * SO_BINDTODEVICE or similar. This is the next best thing to the
1701 * correct behaviour, which would be to update the MTU on all
1704 rt6_do_pmtu_disc(daddr
, saddr
, net
, pmtu
, 0);
1705 rt6_do_pmtu_disc(daddr
, saddr
, net
, pmtu
, dev
->ifindex
);
1709 * Misc support functions
1712 static struct rt6_info
* ip6_rt_copy(struct rt6_info
*ort
)
1714 struct net
*net
= dev_net(ort
->rt6i_dev
);
1715 struct rt6_info
*rt
= ip6_dst_alloc(&net
->ipv6
.ip6_dst_ops
);
1718 rt
->dst
.input
= ort
->dst
.input
;
1719 rt
->dst
.output
= ort
->dst
.output
;
1721 dst_copy_metrics(&rt
->dst
, &ort
->dst
);
1722 rt
->dst
.error
= ort
->dst
.error
;
1723 rt
->dst
.dev
= ort
->dst
.dev
;
1725 dev_hold(rt
->dst
.dev
);
1726 rt
->rt6i_idev
= ort
->rt6i_idev
;
1728 in6_dev_hold(rt
->rt6i_idev
);
1729 rt
->dst
.lastuse
= jiffies
;
1730 rt
->rt6i_expires
= 0;
1732 ipv6_addr_copy(&rt
->rt6i_gateway
, &ort
->rt6i_gateway
);
1733 rt
->rt6i_flags
= ort
->rt6i_flags
& ~RTF_EXPIRES
;
1734 rt
->rt6i_metric
= 0;
1736 memcpy(&rt
->rt6i_dst
, &ort
->rt6i_dst
, sizeof(struct rt6key
));
1737 #ifdef CONFIG_IPV6_SUBTREES
1738 memcpy(&rt
->rt6i_src
, &ort
->rt6i_src
, sizeof(struct rt6key
));
1740 rt
->rt6i_table
= ort
->rt6i_table
;
1745 #ifdef CONFIG_IPV6_ROUTE_INFO
1746 static struct rt6_info
*rt6_get_route_info(struct net
*net
,
1747 struct in6_addr
*prefix
, int prefixlen
,
1748 struct in6_addr
*gwaddr
, int ifindex
)
1750 struct fib6_node
*fn
;
1751 struct rt6_info
*rt
= NULL
;
1752 struct fib6_table
*table
;
1754 table
= fib6_get_table(net
, RT6_TABLE_INFO
);
1758 write_lock_bh(&table
->tb6_lock
);
1759 fn
= fib6_locate(&table
->tb6_root
, prefix
,prefixlen
, NULL
, 0);
1763 for (rt
= fn
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1764 if (rt
->rt6i_dev
->ifindex
!= ifindex
)
1766 if ((rt
->rt6i_flags
& (RTF_ROUTEINFO
|RTF_GATEWAY
)) != (RTF_ROUTEINFO
|RTF_GATEWAY
))
1768 if (!ipv6_addr_equal(&rt
->rt6i_gateway
, gwaddr
))
1774 write_unlock_bh(&table
->tb6_lock
);
1778 static struct rt6_info
*rt6_add_route_info(struct net
*net
,
1779 struct in6_addr
*prefix
, int prefixlen
,
1780 struct in6_addr
*gwaddr
, int ifindex
,
1783 struct fib6_config cfg
= {
1784 .fc_table
= RT6_TABLE_INFO
,
1785 .fc_metric
= IP6_RT_PRIO_USER
,
1786 .fc_ifindex
= ifindex
,
1787 .fc_dst_len
= prefixlen
,
1788 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_ROUTEINFO
|
1789 RTF_UP
| RTF_PREF(pref
),
1791 .fc_nlinfo
.nlh
= NULL
,
1792 .fc_nlinfo
.nl_net
= net
,
1795 ipv6_addr_copy(&cfg
.fc_dst
, prefix
);
1796 ipv6_addr_copy(&cfg
.fc_gateway
, gwaddr
);
1798 /* We should treat it as a default route if prefix length is 0. */
1800 cfg
.fc_flags
|= RTF_DEFAULT
;
1802 ip6_route_add(&cfg
);
1804 return rt6_get_route_info(net
, prefix
, prefixlen
, gwaddr
, ifindex
);
1808 struct rt6_info
*rt6_get_dflt_router(struct in6_addr
*addr
, struct net_device
*dev
)
1810 struct rt6_info
*rt
;
1811 struct fib6_table
*table
;
1813 table
= fib6_get_table(dev_net(dev
), RT6_TABLE_DFLT
);
1817 write_lock_bh(&table
->tb6_lock
);
1818 for (rt
= table
->tb6_root
.leaf
; rt
; rt
=rt
->dst
.rt6_next
) {
1819 if (dev
== rt
->rt6i_dev
&&
1820 ((rt
->rt6i_flags
& (RTF_ADDRCONF
| RTF_DEFAULT
)) == (RTF_ADDRCONF
| RTF_DEFAULT
)) &&
1821 ipv6_addr_equal(&rt
->rt6i_gateway
, addr
))
1826 write_unlock_bh(&table
->tb6_lock
);
1830 struct rt6_info
*rt6_add_dflt_router(struct in6_addr
*gwaddr
,
1831 struct net_device
*dev
,
1834 struct fib6_config cfg
= {
1835 .fc_table
= RT6_TABLE_DFLT
,
1836 .fc_metric
= IP6_RT_PRIO_USER
,
1837 .fc_ifindex
= dev
->ifindex
,
1838 .fc_flags
= RTF_GATEWAY
| RTF_ADDRCONF
| RTF_DEFAULT
|
1839 RTF_UP
| RTF_EXPIRES
| RTF_PREF(pref
),
1841 .fc_nlinfo
.nlh
= NULL
,
1842 .fc_nlinfo
.nl_net
= dev_net(dev
),
1845 ipv6_addr_copy(&cfg
.fc_gateway
, gwaddr
);
1847 ip6_route_add(&cfg
);
1849 return rt6_get_dflt_router(gwaddr
, dev
);
1852 void rt6_purge_dflt_routers(struct net
*net
)
1854 struct rt6_info
*rt
;
1855 struct fib6_table
*table
;
1857 /* NOTE: Keep consistent with rt6_get_dflt_router */
1858 table
= fib6_get_table(net
, RT6_TABLE_DFLT
);
1863 read_lock_bh(&table
->tb6_lock
);
1864 for (rt
= table
->tb6_root
.leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1865 if (rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) {
1867 read_unlock_bh(&table
->tb6_lock
);
1872 read_unlock_bh(&table
->tb6_lock
);
1875 static void rtmsg_to_fib6_config(struct net
*net
,
1876 struct in6_rtmsg
*rtmsg
,
1877 struct fib6_config
*cfg
)
1879 memset(cfg
, 0, sizeof(*cfg
));
1881 cfg
->fc_table
= RT6_TABLE_MAIN
;
1882 cfg
->fc_ifindex
= rtmsg
->rtmsg_ifindex
;
1883 cfg
->fc_metric
= rtmsg
->rtmsg_metric
;
1884 cfg
->fc_expires
= rtmsg
->rtmsg_info
;
1885 cfg
->fc_dst_len
= rtmsg
->rtmsg_dst_len
;
1886 cfg
->fc_src_len
= rtmsg
->rtmsg_src_len
;
1887 cfg
->fc_flags
= rtmsg
->rtmsg_flags
;
1889 cfg
->fc_nlinfo
.nl_net
= net
;
1891 ipv6_addr_copy(&cfg
->fc_dst
, &rtmsg
->rtmsg_dst
);
1892 ipv6_addr_copy(&cfg
->fc_src
, &rtmsg
->rtmsg_src
);
1893 ipv6_addr_copy(&cfg
->fc_gateway
, &rtmsg
->rtmsg_gateway
);
1896 int ipv6_route_ioctl(struct net
*net
, unsigned int cmd
, void __user
*arg
)
1898 struct fib6_config cfg
;
1899 struct in6_rtmsg rtmsg
;
1903 case SIOCADDRT
: /* Add a route */
1904 case SIOCDELRT
: /* Delete a route */
1905 if (!capable(CAP_NET_ADMIN
))
1907 err
= copy_from_user(&rtmsg
, arg
,
1908 sizeof(struct in6_rtmsg
));
1912 rtmsg_to_fib6_config(net
, &rtmsg
, &cfg
);
1917 err
= ip6_route_add(&cfg
);
1920 err
= ip6_route_del(&cfg
);
1934 * Drop the packet on the floor
1937 static int ip6_pkt_drop(struct sk_buff
*skb
, u8 code
, int ipstats_mib_noroutes
)
1940 struct dst_entry
*dst
= skb_dst(skb
);
1941 switch (ipstats_mib_noroutes
) {
1942 case IPSTATS_MIB_INNOROUTES
:
1943 type
= ipv6_addr_type(&ipv6_hdr(skb
)->daddr
);
1944 if (type
== IPV6_ADDR_ANY
) {
1945 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
1946 IPSTATS_MIB_INADDRERRORS
);
1950 case IPSTATS_MIB_OUTNOROUTES
:
1951 IP6_INC_STATS(dev_net(dst
->dev
), ip6_dst_idev(dst
),
1952 ipstats_mib_noroutes
);
1955 icmpv6_send(skb
, ICMPV6_DEST_UNREACH
, code
, 0);
1960 static int ip6_pkt_discard(struct sk_buff
*skb
)
1962 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_INNOROUTES
);
1965 static int ip6_pkt_discard_out(struct sk_buff
*skb
)
1967 skb
->dev
= skb_dst(skb
)->dev
;
1968 return ip6_pkt_drop(skb
, ICMPV6_NOROUTE
, IPSTATS_MIB_OUTNOROUTES
);
1971 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1973 static int ip6_pkt_prohibit(struct sk_buff
*skb
)
1975 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_INNOROUTES
);
1978 static int ip6_pkt_prohibit_out(struct sk_buff
*skb
)
1980 skb
->dev
= skb_dst(skb
)->dev
;
1981 return ip6_pkt_drop(skb
, ICMPV6_ADM_PROHIBITED
, IPSTATS_MIB_OUTNOROUTES
);
1987 * Allocate a dst for local (unicast / anycast) address.
1990 struct rt6_info
*addrconf_dst_alloc(struct inet6_dev
*idev
,
1991 const struct in6_addr
*addr
,
1994 struct net
*net
= dev_net(idev
->dev
);
1995 struct rt6_info
*rt
= ip6_dst_alloc(&net
->ipv6
.ip6_dst_ops
);
1996 struct neighbour
*neigh
;
1999 if (net_ratelimit())
2000 pr_warning("IPv6: Maximum number of routes reached,"
2001 " consider increasing route/max_size.\n");
2002 return ERR_PTR(-ENOMEM
);
2005 dev_hold(net
->loopback_dev
);
2008 rt
->dst
.flags
= DST_HOST
;
2009 rt
->dst
.input
= ip6_input
;
2010 rt
->dst
.output
= ip6_output
;
2011 rt
->rt6i_dev
= net
->loopback_dev
;
2012 rt
->rt6i_idev
= idev
;
2013 dst_metric_set(&rt
->dst
, RTAX_HOPLIMIT
, -1);
2014 rt
->dst
.obsolete
= -1;
2016 rt
->rt6i_flags
= RTF_UP
| RTF_NONEXTHOP
;
2018 rt
->rt6i_flags
|= RTF_ANYCAST
;
2020 rt
->rt6i_flags
|= RTF_LOCAL
;
2021 neigh
= ndisc_get_neigh(rt
->rt6i_dev
, &rt
->rt6i_gateway
);
2022 if (IS_ERR(neigh
)) {
2025 /* We are casting this because that is the return
2026 * value type. But an errno encoded pointer is the
2027 * same regardless of the underlying pointer type,
2028 * and that's what we are returning. So this is OK.
2030 return (struct rt6_info
*) neigh
;
2032 rt
->rt6i_nexthop
= neigh
;
2034 ipv6_addr_copy(&rt
->rt6i_dst
.addr
, addr
);
2035 rt
->rt6i_dst
.plen
= 128;
2036 rt
->rt6i_table
= fib6_get_table(net
, RT6_TABLE_LOCAL
);
2038 atomic_set(&rt
->dst
.__refcnt
, 1);
2043 struct arg_dev_net
{
2044 struct net_device
*dev
;
2048 static int fib6_ifdown(struct rt6_info
*rt
, void *arg
)
2050 const struct arg_dev_net
*adn
= arg
;
2051 const struct net_device
*dev
= adn
->dev
;
2053 if ((rt
->rt6i_dev
== dev
|| dev
== NULL
) &&
2054 rt
!= adn
->net
->ipv6
.ip6_null_entry
) {
2055 RT6_TRACE("deleted by ifdown %p\n", rt
);
2061 void rt6_ifdown(struct net
*net
, struct net_device
*dev
)
2063 struct arg_dev_net adn
= {
2068 fib6_clean_all(net
, fib6_ifdown
, 0, &adn
);
2069 icmp6_clean_all(fib6_ifdown
, &adn
);
2072 struct rt6_mtu_change_arg
2074 struct net_device
*dev
;
2078 static int rt6_mtu_change_route(struct rt6_info
*rt
, void *p_arg
)
2080 struct rt6_mtu_change_arg
*arg
= (struct rt6_mtu_change_arg
*) p_arg
;
2081 struct inet6_dev
*idev
;
2083 /* In IPv6 pmtu discovery is not optional,
2084 so that RTAX_MTU lock cannot disable it.
2085 We still use this lock to block changes
2086 caused by addrconf/ndisc.
2089 idev
= __in6_dev_get(arg
->dev
);
2093 /* For administrative MTU increase, there is no way to discover
2094 IPv6 PMTU increase, so PMTU increase should be updated here.
2095 Since RFC 1981 doesn't include administrative MTU increase
2096 update PMTU increase is a MUST. (i.e. jumbo frame)
2099 If new MTU is less than route PMTU, this new MTU will be the
2100 lowest MTU in the path, update the route PMTU to reflect PMTU
2101 decreases; if new MTU is greater than route PMTU, and the
2102 old MTU is the lowest MTU in the path, update the route PMTU
2103 to reflect the increase. In this case if the other nodes' MTU
2104 also have the lowest MTU, TOO BIG MESSAGE will be lead to
2107 if (rt
->rt6i_dev
== arg
->dev
&&
2108 !dst_metric_locked(&rt
->dst
, RTAX_MTU
) &&
2109 (dst_mtu(&rt
->dst
) >= arg
->mtu
||
2110 (dst_mtu(&rt
->dst
) < arg
->mtu
&&
2111 dst_mtu(&rt
->dst
) == idev
->cnf
.mtu6
))) {
2112 dst_metric_set(&rt
->dst
, RTAX_MTU
, arg
->mtu
);
2117 void rt6_mtu_change(struct net_device
*dev
, unsigned mtu
)
2119 struct rt6_mtu_change_arg arg
= {
2124 fib6_clean_all(dev_net(dev
), rt6_mtu_change_route
, 0, &arg
);
2127 static const struct nla_policy rtm_ipv6_policy
[RTA_MAX
+1] = {
2128 [RTA_GATEWAY
] = { .len
= sizeof(struct in6_addr
) },
2129 [RTA_OIF
] = { .type
= NLA_U32
},
2130 [RTA_IIF
] = { .type
= NLA_U32
},
2131 [RTA_PRIORITY
] = { .type
= NLA_U32
},
2132 [RTA_METRICS
] = { .type
= NLA_NESTED
},
2135 static int rtm_to_fib6_config(struct sk_buff
*skb
, struct nlmsghdr
*nlh
,
2136 struct fib6_config
*cfg
)
2139 struct nlattr
*tb
[RTA_MAX
+1];
2142 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2147 rtm
= nlmsg_data(nlh
);
2148 memset(cfg
, 0, sizeof(*cfg
));
2150 cfg
->fc_table
= rtm
->rtm_table
;
2151 cfg
->fc_dst_len
= rtm
->rtm_dst_len
;
2152 cfg
->fc_src_len
= rtm
->rtm_src_len
;
2153 cfg
->fc_flags
= RTF_UP
;
2154 cfg
->fc_protocol
= rtm
->rtm_protocol
;
2156 if (rtm
->rtm_type
== RTN_UNREACHABLE
)
2157 cfg
->fc_flags
|= RTF_REJECT
;
2159 if (rtm
->rtm_type
== RTN_LOCAL
)
2160 cfg
->fc_flags
|= RTF_LOCAL
;
2162 cfg
->fc_nlinfo
.pid
= NETLINK_CB(skb
).pid
;
2163 cfg
->fc_nlinfo
.nlh
= nlh
;
2164 cfg
->fc_nlinfo
.nl_net
= sock_net(skb
->sk
);
2166 if (tb
[RTA_GATEWAY
]) {
2167 nla_memcpy(&cfg
->fc_gateway
, tb
[RTA_GATEWAY
], 16);
2168 cfg
->fc_flags
|= RTF_GATEWAY
;
2172 int plen
= (rtm
->rtm_dst_len
+ 7) >> 3;
2174 if (nla_len(tb
[RTA_DST
]) < plen
)
2177 nla_memcpy(&cfg
->fc_dst
, tb
[RTA_DST
], plen
);
2181 int plen
= (rtm
->rtm_src_len
+ 7) >> 3;
2183 if (nla_len(tb
[RTA_SRC
]) < plen
)
2186 nla_memcpy(&cfg
->fc_src
, tb
[RTA_SRC
], plen
);
2190 cfg
->fc_ifindex
= nla_get_u32(tb
[RTA_OIF
]);
2192 if (tb
[RTA_PRIORITY
])
2193 cfg
->fc_metric
= nla_get_u32(tb
[RTA_PRIORITY
]);
2195 if (tb
[RTA_METRICS
]) {
2196 cfg
->fc_mx
= nla_data(tb
[RTA_METRICS
]);
2197 cfg
->fc_mx_len
= nla_len(tb
[RTA_METRICS
]);
2201 cfg
->fc_table
= nla_get_u32(tb
[RTA_TABLE
]);
2208 static int inet6_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
2210 struct fib6_config cfg
;
2213 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2217 return ip6_route_del(&cfg
);
2220 static int inet6_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
* nlh
, void *arg
)
2222 struct fib6_config cfg
;
2225 err
= rtm_to_fib6_config(skb
, nlh
, &cfg
);
2229 return ip6_route_add(&cfg
);
2232 static inline size_t rt6_nlmsg_size(void)
2234 return NLMSG_ALIGN(sizeof(struct rtmsg
))
2235 + nla_total_size(16) /* RTA_SRC */
2236 + nla_total_size(16) /* RTA_DST */
2237 + nla_total_size(16) /* RTA_GATEWAY */
2238 + nla_total_size(16) /* RTA_PREFSRC */
2239 + nla_total_size(4) /* RTA_TABLE */
2240 + nla_total_size(4) /* RTA_IIF */
2241 + nla_total_size(4) /* RTA_OIF */
2242 + nla_total_size(4) /* RTA_PRIORITY */
2243 + RTAX_MAX
* nla_total_size(4) /* RTA_METRICS */
2244 + nla_total_size(sizeof(struct rta_cacheinfo
));
2247 static int rt6_fill_node(struct net
*net
,
2248 struct sk_buff
*skb
, struct rt6_info
*rt
,
2249 struct in6_addr
*dst
, struct in6_addr
*src
,
2250 int iif
, int type
, u32 pid
, u32 seq
,
2251 int prefix
, int nowait
, unsigned int flags
)
2254 struct nlmsghdr
*nlh
;
2258 if (prefix
) { /* user wants prefix routes only */
2259 if (!(rt
->rt6i_flags
& RTF_PREFIX_RT
)) {
2260 /* success since this is not a prefix route */
2265 nlh
= nlmsg_put(skb
, pid
, seq
, type
, sizeof(*rtm
), flags
);
2269 rtm
= nlmsg_data(nlh
);
2270 rtm
->rtm_family
= AF_INET6
;
2271 rtm
->rtm_dst_len
= rt
->rt6i_dst
.plen
;
2272 rtm
->rtm_src_len
= rt
->rt6i_src
.plen
;
2275 table
= rt
->rt6i_table
->tb6_id
;
2277 table
= RT6_TABLE_UNSPEC
;
2278 rtm
->rtm_table
= table
;
2279 NLA_PUT_U32(skb
, RTA_TABLE
, table
);
2280 if (rt
->rt6i_flags
&RTF_REJECT
)
2281 rtm
->rtm_type
= RTN_UNREACHABLE
;
2282 else if (rt
->rt6i_flags
&RTF_LOCAL
)
2283 rtm
->rtm_type
= RTN_LOCAL
;
2284 else if (rt
->rt6i_dev
&& (rt
->rt6i_dev
->flags
&IFF_LOOPBACK
))
2285 rtm
->rtm_type
= RTN_LOCAL
;
2287 rtm
->rtm_type
= RTN_UNICAST
;
2289 rtm
->rtm_scope
= RT_SCOPE_UNIVERSE
;
2290 rtm
->rtm_protocol
= rt
->rt6i_protocol
;
2291 if (rt
->rt6i_flags
&RTF_DYNAMIC
)
2292 rtm
->rtm_protocol
= RTPROT_REDIRECT
;
2293 else if (rt
->rt6i_flags
& RTF_ADDRCONF
)
2294 rtm
->rtm_protocol
= RTPROT_KERNEL
;
2295 else if (rt
->rt6i_flags
&RTF_DEFAULT
)
2296 rtm
->rtm_protocol
= RTPROT_RA
;
2298 if (rt
->rt6i_flags
&RTF_CACHE
)
2299 rtm
->rtm_flags
|= RTM_F_CLONED
;
2302 NLA_PUT(skb
, RTA_DST
, 16, dst
);
2303 rtm
->rtm_dst_len
= 128;
2304 } else if (rtm
->rtm_dst_len
)
2305 NLA_PUT(skb
, RTA_DST
, 16, &rt
->rt6i_dst
.addr
);
2306 #ifdef CONFIG_IPV6_SUBTREES
2308 NLA_PUT(skb
, RTA_SRC
, 16, src
);
2309 rtm
->rtm_src_len
= 128;
2310 } else if (rtm
->rtm_src_len
)
2311 NLA_PUT(skb
, RTA_SRC
, 16, &rt
->rt6i_src
.addr
);
2314 #ifdef CONFIG_IPV6_MROUTE
2315 if (ipv6_addr_is_multicast(&rt
->rt6i_dst
.addr
)) {
2316 int err
= ip6mr_get_route(net
, skb
, rtm
, nowait
);
2321 goto nla_put_failure
;
2323 if (err
== -EMSGSIZE
)
2324 goto nla_put_failure
;
2329 NLA_PUT_U32(skb
, RTA_IIF
, iif
);
2331 struct inet6_dev
*idev
= ip6_dst_idev(&rt
->dst
);
2332 struct in6_addr saddr_buf
;
2333 if (ipv6_dev_get_saddr(net
, idev
? idev
->dev
: NULL
,
2334 dst
, 0, &saddr_buf
) == 0)
2335 NLA_PUT(skb
, RTA_PREFSRC
, 16, &saddr_buf
);
2338 if (rtnetlink_put_metrics(skb
, dst_metrics_ptr(&rt
->dst
)) < 0)
2339 goto nla_put_failure
;
2341 if (rt
->dst
.neighbour
)
2342 NLA_PUT(skb
, RTA_GATEWAY
, 16, &rt
->dst
.neighbour
->primary_key
);
2345 NLA_PUT_U32(skb
, RTA_OIF
, rt
->rt6i_dev
->ifindex
);
2347 NLA_PUT_U32(skb
, RTA_PRIORITY
, rt
->rt6i_metric
);
2349 if (!(rt
->rt6i_flags
& RTF_EXPIRES
))
2351 else if (rt
->rt6i_expires
- jiffies
< INT_MAX
)
2352 expires
= rt
->rt6i_expires
- jiffies
;
2356 if (rtnl_put_cacheinfo(skb
, &rt
->dst
, 0, 0, 0,
2357 expires
, rt
->dst
.error
) < 0)
2358 goto nla_put_failure
;
2360 return nlmsg_end(skb
, nlh
);
2363 nlmsg_cancel(skb
, nlh
);
2367 int rt6_dump_route(struct rt6_info
*rt
, void *p_arg
)
2369 struct rt6_rtnl_dump_arg
*arg
= (struct rt6_rtnl_dump_arg
*) p_arg
;
2372 if (nlmsg_len(arg
->cb
->nlh
) >= sizeof(struct rtmsg
)) {
2373 struct rtmsg
*rtm
= nlmsg_data(arg
->cb
->nlh
);
2374 prefix
= (rtm
->rtm_flags
& RTM_F_PREFIX
) != 0;
2378 return rt6_fill_node(arg
->net
,
2379 arg
->skb
, rt
, NULL
, NULL
, 0, RTM_NEWROUTE
,
2380 NETLINK_CB(arg
->cb
->skb
).pid
, arg
->cb
->nlh
->nlmsg_seq
,
2381 prefix
, 0, NLM_F_MULTI
);
2384 static int inet6_rtm_getroute(struct sk_buff
*in_skb
, struct nlmsghdr
* nlh
, void *arg
)
2386 struct net
*net
= sock_net(in_skb
->sk
);
2387 struct nlattr
*tb
[RTA_MAX
+1];
2388 struct rt6_info
*rt
;
2389 struct sk_buff
*skb
;
2394 err
= nlmsg_parse(nlh
, sizeof(*rtm
), tb
, RTA_MAX
, rtm_ipv6_policy
);
2399 memset(&fl
, 0, sizeof(fl
));
2402 if (nla_len(tb
[RTA_SRC
]) < sizeof(struct in6_addr
))
2405 ipv6_addr_copy(&fl
.fl6_src
, nla_data(tb
[RTA_SRC
]));
2409 if (nla_len(tb
[RTA_DST
]) < sizeof(struct in6_addr
))
2412 ipv6_addr_copy(&fl
.fl6_dst
, nla_data(tb
[RTA_DST
]));
2416 iif
= nla_get_u32(tb
[RTA_IIF
]);
2419 fl
.oif
= nla_get_u32(tb
[RTA_OIF
]);
2422 struct net_device
*dev
;
2423 dev
= __dev_get_by_index(net
, iif
);
2430 skb
= alloc_skb(NLMSG_GOODSIZE
, GFP_KERNEL
);
2436 /* Reserve room for dummy headers, this skb can pass
2437 through good chunk of routing engine.
2439 skb_reset_mac_header(skb
);
2440 skb_reserve(skb
, MAX_HEADER
+ sizeof(struct ipv6hdr
));
2442 rt
= (struct rt6_info
*) ip6_route_output(net
, NULL
, &fl
);
2443 skb_dst_set(skb
, &rt
->dst
);
2445 err
= rt6_fill_node(net
, skb
, rt
, &fl
.fl6_dst
, &fl
.fl6_src
, iif
,
2446 RTM_NEWROUTE
, NETLINK_CB(in_skb
).pid
,
2447 nlh
->nlmsg_seq
, 0, 0, 0);
2453 err
= rtnl_unicast(skb
, net
, NETLINK_CB(in_skb
).pid
);
2458 void inet6_rt_notify(int event
, struct rt6_info
*rt
, struct nl_info
*info
)
2460 struct sk_buff
*skb
;
2461 struct net
*net
= info
->nl_net
;
2466 seq
= info
->nlh
!= NULL
? info
->nlh
->nlmsg_seq
: 0;
2468 skb
= nlmsg_new(rt6_nlmsg_size(), gfp_any());
2472 err
= rt6_fill_node(net
, skb
, rt
, NULL
, NULL
, 0,
2473 event
, info
->pid
, seq
, 0, 0, 0);
2475 /* -EMSGSIZE implies BUG in rt6_nlmsg_size() */
2476 WARN_ON(err
== -EMSGSIZE
);
2480 rtnl_notify(skb
, net
, info
->pid
, RTNLGRP_IPV6_ROUTE
,
2481 info
->nlh
, gfp_any());
2485 rtnl_set_sk_err(net
, RTNLGRP_IPV6_ROUTE
, err
);
2488 static int ip6_route_dev_notify(struct notifier_block
*this,
2489 unsigned long event
, void *data
)
2491 struct net_device
*dev
= (struct net_device
*)data
;
2492 struct net
*net
= dev_net(dev
);
2494 if (event
== NETDEV_REGISTER
&& (dev
->flags
& IFF_LOOPBACK
)) {
2495 net
->ipv6
.ip6_null_entry
->dst
.dev
= dev
;
2496 net
->ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(dev
);
2497 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2498 net
->ipv6
.ip6_prohibit_entry
->dst
.dev
= dev
;
2499 net
->ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(dev
);
2500 net
->ipv6
.ip6_blk_hole_entry
->dst
.dev
= dev
;
2501 net
->ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(dev
);
2512 #ifdef CONFIG_PROC_FS
2523 static int rt6_info_route(struct rt6_info
*rt
, void *p_arg
)
2525 struct seq_file
*m
= p_arg
;
2527 seq_printf(m
, "%pi6 %02x ", &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
);
2529 #ifdef CONFIG_IPV6_SUBTREES
2530 seq_printf(m
, "%pi6 %02x ", &rt
->rt6i_src
.addr
, rt
->rt6i_src
.plen
);
2532 seq_puts(m
, "00000000000000000000000000000000 00 ");
2535 if (rt
->rt6i_nexthop
) {
2536 seq_printf(m
, "%pi6", rt
->rt6i_nexthop
->primary_key
);
2538 seq_puts(m
, "00000000000000000000000000000000");
2540 seq_printf(m
, " %08x %08x %08x %08x %8s\n",
2541 rt
->rt6i_metric
, atomic_read(&rt
->dst
.__refcnt
),
2542 rt
->dst
.__use
, rt
->rt6i_flags
,
2543 rt
->rt6i_dev
? rt
->rt6i_dev
->name
: "");
2547 static int ipv6_route_show(struct seq_file
*m
, void *v
)
2549 struct net
*net
= (struct net
*)m
->private;
2550 fib6_clean_all(net
, rt6_info_route
, 0, m
);
2554 static int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2556 return single_open_net(inode
, file
, ipv6_route_show
);
2559 static const struct file_operations ipv6_route_proc_fops
= {
2560 .owner
= THIS_MODULE
,
2561 .open
= ipv6_route_open
,
2563 .llseek
= seq_lseek
,
2564 .release
= single_release_net
,
2567 static int rt6_stats_seq_show(struct seq_file
*seq
, void *v
)
2569 struct net
*net
= (struct net
*)seq
->private;
2570 seq_printf(seq
, "%04x %04x %04x %04x %04x %04x %04x\n",
2571 net
->ipv6
.rt6_stats
->fib_nodes
,
2572 net
->ipv6
.rt6_stats
->fib_route_nodes
,
2573 net
->ipv6
.rt6_stats
->fib_rt_alloc
,
2574 net
->ipv6
.rt6_stats
->fib_rt_entries
,
2575 net
->ipv6
.rt6_stats
->fib_rt_cache
,
2576 dst_entries_get_slow(&net
->ipv6
.ip6_dst_ops
),
2577 net
->ipv6
.rt6_stats
->fib_discarded_routes
);
2582 static int rt6_stats_seq_open(struct inode
*inode
, struct file
*file
)
2584 return single_open_net(inode
, file
, rt6_stats_seq_show
);
2587 static const struct file_operations rt6_stats_seq_fops
= {
2588 .owner
= THIS_MODULE
,
2589 .open
= rt6_stats_seq_open
,
2591 .llseek
= seq_lseek
,
2592 .release
= single_release_net
,
2594 #endif /* CONFIG_PROC_FS */
2596 #ifdef CONFIG_SYSCTL
2599 int ipv6_sysctl_rtcache_flush(ctl_table
*ctl
, int write
,
2600 void __user
*buffer
, size_t *lenp
, loff_t
*ppos
)
2602 struct net
*net
= current
->nsproxy
->net_ns
;
2603 int delay
= net
->ipv6
.sysctl
.flush_delay
;
2605 proc_dointvec(ctl
, write
, buffer
, lenp
, ppos
);
2606 fib6_run_gc(delay
<= 0 ? ~0UL : (unsigned long)delay
, net
);
2612 ctl_table ipv6_route_table_template
[] = {
2614 .procname
= "flush",
2615 .data
= &init_net
.ipv6
.sysctl
.flush_delay
,
2616 .maxlen
= sizeof(int),
2618 .proc_handler
= ipv6_sysctl_rtcache_flush
2621 .procname
= "gc_thresh",
2622 .data
= &ip6_dst_ops_template
.gc_thresh
,
2623 .maxlen
= sizeof(int),
2625 .proc_handler
= proc_dointvec
,
2628 .procname
= "max_size",
2629 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_max_size
,
2630 .maxlen
= sizeof(int),
2632 .proc_handler
= proc_dointvec
,
2635 .procname
= "gc_min_interval",
2636 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2637 .maxlen
= sizeof(int),
2639 .proc_handler
= proc_dointvec_jiffies
,
2642 .procname
= "gc_timeout",
2643 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_timeout
,
2644 .maxlen
= sizeof(int),
2646 .proc_handler
= proc_dointvec_jiffies
,
2649 .procname
= "gc_interval",
2650 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_interval
,
2651 .maxlen
= sizeof(int),
2653 .proc_handler
= proc_dointvec_jiffies
,
2656 .procname
= "gc_elasticity",
2657 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_elasticity
,
2658 .maxlen
= sizeof(int),
2660 .proc_handler
= proc_dointvec
,
2663 .procname
= "mtu_expires",
2664 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_mtu_expires
,
2665 .maxlen
= sizeof(int),
2667 .proc_handler
= proc_dointvec_jiffies
,
2670 .procname
= "min_adv_mss",
2671 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_min_advmss
,
2672 .maxlen
= sizeof(int),
2674 .proc_handler
= proc_dointvec
,
2677 .procname
= "gc_min_interval_ms",
2678 .data
= &init_net
.ipv6
.sysctl
.ip6_rt_gc_min_interval
,
2679 .maxlen
= sizeof(int),
2681 .proc_handler
= proc_dointvec_ms_jiffies
,
2686 struct ctl_table
* __net_init
ipv6_route_sysctl_init(struct net
*net
)
2688 struct ctl_table
*table
;
2690 table
= kmemdup(ipv6_route_table_template
,
2691 sizeof(ipv6_route_table_template
),
2695 table
[0].data
= &net
->ipv6
.sysctl
.flush_delay
;
2696 table
[1].data
= &net
->ipv6
.ip6_dst_ops
.gc_thresh
;
2697 table
[2].data
= &net
->ipv6
.sysctl
.ip6_rt_max_size
;
2698 table
[3].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2699 table
[4].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_timeout
;
2700 table
[5].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
2701 table
[6].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
;
2702 table
[7].data
= &net
->ipv6
.sysctl
.ip6_rt_mtu_expires
;
2703 table
[8].data
= &net
->ipv6
.sysctl
.ip6_rt_min_advmss
;
2704 table
[9].data
= &net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
;
2711 static int __net_init
ip6_route_net_init(struct net
*net
)
2715 memcpy(&net
->ipv6
.ip6_dst_ops
, &ip6_dst_ops_template
,
2716 sizeof(net
->ipv6
.ip6_dst_ops
));
2718 if (dst_entries_init(&net
->ipv6
.ip6_dst_ops
) < 0)
2719 goto out_ip6_dst_ops
;
2721 net
->ipv6
.ip6_null_entry
= kmemdup(&ip6_null_entry_template
,
2722 sizeof(*net
->ipv6
.ip6_null_entry
),
2724 if (!net
->ipv6
.ip6_null_entry
)
2725 goto out_ip6_dst_entries
;
2726 net
->ipv6
.ip6_null_entry
->dst
.path
=
2727 (struct dst_entry
*)net
->ipv6
.ip6_null_entry
;
2728 net
->ipv6
.ip6_null_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
2729 dst_init_metrics(&net
->ipv6
.ip6_null_entry
->dst
,
2730 ip6_template_metrics
, true);
2732 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2733 net
->ipv6
.ip6_prohibit_entry
= kmemdup(&ip6_prohibit_entry_template
,
2734 sizeof(*net
->ipv6
.ip6_prohibit_entry
),
2736 if (!net
->ipv6
.ip6_prohibit_entry
)
2737 goto out_ip6_null_entry
;
2738 net
->ipv6
.ip6_prohibit_entry
->dst
.path
=
2739 (struct dst_entry
*)net
->ipv6
.ip6_prohibit_entry
;
2740 net
->ipv6
.ip6_prohibit_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
2741 dst_init_metrics(&net
->ipv6
.ip6_prohibit_entry
->dst
,
2742 ip6_template_metrics
, true);
2744 net
->ipv6
.ip6_blk_hole_entry
= kmemdup(&ip6_blk_hole_entry_template
,
2745 sizeof(*net
->ipv6
.ip6_blk_hole_entry
),
2747 if (!net
->ipv6
.ip6_blk_hole_entry
)
2748 goto out_ip6_prohibit_entry
;
2749 net
->ipv6
.ip6_blk_hole_entry
->dst
.path
=
2750 (struct dst_entry
*)net
->ipv6
.ip6_blk_hole_entry
;
2751 net
->ipv6
.ip6_blk_hole_entry
->dst
.ops
= &net
->ipv6
.ip6_dst_ops
;
2752 dst_init_metrics(&net
->ipv6
.ip6_blk_hole_entry
->dst
,
2753 ip6_template_metrics
, true);
2756 net
->ipv6
.sysctl
.flush_delay
= 0;
2757 net
->ipv6
.sysctl
.ip6_rt_max_size
= 4096;
2758 net
->ipv6
.sysctl
.ip6_rt_gc_min_interval
= HZ
/ 2;
2759 net
->ipv6
.sysctl
.ip6_rt_gc_timeout
= 60*HZ
;
2760 net
->ipv6
.sysctl
.ip6_rt_gc_interval
= 30*HZ
;
2761 net
->ipv6
.sysctl
.ip6_rt_gc_elasticity
= 9;
2762 net
->ipv6
.sysctl
.ip6_rt_mtu_expires
= 10*60*HZ
;
2763 net
->ipv6
.sysctl
.ip6_rt_min_advmss
= IPV6_MIN_MTU
- 20 - 40;
2765 #ifdef CONFIG_PROC_FS
2766 proc_net_fops_create(net
, "ipv6_route", 0, &ipv6_route_proc_fops
);
2767 proc_net_fops_create(net
, "rt6_stats", S_IRUGO
, &rt6_stats_seq_fops
);
2769 net
->ipv6
.ip6_rt_gc_expire
= 30*HZ
;
2775 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2776 out_ip6_prohibit_entry
:
2777 kfree(net
->ipv6
.ip6_prohibit_entry
);
2779 kfree(net
->ipv6
.ip6_null_entry
);
2781 out_ip6_dst_entries
:
2782 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
2787 static void __net_exit
ip6_route_net_exit(struct net
*net
)
2789 #ifdef CONFIG_PROC_FS
2790 proc_net_remove(net
, "ipv6_route");
2791 proc_net_remove(net
, "rt6_stats");
2793 kfree(net
->ipv6
.ip6_null_entry
);
2794 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2795 kfree(net
->ipv6
.ip6_prohibit_entry
);
2796 kfree(net
->ipv6
.ip6_blk_hole_entry
);
2798 dst_entries_destroy(&net
->ipv6
.ip6_dst_ops
);
2801 static struct pernet_operations ip6_route_net_ops
= {
2802 .init
= ip6_route_net_init
,
2803 .exit
= ip6_route_net_exit
,
2806 static struct notifier_block ip6_route_dev_notifier
= {
2807 .notifier_call
= ip6_route_dev_notify
,
2811 int __init
ip6_route_init(void)
2816 ip6_dst_ops_template
.kmem_cachep
=
2817 kmem_cache_create("ip6_dst_cache", sizeof(struct rt6_info
), 0,
2818 SLAB_HWCACHE_ALIGN
, NULL
);
2819 if (!ip6_dst_ops_template
.kmem_cachep
)
2822 ret
= dst_entries_init(&ip6_dst_blackhole_ops
);
2824 goto out_kmem_cache
;
2826 ret
= register_pernet_subsys(&ip6_route_net_ops
);
2828 goto out_dst_entries
;
2830 ip6_dst_blackhole_ops
.kmem_cachep
= ip6_dst_ops_template
.kmem_cachep
;
2832 /* Registering of the loopback is done before this portion of code,
2833 * the loopback reference in rt6_info will not be taken, do it
2834 * manually for init_net */
2835 init_net
.ipv6
.ip6_null_entry
->dst
.dev
= init_net
.loopback_dev
;
2836 init_net
.ipv6
.ip6_null_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
2837 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
2838 init_net
.ipv6
.ip6_prohibit_entry
->dst
.dev
= init_net
.loopback_dev
;
2839 init_net
.ipv6
.ip6_prohibit_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
2840 init_net
.ipv6
.ip6_blk_hole_entry
->dst
.dev
= init_net
.loopback_dev
;
2841 init_net
.ipv6
.ip6_blk_hole_entry
->rt6i_idev
= in6_dev_get(init_net
.loopback_dev
);
2845 goto out_register_subsys
;
2851 ret
= fib6_rules_init();
2856 if (__rtnl_register(PF_INET6
, RTM_NEWROUTE
, inet6_rtm_newroute
, NULL
) ||
2857 __rtnl_register(PF_INET6
, RTM_DELROUTE
, inet6_rtm_delroute
, NULL
) ||
2858 __rtnl_register(PF_INET6
, RTM_GETROUTE
, inet6_rtm_getroute
, NULL
))
2859 goto fib6_rules_init
;
2861 ret
= register_netdevice_notifier(&ip6_route_dev_notifier
);
2863 goto fib6_rules_init
;
2869 fib6_rules_cleanup();
2874 out_register_subsys
:
2875 unregister_pernet_subsys(&ip6_route_net_ops
);
2877 dst_entries_destroy(&ip6_dst_blackhole_ops
);
2879 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);
2883 void ip6_route_cleanup(void)
2885 unregister_netdevice_notifier(&ip6_route_dev_notifier
);
2886 fib6_rules_cleanup();
2889 unregister_pernet_subsys(&ip6_route_net_ops
);
2890 dst_entries_destroy(&ip6_dst_blackhole_ops
);
2891 kmem_cache_destroy(ip6_dst_ops_template
.kmem_cachep
);