2 * Linux INET6 implementation
3 * Forwarding Information Database
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 * Yuji SEKIYA @USAGI: Support default route on router node;
17 * remove ip6_null_entry from the top of
19 * Ville Nuorvala: Fixed routing subtrees.
21 #include <linux/errno.h>
22 #include <linux/types.h>
23 #include <linux/net.h>
24 #include <linux/route.h>
25 #include <linux/netdevice.h>
26 #include <linux/in6.h>
27 #include <linux/init.h>
28 #include <linux/list.h>
31 #include <linux/proc_fs.h>
35 #include <net/ndisc.h>
36 #include <net/addrconf.h>
38 #include <net/ip6_fib.h>
39 #include <net/ip6_route.h>
44 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
46 #define RT6_TRACE(x...) do { ; } while (0)
49 static struct kmem_cache
* fib6_node_kmem __read_mostly
;
53 #ifdef CONFIG_IPV6_SUBTREES
64 struct fib6_walker_t w
;
66 int (*func
)(struct rt6_info
*, void *arg
);
70 static DEFINE_RWLOCK(fib6_walker_lock
);
72 #ifdef CONFIG_IPV6_SUBTREES
73 #define FWS_INIT FWS_S
75 #define FWS_INIT FWS_L
78 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
,
80 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
);
81 static struct fib6_node
*fib6_repair_tree(struct net
*net
, struct fib6_node
*fn
);
82 static int fib6_walk(struct fib6_walker_t
*w
);
83 static int fib6_walk_continue(struct fib6_walker_t
*w
);
86 * A routing update causes an increase of the serial number on the
87 * affected subtree. This allows for cached routes to be asynchronously
88 * tested when modifications are made to the destination cache as a
89 * result of redirects, path MTU changes, etc.
92 static __u32 rt_sernum
;
94 static void fib6_gc_timer_cb(unsigned long arg
);
96 static struct fib6_walker_t fib6_walker_list
= {
97 .prev
= &fib6_walker_list
,
98 .next
= &fib6_walker_list
,
101 #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next)
103 static inline void fib6_walker_link(struct fib6_walker_t
*w
)
105 write_lock_bh(&fib6_walker_lock
);
106 w
->next
= fib6_walker_list
.next
;
107 w
->prev
= &fib6_walker_list
;
110 write_unlock_bh(&fib6_walker_lock
);
113 static inline void fib6_walker_unlink(struct fib6_walker_t
*w
)
115 write_lock_bh(&fib6_walker_lock
);
116 w
->next
->prev
= w
->prev
;
117 w
->prev
->next
= w
->next
;
118 w
->prev
= w
->next
= w
;
119 write_unlock_bh(&fib6_walker_lock
);
121 static __inline__ u32
fib6_new_sernum(void)
130 * Auxiliary address test functions for the radix tree.
132 * These assume a 32bit processor (although it will work on
140 static __inline__ __be32
addr_bit_set(void *token
, int fn_bit
)
142 __be32
*addr
= token
;
144 return htonl(1 << ((~fn_bit
)&0x1F)) & addr
[fn_bit
>>5];
147 static __inline__
struct fib6_node
* node_alloc(void)
149 struct fib6_node
*fn
;
151 fn
= kmem_cache_zalloc(fib6_node_kmem
, GFP_ATOMIC
);
156 static __inline__
void node_free(struct fib6_node
* fn
)
158 kmem_cache_free(fib6_node_kmem
, fn
);
161 static __inline__
void rt6_release(struct rt6_info
*rt
)
163 if (atomic_dec_and_test(&rt
->rt6i_ref
))
164 dst_free(&rt
->u
.dst
);
167 static void fib6_link_table(struct net
*net
, struct fib6_table
*tb
)
172 * Initialize table lock at a single place to give lockdep a key,
173 * tables aren't visible prior to being linked to the list.
175 rwlock_init(&tb
->tb6_lock
);
177 h
= tb
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1);
180 * No protection necessary, this is the only list mutatation
181 * operation, tables never disappear once they exist.
183 hlist_add_head_rcu(&tb
->tb6_hlist
, &net
->ipv6
.fib_table_hash
[h
]);
186 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
188 static struct fib6_table
*fib6_alloc_table(struct net
*net
, u32 id
)
190 struct fib6_table
*table
;
192 table
= kzalloc(sizeof(*table
), GFP_ATOMIC
);
195 table
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
196 table
->tb6_root
.fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
202 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
204 struct fib6_table
*tb
;
208 tb
= fib6_get_table(net
, id
);
212 tb
= fib6_alloc_table(net
, id
);
214 fib6_link_table(net
, tb
);
219 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
221 struct fib6_table
*tb
;
222 struct hlist_head
*head
;
223 struct hlist_node
*node
;
228 h
= id
& (FIB6_TABLE_HASHSZ
- 1);
230 head
= &net
->ipv6
.fib_table_hash
[h
];
231 hlist_for_each_entry_rcu(tb
, node
, head
, tb6_hlist
) {
232 if (tb
->tb6_id
== id
) {
242 static void fib6_tables_init(struct net
*net
)
244 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
245 fib6_link_table(net
, net
->ipv6
.fib6_local_tbl
);
249 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
251 return fib6_get_table(net
, id
);
254 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
256 return net
->ipv6
.fib6_main_tbl
;
259 struct dst_entry
*fib6_rule_lookup(struct net
*net
, struct flowi
*fl
,
260 int flags
, pol_lookup_t lookup
)
262 return (struct dst_entry
*) lookup(net
, net
->ipv6
.fib6_main_tbl
, fl
, flags
);
265 static void fib6_tables_init(struct net
*net
)
267 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
272 static int fib6_dump_node(struct fib6_walker_t
*w
)
277 for (rt
= w
->leaf
; rt
; rt
= rt
->u
.dst
.rt6_next
) {
278 res
= rt6_dump_route(rt
, w
->args
);
280 /* Frame is full, suspend walking */
290 static void fib6_dump_end(struct netlink_callback
*cb
)
292 struct fib6_walker_t
*w
= (void*)cb
->args
[2];
297 fib6_walker_unlink(w
);
302 cb
->done
= (void*)cb
->args
[3];
306 static int fib6_dump_done(struct netlink_callback
*cb
)
309 return cb
->done
? cb
->done(cb
) : 0;
312 static int fib6_dump_table(struct fib6_table
*table
, struct sk_buff
*skb
,
313 struct netlink_callback
*cb
)
315 struct fib6_walker_t
*w
;
318 w
= (void *)cb
->args
[2];
319 w
->root
= &table
->tb6_root
;
321 if (cb
->args
[4] == 0) {
322 read_lock_bh(&table
->tb6_lock
);
324 read_unlock_bh(&table
->tb6_lock
);
328 read_lock_bh(&table
->tb6_lock
);
329 res
= fib6_walk_continue(w
);
330 read_unlock_bh(&table
->tb6_lock
);
332 fib6_walker_unlink(w
);
340 static int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
342 struct net
*net
= sock_net(skb
->sk
);
344 unsigned int e
= 0, s_e
;
345 struct rt6_rtnl_dump_arg arg
;
346 struct fib6_walker_t
*w
;
347 struct fib6_table
*tb
;
348 struct hlist_node
*node
;
349 struct hlist_head
*head
;
355 w
= (void *)cb
->args
[2];
359 * 1. hook callback destructor.
361 cb
->args
[3] = (long)cb
->done
;
362 cb
->done
= fib6_dump_done
;
365 * 2. allocate and initialize walker.
367 w
= kzalloc(sizeof(*w
), GFP_ATOMIC
);
370 w
->func
= fib6_dump_node
;
371 cb
->args
[2] = (long)w
;
379 for (h
= s_h
; h
< FIB6_TABLE_HASHSZ
; h
++, s_e
= 0) {
381 head
= &net
->ipv6
.fib_table_hash
[h
];
382 hlist_for_each_entry(tb
, node
, head
, tb6_hlist
) {
385 res
= fib6_dump_table(tb
, skb
, cb
);
396 res
= res
< 0 ? res
: skb
->len
;
405 * return the appropriate node for a routing tree "add" operation
406 * by either creating and inserting or by returning an existing
410 static struct fib6_node
* fib6_add_1(struct fib6_node
*root
, void *addr
,
411 int addrlen
, int plen
,
414 struct fib6_node
*fn
, *in
, *ln
;
415 struct fib6_node
*pn
= NULL
;
419 __u32 sernum
= fib6_new_sernum();
421 RT6_TRACE("fib6_add_1\n");
423 /* insert node in tree */
428 key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
433 if (plen
< fn
->fn_bit
||
434 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
441 if (plen
== fn
->fn_bit
) {
442 /* clean up an intermediate node */
443 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
444 rt6_release(fn
->leaf
);
448 fn
->fn_sernum
= sernum
;
454 * We have more bits to go
457 /* Try to walk down on tree. */
458 fn
->fn_sernum
= sernum
;
459 dir
= addr_bit_set(addr
, fn
->fn_bit
);
461 fn
= dir
? fn
->right
: fn
->left
;
465 * We walked to the bottom of tree.
466 * Create new leaf node without children.
476 ln
->fn_sernum
= sernum
;
488 * split since we don't have a common prefix anymore or
489 * we have a less significant route.
490 * we've to insert an intermediate node on the list
491 * this new node will point to the one we need to create
497 /* find 1st bit in difference between the 2 addrs.
499 See comment in __ipv6_addr_diff: bit may be an invalid value,
500 but if it is >= plen, the value is ignored in any case.
503 bit
= __ipv6_addr_diff(addr
, &key
->addr
, addrlen
);
508 * (new leaf node)[ln] (old node)[fn]
514 if (in
== NULL
|| ln
== NULL
) {
523 * new intermediate node.
525 * be off since that an address that chooses one of
526 * the branches would not match less specific routes
527 * in the other branch
534 atomic_inc(&in
->leaf
->rt6i_ref
);
536 in
->fn_sernum
= sernum
;
538 /* update parent pointer */
549 ln
->fn_sernum
= sernum
;
551 if (addr_bit_set(addr
, bit
)) {
558 } else { /* plen <= bit */
561 * (new leaf node)[ln]
563 * (old node)[fn] NULL
575 ln
->fn_sernum
= sernum
;
582 if (addr_bit_set(&key
->addr
, plen
))
593 * Insert routing information in a node.
596 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
597 struct nl_info
*info
)
599 struct rt6_info
*iter
= NULL
;
600 struct rt6_info
**ins
;
604 for (iter
= fn
->leaf
; iter
; iter
=iter
->u
.dst
.rt6_next
) {
606 * Search for duplicates
609 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
611 * Same priority level
614 if (iter
->rt6i_dev
== rt
->rt6i_dev
&&
615 iter
->rt6i_idev
== rt
->rt6i_idev
&&
616 ipv6_addr_equal(&iter
->rt6i_gateway
,
617 &rt
->rt6i_gateway
)) {
618 if (!(iter
->rt6i_flags
&RTF_EXPIRES
))
620 iter
->rt6i_expires
= rt
->rt6i_expires
;
621 if (!(rt
->rt6i_flags
&RTF_EXPIRES
)) {
622 iter
->rt6i_flags
&= ~RTF_EXPIRES
;
623 iter
->rt6i_expires
= 0;
629 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
632 ins
= &iter
->u
.dst
.rt6_next
;
635 /* Reset round-robin state, if necessary */
636 if (ins
== &fn
->leaf
)
643 rt
->u
.dst
.rt6_next
= iter
;
646 atomic_inc(&rt
->rt6i_ref
);
647 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
);
648 info
->nl_net
->ipv6
.rt6_stats
->fib_rt_entries
++;
650 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
651 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
652 fn
->fn_flags
|= RTN_RTINFO
;
658 static __inline__
void fib6_start_gc(struct net
*net
, struct rt6_info
*rt
)
660 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
) &&
661 (rt
->rt6i_flags
& (RTF_EXPIRES
|RTF_CACHE
)))
662 mod_timer(&net
->ipv6
.ip6_fib_timer
,
663 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
666 void fib6_force_start_gc(struct net
*net
)
668 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
))
669 mod_timer(&net
->ipv6
.ip6_fib_timer
,
670 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
674 * Add routing information to the routing tree.
675 * <destination addr>/<source addr>
676 * with source addr info in sub-trees
679 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
, struct nl_info
*info
)
681 struct fib6_node
*fn
, *pn
= NULL
;
684 fn
= fib6_add_1(root
, &rt
->rt6i_dst
.addr
, sizeof(struct in6_addr
),
685 rt
->rt6i_dst
.plen
, offsetof(struct rt6_info
, rt6i_dst
));
692 #ifdef CONFIG_IPV6_SUBTREES
693 if (rt
->rt6i_src
.plen
) {
694 struct fib6_node
*sn
;
696 if (fn
->subtree
== NULL
) {
697 struct fib6_node
*sfn
;
709 /* Create subtree root node */
714 sfn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
715 atomic_inc(&info
->nl_net
->ipv6
.ip6_null_entry
->rt6i_ref
);
716 sfn
->fn_flags
= RTN_ROOT
;
717 sfn
->fn_sernum
= fib6_new_sernum();
719 /* Now add the first leaf node to new subtree */
721 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
722 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
723 offsetof(struct rt6_info
, rt6i_src
));
726 /* If it is failed, discard just allocated
727 root, and then (in st_failure) stale node
734 /* Now link new subtree to main tree */
738 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
739 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
740 offsetof(struct rt6_info
, rt6i_src
));
746 if (fn
->leaf
== NULL
) {
748 atomic_inc(&rt
->rt6i_ref
);
754 err
= fib6_add_rt2node(fn
, rt
, info
);
757 fib6_start_gc(info
->nl_net
, rt
);
758 if (!(rt
->rt6i_flags
&RTF_CACHE
))
759 fib6_prune_clones(info
->nl_net
, pn
, rt
);
764 #ifdef CONFIG_IPV6_SUBTREES
766 * If fib6_add_1 has cleared the old leaf pointer in the
767 * super-tree leaf node we have to find a new one for it.
769 if (pn
!= fn
&& pn
->leaf
== rt
) {
771 atomic_dec(&rt
->rt6i_ref
);
773 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
774 pn
->leaf
= fib6_find_prefix(info
->nl_net
, pn
);
777 WARN_ON(pn
->leaf
== NULL
);
778 pn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
781 atomic_inc(&pn
->leaf
->rt6i_ref
);
784 dst_free(&rt
->u
.dst
);
788 #ifdef CONFIG_IPV6_SUBTREES
789 /* Subtree creation failed, probably main tree node
790 is orphan. If it is, shoot it.
793 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
794 fib6_repair_tree(info
->nl_net
, fn
);
795 dst_free(&rt
->u
.dst
);
801 * Routing tree lookup
806 int offset
; /* key offset on rt6_info */
807 struct in6_addr
*addr
; /* search key */
810 static struct fib6_node
* fib6_lookup_1(struct fib6_node
*root
,
811 struct lookup_args
*args
)
813 struct fib6_node
*fn
;
816 if (unlikely(args
->offset
== 0))
826 struct fib6_node
*next
;
828 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
830 next
= dir
? fn
->right
: fn
->left
;
841 if (FIB6_SUBTREE(fn
) || fn
->fn_flags
& RTN_RTINFO
) {
844 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
847 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
)) {
848 #ifdef CONFIG_IPV6_SUBTREES
850 fn
= fib6_lookup_1(fn
->subtree
, args
+ 1);
852 if (!fn
|| fn
->fn_flags
& RTN_RTINFO
)
857 if (fn
->fn_flags
& RTN_ROOT
)
866 struct fib6_node
* fib6_lookup(struct fib6_node
*root
, struct in6_addr
*daddr
,
867 struct in6_addr
*saddr
)
869 struct fib6_node
*fn
;
870 struct lookup_args args
[] = {
872 .offset
= offsetof(struct rt6_info
, rt6i_dst
),
875 #ifdef CONFIG_IPV6_SUBTREES
877 .offset
= offsetof(struct rt6_info
, rt6i_src
),
882 .offset
= 0, /* sentinel */
886 fn
= fib6_lookup_1(root
, daddr
? args
: args
+ 1);
888 if (fn
== NULL
|| fn
->fn_flags
& RTN_TL_ROOT
)
895 * Get node with specified destination prefix (and source prefix,
896 * if subtrees are used)
900 static struct fib6_node
* fib6_locate_1(struct fib6_node
*root
,
901 struct in6_addr
*addr
,
902 int plen
, int offset
)
904 struct fib6_node
*fn
;
906 for (fn
= root
; fn
; ) {
907 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
912 if (plen
< fn
->fn_bit
||
913 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
916 if (plen
== fn
->fn_bit
)
920 * We have more bits to go
922 if (addr_bit_set(addr
, fn
->fn_bit
))
930 struct fib6_node
* fib6_locate(struct fib6_node
*root
,
931 struct in6_addr
*daddr
, int dst_len
,
932 struct in6_addr
*saddr
, int src_len
)
934 struct fib6_node
*fn
;
936 fn
= fib6_locate_1(root
, daddr
, dst_len
,
937 offsetof(struct rt6_info
, rt6i_dst
));
939 #ifdef CONFIG_IPV6_SUBTREES
941 WARN_ON(saddr
== NULL
);
942 if (fn
&& fn
->subtree
)
943 fn
= fib6_locate_1(fn
->subtree
, saddr
, src_len
,
944 offsetof(struct rt6_info
, rt6i_src
));
948 if (fn
&& fn
->fn_flags
&RTN_RTINFO
)
960 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
)
962 if (fn
->fn_flags
&RTN_ROOT
)
963 return net
->ipv6
.ip6_null_entry
;
967 return fn
->left
->leaf
;
970 return fn
->right
->leaf
;
972 fn
= FIB6_SUBTREE(fn
);
978 * Called to trim the tree of intermediate nodes when possible. "fn"
979 * is the node we want to try and remove.
982 static struct fib6_node
*fib6_repair_tree(struct net
*net
,
983 struct fib6_node
*fn
)
987 struct fib6_node
*child
, *pn
;
988 struct fib6_walker_t
*w
;
992 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
995 WARN_ON(fn
->fn_flags
& RTN_RTINFO
);
996 WARN_ON(fn
->fn_flags
& RTN_TL_ROOT
);
997 WARN_ON(fn
->leaf
!= NULL
);
1001 if (fn
->right
) child
= fn
->right
, children
|= 1;
1002 if (fn
->left
) child
= fn
->left
, children
|= 2;
1004 if (children
== 3 || FIB6_SUBTREE(fn
)
1005 #ifdef CONFIG_IPV6_SUBTREES
1006 /* Subtree root (i.e. fn) may have one child */
1007 || (children
&& fn
->fn_flags
&RTN_ROOT
)
1010 fn
->leaf
= fib6_find_prefix(net
, fn
);
1012 if (fn
->leaf
==NULL
) {
1014 fn
->leaf
= net
->ipv6
.ip6_null_entry
;
1017 atomic_inc(&fn
->leaf
->rt6i_ref
);
1022 #ifdef CONFIG_IPV6_SUBTREES
1023 if (FIB6_SUBTREE(pn
) == fn
) {
1024 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1025 FIB6_SUBTREE(pn
) = NULL
;
1028 WARN_ON(fn
->fn_flags
& RTN_ROOT
);
1030 if (pn
->right
== fn
) pn
->right
= child
;
1031 else if (pn
->left
== fn
) pn
->left
= child
;
1039 #ifdef CONFIG_IPV6_SUBTREES
1043 read_lock(&fib6_walker_lock
);
1045 if (child
== NULL
) {
1046 if (w
->root
== fn
) {
1047 w
->root
= w
->node
= NULL
;
1048 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1049 } else if (w
->node
== fn
) {
1050 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1055 if (w
->root
== fn
) {
1057 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1059 if (w
->node
== fn
) {
1062 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1063 w
->state
= w
->state
>=FWS_R
? FWS_U
: FWS_INIT
;
1065 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1066 w
->state
= w
->state
>=FWS_C
? FWS_U
: FWS_INIT
;
1071 read_unlock(&fib6_walker_lock
);
1074 if (pn
->fn_flags
&RTN_RTINFO
|| FIB6_SUBTREE(pn
))
1077 rt6_release(pn
->leaf
);
1083 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1084 struct nl_info
*info
)
1086 struct fib6_walker_t
*w
;
1087 struct rt6_info
*rt
= *rtp
;
1088 struct net
*net
= info
->nl_net
;
1090 RT6_TRACE("fib6_del_route\n");
1093 *rtp
= rt
->u
.dst
.rt6_next
;
1094 rt
->rt6i_node
= NULL
;
1095 net
->ipv6
.rt6_stats
->fib_rt_entries
--;
1096 net
->ipv6
.rt6_stats
->fib_discarded_routes
++;
1098 /* Reset round-robin state, if necessary */
1099 if (fn
->rr_ptr
== rt
)
1102 /* Adjust walkers */
1103 read_lock(&fib6_walker_lock
);
1105 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1106 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1107 w
->leaf
= rt
->u
.dst
.rt6_next
;
1108 if (w
->leaf
== NULL
)
1112 read_unlock(&fib6_walker_lock
);
1114 rt
->u
.dst
.rt6_next
= NULL
;
1116 /* If it was last route, expunge its radix tree node */
1117 if (fn
->leaf
== NULL
) {
1118 fn
->fn_flags
&= ~RTN_RTINFO
;
1119 net
->ipv6
.rt6_stats
->fib_route_nodes
--;
1120 fn
= fib6_repair_tree(net
, fn
);
1123 if (atomic_read(&rt
->rt6i_ref
) != 1) {
1124 /* This route is used as dummy address holder in some split
1125 * nodes. It is not leaked, but it still holds other resources,
1126 * which must be released in time. So, scan ascendant nodes
1127 * and replace dummy references to this route with references
1128 * to still alive ones.
1131 if (!(fn
->fn_flags
&RTN_RTINFO
) && fn
->leaf
== rt
) {
1132 fn
->leaf
= fib6_find_prefix(net
, fn
);
1133 atomic_inc(&fn
->leaf
->rt6i_ref
);
1138 /* No more references are possible at this point. */
1139 BUG_ON(atomic_read(&rt
->rt6i_ref
) != 1);
1142 inet6_rt_notify(RTM_DELROUTE
, rt
, info
);
1146 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1148 struct net
*net
= info
->nl_net
;
1149 struct fib6_node
*fn
= rt
->rt6i_node
;
1150 struct rt6_info
**rtp
;
1153 if (rt
->u
.dst
.obsolete
>0) {
1154 WARN_ON(fn
!= NULL
);
1158 if (fn
== NULL
|| rt
== net
->ipv6
.ip6_null_entry
)
1161 WARN_ON(!(fn
->fn_flags
& RTN_RTINFO
));
1163 if (!(rt
->rt6i_flags
&RTF_CACHE
)) {
1164 struct fib6_node
*pn
= fn
;
1165 #ifdef CONFIG_IPV6_SUBTREES
1166 /* clones of this route might be in another subtree */
1167 if (rt
->rt6i_src
.plen
) {
1168 while (!(pn
->fn_flags
&RTN_ROOT
))
1173 fib6_prune_clones(info
->nl_net
, pn
, rt
);
1177 * Walk the leaf entries looking for ourself
1180 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->u
.dst
.rt6_next
) {
1182 fib6_del_route(fn
, rtp
, info
);
1190 * Tree traversal function.
1192 * Certainly, it is not interrupt safe.
1193 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1194 * It means, that we can modify tree during walking
1195 * and use this function for garbage collection, clone pruning,
1196 * cleaning tree when a device goes down etc. etc.
1198 * It guarantees that every node will be traversed,
1199 * and that it will be traversed only once.
1201 * Callback function w->func may return:
1202 * 0 -> continue walking.
1203 * positive value -> walking is suspended (used by tree dumps,
1204 * and probably by gc, if it will be split to several slices)
1205 * negative value -> terminate walking.
1207 * The function itself returns:
1208 * 0 -> walk is complete.
1209 * >0 -> walk is incomplete (i.e. suspended)
1210 * <0 -> walk is terminated by an error.
1213 static int fib6_walk_continue(struct fib6_walker_t
*w
)
1215 struct fib6_node
*fn
, *pn
;
1222 if (w
->prune
&& fn
!= w
->root
&&
1223 fn
->fn_flags
&RTN_RTINFO
&& w
->state
< FWS_C
) {
1228 #ifdef CONFIG_IPV6_SUBTREES
1230 if (FIB6_SUBTREE(fn
)) {
1231 w
->node
= FIB6_SUBTREE(fn
);
1239 w
->state
= FWS_INIT
;
1245 w
->node
= fn
->right
;
1246 w
->state
= FWS_INIT
;
1252 if (w
->leaf
&& fn
->fn_flags
&RTN_RTINFO
) {
1253 int err
= w
->func(w
);
1264 #ifdef CONFIG_IPV6_SUBTREES
1265 if (FIB6_SUBTREE(pn
) == fn
) {
1266 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1271 if (pn
->left
== fn
) {
1275 if (pn
->right
== fn
) {
1277 w
->leaf
= w
->node
->leaf
;
1287 static int fib6_walk(struct fib6_walker_t
*w
)
1291 w
->state
= FWS_INIT
;
1294 fib6_walker_link(w
);
1295 res
= fib6_walk_continue(w
);
1297 fib6_walker_unlink(w
);
1301 static int fib6_clean_node(struct fib6_walker_t
*w
)
1304 struct rt6_info
*rt
;
1305 struct fib6_cleaner_t
*c
= container_of(w
, struct fib6_cleaner_t
, w
);
1306 struct nl_info info
= {
1310 for (rt
= w
->leaf
; rt
; rt
= rt
->u
.dst
.rt6_next
) {
1311 res
= c
->func(rt
, c
->arg
);
1314 res
= fib6_del(rt
, &info
);
1317 printk(KERN_DEBUG
"fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt
, rt
->rt6i_node
, res
);
1330 * Convenient frontend to tree walker.
1332 * func is called on each route.
1333 * It may return -1 -> delete this route.
1334 * 0 -> continue walking
1336 * prune==1 -> only immediate children of node (certainly,
1337 * ignoring pure split nodes) will be scanned.
1340 static void fib6_clean_tree(struct net
*net
, struct fib6_node
*root
,
1341 int (*func
)(struct rt6_info
*, void *arg
),
1342 int prune
, void *arg
)
1344 struct fib6_cleaner_t c
;
1347 c
.w
.func
= fib6_clean_node
;
1356 void fib6_clean_all(struct net
*net
, int (*func
)(struct rt6_info
*, void *arg
),
1357 int prune
, void *arg
)
1359 struct fib6_table
*table
;
1360 struct hlist_node
*node
;
1361 struct hlist_head
*head
;
1365 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1366 head
= &net
->ipv6
.fib_table_hash
[h
];
1367 hlist_for_each_entry_rcu(table
, node
, head
, tb6_hlist
) {
1368 write_lock_bh(&table
->tb6_lock
);
1369 fib6_clean_tree(net
, &table
->tb6_root
,
1371 write_unlock_bh(&table
->tb6_lock
);
1377 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1379 if (rt
->rt6i_flags
& RTF_CACHE
) {
1380 RT6_TRACE("pruning clone %p\n", rt
);
1387 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
,
1388 struct rt6_info
*rt
)
1390 fib6_clean_tree(net
, fn
, fib6_prune_clone
, 1, rt
);
1394 * Garbage collection
1397 static struct fib6_gc_args
1403 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1405 unsigned long now
= jiffies
;
1408 * check addrconf expiration here.
1409 * Routes are expired even if they are in use.
1411 * Also age clones. Note, that clones are aged out
1412 * only if they are not in use now.
1415 if (rt
->rt6i_flags
&RTF_EXPIRES
&& rt
->rt6i_expires
) {
1416 if (time_after(now
, rt
->rt6i_expires
)) {
1417 RT6_TRACE("expiring %p\n", rt
);
1421 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1422 if (atomic_read(&rt
->u
.dst
.__refcnt
) == 0 &&
1423 time_after_eq(now
, rt
->u
.dst
.lastuse
+ gc_args
.timeout
)) {
1424 RT6_TRACE("aging clone %p\n", rt
);
1426 } else if ((rt
->rt6i_flags
& RTF_GATEWAY
) &&
1427 (!(rt
->rt6i_nexthop
->flags
& NTF_ROUTER
))) {
1428 RT6_TRACE("purging route %p via non-router but gateway\n",
1438 static DEFINE_SPINLOCK(fib6_gc_lock
);
1440 void fib6_run_gc(unsigned long expires
, struct net
*net
)
1442 if (expires
!= ~0UL) {
1443 spin_lock_bh(&fib6_gc_lock
);
1444 gc_args
.timeout
= expires
? (int)expires
:
1445 net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1447 if (!spin_trylock_bh(&fib6_gc_lock
)) {
1448 mod_timer(&net
->ipv6
.ip6_fib_timer
, jiffies
+ HZ
);
1451 gc_args
.timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1454 gc_args
.more
= icmp6_dst_gc();
1456 fib6_clean_all(net
, fib6_age
, 0, NULL
);
1459 mod_timer(&net
->ipv6
.ip6_fib_timer
,
1460 round_jiffies(jiffies
1461 + net
->ipv6
.sysctl
.ip6_rt_gc_interval
));
1463 del_timer(&net
->ipv6
.ip6_fib_timer
);
1464 spin_unlock_bh(&fib6_gc_lock
);
1467 static void fib6_gc_timer_cb(unsigned long arg
)
1469 fib6_run_gc(0, (struct net
*)arg
);
1472 static int fib6_net_init(struct net
*net
)
1474 setup_timer(&net
->ipv6
.ip6_fib_timer
, fib6_gc_timer_cb
, (unsigned long)net
);
1476 net
->ipv6
.rt6_stats
= kzalloc(sizeof(*net
->ipv6
.rt6_stats
), GFP_KERNEL
);
1477 if (!net
->ipv6
.rt6_stats
)
1480 net
->ipv6
.fib_table_hash
= kcalloc(FIB6_TABLE_HASHSZ
,
1481 sizeof(*net
->ipv6
.fib_table_hash
),
1483 if (!net
->ipv6
.fib_table_hash
)
1486 net
->ipv6
.fib6_main_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_main_tbl
),
1488 if (!net
->ipv6
.fib6_main_tbl
)
1489 goto out_fib_table_hash
;
1491 net
->ipv6
.fib6_main_tbl
->tb6_id
= RT6_TABLE_MAIN
;
1492 net
->ipv6
.fib6_main_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1493 net
->ipv6
.fib6_main_tbl
->tb6_root
.fn_flags
=
1494 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1496 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1497 net
->ipv6
.fib6_local_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_local_tbl
),
1499 if (!net
->ipv6
.fib6_local_tbl
)
1500 goto out_fib6_main_tbl
;
1501 net
->ipv6
.fib6_local_tbl
->tb6_id
= RT6_TABLE_LOCAL
;
1502 net
->ipv6
.fib6_local_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1503 net
->ipv6
.fib6_local_tbl
->tb6_root
.fn_flags
=
1504 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1506 fib6_tables_init(net
);
1510 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1512 kfree(net
->ipv6
.fib6_main_tbl
);
1515 kfree(net
->ipv6
.fib_table_hash
);
1517 kfree(net
->ipv6
.rt6_stats
);
1522 static void fib6_net_exit(struct net
*net
)
1524 rt6_ifdown(net
, NULL
);
1525 del_timer_sync(&net
->ipv6
.ip6_fib_timer
);
1527 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1528 kfree(net
->ipv6
.fib6_local_tbl
);
1530 kfree(net
->ipv6
.fib6_main_tbl
);
1531 kfree(net
->ipv6
.fib_table_hash
);
1532 kfree(net
->ipv6
.rt6_stats
);
1535 static struct pernet_operations fib6_net_ops
= {
1536 .init
= fib6_net_init
,
1537 .exit
= fib6_net_exit
,
1540 int __init
fib6_init(void)
1544 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1545 sizeof(struct fib6_node
),
1546 0, SLAB_HWCACHE_ALIGN
,
1548 if (!fib6_node_kmem
)
1551 ret
= register_pernet_subsys(&fib6_net_ops
);
1553 goto out_kmem_cache_create
;
1555 ret
= __rtnl_register(PF_INET6
, RTM_GETROUTE
, NULL
, inet6_dump_fib
);
1557 goto out_unregister_subsys
;
1561 out_unregister_subsys
:
1562 unregister_pernet_subsys(&fib6_net_ops
);
1563 out_kmem_cache_create
:
1564 kmem_cache_destroy(fib6_node_kmem
);
1568 void fib6_gc_cleanup(void)
1570 unregister_pernet_subsys(&fib6_net_ops
);
1571 kmem_cache_destroy(fib6_node_kmem
);