2 * Linux INET6 implementation
3 * Forwarding Information Database
6 * Pedro Roque <roque@di.fc.ul.pt>
8 * $Id: ip6_fib.c,v 1.25 2001/10/31 21:55:55 davem Exp $
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public License
12 * as published by the Free Software Foundation; either version
13 * 2 of the License, or (at your option) any later version.
18 * Yuji SEKIYA @USAGI: Support default route on router node;
19 * remove ip6_null_entry from the top of
22 #include <linux/errno.h>
23 #include <linux/types.h>
24 #include <linux/net.h>
25 #include <linux/route.h>
26 #include <linux/netdevice.h>
27 #include <linux/in6.h>
28 #include <linux/init.h>
29 #include <linux/list.h>
32 #include <linux/proc_fs.h>
36 #include <net/ndisc.h>
37 #include <net/addrconf.h>
39 #include <net/ip6_fib.h>
40 #include <net/ip6_route.h>
45 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
47 #define RT6_TRACE(x...) do { ; } while (0)
50 struct rt6_statistics rt6_stats
;
52 static kmem_cache_t
* fib6_node_kmem __read_mostly
;
56 #ifdef CONFIG_IPV6_SUBTREES
67 struct fib6_walker_t w
;
68 int (*func
)(struct rt6_info
*, void *arg
);
72 static DEFINE_RWLOCK(fib6_walker_lock
);
74 #ifdef CONFIG_IPV6_SUBTREES
75 #define FWS_INIT FWS_S
76 #define SUBTREE(fn) ((fn)->subtree)
78 #define FWS_INIT FWS_L
79 #define SUBTREE(fn) NULL
82 static void fib6_prune_clones(struct fib6_node
*fn
, struct rt6_info
*rt
);
83 static struct rt6_info
* fib6_find_prefix(struct fib6_node
*fn
);
84 static struct fib6_node
* fib6_repair_tree(struct fib6_node
*fn
);
85 static int fib6_walk(struct fib6_walker_t
*w
);
86 static int fib6_walk_continue(struct fib6_walker_t
*w
);
89 * A routing update causes an increase of the serial number on the
90 * affected subtree. This allows for cached routes to be asynchronously
91 * tested when modifications are made to the destination cache as a
92 * result of redirects, path MTU changes, etc.
95 static __u32 rt_sernum
;
97 static DEFINE_TIMER(ip6_fib_timer
, fib6_run_gc
, 0, 0);
99 static struct fib6_walker_t fib6_walker_list
= {
100 .prev
= &fib6_walker_list
,
101 .next
= &fib6_walker_list
,
104 #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next)
106 static inline void fib6_walker_link(struct fib6_walker_t
*w
)
108 write_lock_bh(&fib6_walker_lock
);
109 w
->next
= fib6_walker_list
.next
;
110 w
->prev
= &fib6_walker_list
;
113 write_unlock_bh(&fib6_walker_lock
);
116 static inline void fib6_walker_unlink(struct fib6_walker_t
*w
)
118 write_lock_bh(&fib6_walker_lock
);
119 w
->next
->prev
= w
->prev
;
120 w
->prev
->next
= w
->next
;
121 w
->prev
= w
->next
= w
;
122 write_unlock_bh(&fib6_walker_lock
);
124 static __inline__ u32
fib6_new_sernum(void)
133 * Auxiliary address test functions for the radix tree.
135 * These assume a 32bit processor (although it will work on
143 static __inline__
int addr_bit_set(void *token
, int fn_bit
)
147 return htonl(1 << ((~fn_bit
)&0x1F)) & addr
[fn_bit
>>5];
150 static __inline__
struct fib6_node
* node_alloc(void)
152 struct fib6_node
*fn
;
154 if ((fn
= kmem_cache_alloc(fib6_node_kmem
, SLAB_ATOMIC
)) != NULL
)
155 memset(fn
, 0, sizeof(struct fib6_node
));
160 static __inline__
void node_free(struct fib6_node
* fn
)
162 kmem_cache_free(fib6_node_kmem
, fn
);
165 static __inline__
void rt6_release(struct rt6_info
*rt
)
167 if (atomic_dec_and_test(&rt
->rt6i_ref
))
168 dst_free(&rt
->u
.dst
);
171 static struct fib6_table fib6_main_tbl
= {
172 .tb6_id
= RT6_TABLE_MAIN
,
173 .tb6_lock
= RW_LOCK_UNLOCKED
,
175 .leaf
= &ip6_null_entry
,
176 .fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
,
180 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
181 #define FIB_TABLE_HASHSZ 256
183 #define FIB_TABLE_HASHSZ 1
185 static struct hlist_head fib_table_hash
[FIB_TABLE_HASHSZ
];
187 static void fib6_link_table(struct fib6_table
*tb
)
191 h
= tb
->tb6_id
& (FIB_TABLE_HASHSZ
- 1);
194 * No protection necessary, this is the only list mutatation
195 * operation, tables never disappear once they exist.
197 hlist_add_head_rcu(&tb
->tb6_hlist
, &fib_table_hash
[h
]);
200 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
201 static struct fib6_table fib6_local_tbl
= {
202 .tb6_id
= RT6_TABLE_LOCAL
,
203 .tb6_lock
= RW_LOCK_UNLOCKED
,
205 .leaf
= &ip6_null_entry
,
206 .fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
,
210 static struct fib6_table
*fib6_alloc_table(u32 id
)
212 struct fib6_table
*table
;
214 table
= kzalloc(sizeof(*table
), GFP_ATOMIC
);
217 table
->tb6_lock
= RW_LOCK_UNLOCKED
;
218 table
->tb6_root
.leaf
= &ip6_null_entry
;
219 table
->tb6_root
.fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
225 struct fib6_table
*fib6_new_table(u32 id
)
227 struct fib6_table
*tb
;
231 tb
= fib6_get_table(id
);
235 tb
= fib6_alloc_table(id
);
242 struct fib6_table
*fib6_get_table(u32 id
)
244 struct fib6_table
*tb
;
245 struct hlist_node
*node
;
250 h
= id
& (FIB_TABLE_HASHSZ
- 1);
252 hlist_for_each_entry_rcu(tb
, node
, &fib_table_hash
[h
], tb6_hlist
) {
253 if (tb
->tb6_id
== id
) {
263 static void __init
fib6_tables_init(void)
265 fib6_link_table(&fib6_main_tbl
);
266 fib6_link_table(&fib6_local_tbl
);
271 struct fib6_table
*fib6_new_table(u32 id
)
273 return fib6_get_table(id
);
276 struct fib6_table
*fib6_get_table(u32 id
)
278 return &fib6_main_tbl
;
281 struct dst_entry
*fib6_rule_lookup(struct flowi
*fl
, int flags
,
284 return (struct dst_entry
*) lookup(&fib6_main_tbl
, fl
, flags
);
287 static void __init
fib6_tables_init(void)
289 fib6_link_table(&fib6_main_tbl
);
294 static int fib6_dump_node(struct fib6_walker_t
*w
)
299 for (rt
= w
->leaf
; rt
; rt
= rt
->u
.next
) {
300 res
= rt6_dump_route(rt
, w
->args
);
302 /* Frame is full, suspend walking */
312 static void fib6_dump_end(struct netlink_callback
*cb
)
314 struct fib6_walker_t
*w
= (void*)cb
->args
[2];
320 cb
->done
= (void*)cb
->args
[3];
324 static int fib6_dump_done(struct netlink_callback
*cb
)
327 return cb
->done
? cb
->done(cb
) : 0;
330 static int fib6_dump_table(struct fib6_table
*table
, struct sk_buff
*skb
,
331 struct netlink_callback
*cb
)
333 struct fib6_walker_t
*w
;
336 w
= (void *)cb
->args
[2];
337 w
->root
= &table
->tb6_root
;
339 if (cb
->args
[4] == 0) {
340 read_lock_bh(&table
->tb6_lock
);
342 read_unlock_bh(&table
->tb6_lock
);
346 read_lock_bh(&table
->tb6_lock
);
347 res
= fib6_walk_continue(w
);
348 read_unlock_bh(&table
->tb6_lock
);
351 fib6_walker_unlink(w
);
354 fib6_walker_unlink(w
);
361 int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
364 unsigned int e
= 0, s_e
;
365 struct rt6_rtnl_dump_arg arg
;
366 struct fib6_walker_t
*w
;
367 struct fib6_table
*tb
;
368 struct hlist_node
*node
;
374 w
= (void *)cb
->args
[2];
378 * 1. hook callback destructor.
380 cb
->args
[3] = (long)cb
->done
;
381 cb
->done
= fib6_dump_done
;
384 * 2. allocate and initialize walker.
386 w
= kzalloc(sizeof(*w
), GFP_ATOMIC
);
389 w
->func
= fib6_dump_node
;
390 cb
->args
[2] = (long)w
;
397 for (h
= s_h
; h
< FIB_TABLE_HASHSZ
; h
++, s_e
= 0) {
399 hlist_for_each_entry(tb
, node
, &fib_table_hash
[h
], tb6_hlist
) {
402 res
= fib6_dump_table(tb
, skb
, cb
);
413 res
= res
< 0 ? res
: skb
->len
;
422 * return the appropriate node for a routing tree "add" operation
423 * by either creating and inserting or by returning an existing
427 static struct fib6_node
* fib6_add_1(struct fib6_node
*root
, void *addr
,
428 int addrlen
, int plen
,
431 struct fib6_node
*fn
, *in
, *ln
;
432 struct fib6_node
*pn
= NULL
;
436 __u32 sernum
= fib6_new_sernum();
438 RT6_TRACE("fib6_add_1\n");
440 /* insert node in tree */
445 key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
450 if (plen
< fn
->fn_bit
||
451 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
458 if (plen
== fn
->fn_bit
) {
459 /* clean up an intermediate node */
460 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
461 rt6_release(fn
->leaf
);
465 fn
->fn_sernum
= sernum
;
471 * We have more bits to go
474 /* Try to walk down on tree. */
475 fn
->fn_sernum
= sernum
;
476 dir
= addr_bit_set(addr
, fn
->fn_bit
);
478 fn
= dir
? fn
->right
: fn
->left
;
482 * We walked to the bottom of tree.
483 * Create new leaf node without children.
493 ln
->fn_sernum
= sernum
;
505 * split since we don't have a common prefix anymore or
506 * we have a less significant route.
507 * we've to insert an intermediate node on the list
508 * this new node will point to the one we need to create
514 /* find 1st bit in difference between the 2 addrs.
516 See comment in __ipv6_addr_diff: bit may be an invalid value,
517 but if it is >= plen, the value is ignored in any case.
520 bit
= __ipv6_addr_diff(addr
, &key
->addr
, addrlen
);
525 * (new leaf node)[ln] (old node)[fn]
531 if (in
== NULL
|| ln
== NULL
) {
540 * new intermediate node.
542 * be off since that an address that chooses one of
543 * the branches would not match less specific routes
544 * in the other branch
551 atomic_inc(&in
->leaf
->rt6i_ref
);
553 in
->fn_sernum
= sernum
;
555 /* update parent pointer */
566 ln
->fn_sernum
= sernum
;
568 if (addr_bit_set(addr
, bit
)) {
575 } else { /* plen <= bit */
578 * (new leaf node)[ln]
580 * (old node)[fn] NULL
592 ln
->fn_sernum
= sernum
;
599 if (addr_bit_set(&key
->addr
, plen
))
610 * Insert routing information in a node.
613 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
614 struct nl_info
*info
)
616 struct rt6_info
*iter
= NULL
;
617 struct rt6_info
**ins
;
621 if (fn
->fn_flags
&RTN_TL_ROOT
&&
622 fn
->leaf
== &ip6_null_entry
&&
623 !(rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) ){
629 for (iter
= fn
->leaf
; iter
; iter
=iter
->u
.next
) {
631 * Search for duplicates
634 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
636 * Same priority level
639 if (iter
->rt6i_dev
== rt
->rt6i_dev
&&
640 iter
->rt6i_idev
== rt
->rt6i_idev
&&
641 ipv6_addr_equal(&iter
->rt6i_gateway
,
642 &rt
->rt6i_gateway
)) {
643 if (!(iter
->rt6i_flags
&RTF_EXPIRES
))
645 iter
->rt6i_expires
= rt
->rt6i_expires
;
646 if (!(rt
->rt6i_flags
&RTF_EXPIRES
)) {
647 iter
->rt6i_flags
&= ~RTF_EXPIRES
;
648 iter
->rt6i_expires
= 0;
654 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
668 atomic_inc(&rt
->rt6i_ref
);
669 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
);
670 rt6_stats
.fib_rt_entries
++;
672 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
673 rt6_stats
.fib_route_nodes
++;
674 fn
->fn_flags
|= RTN_RTINFO
;
680 static __inline__
void fib6_start_gc(struct rt6_info
*rt
)
682 if (ip6_fib_timer
.expires
== 0 &&
683 (rt
->rt6i_flags
& (RTF_EXPIRES
|RTF_CACHE
)))
684 mod_timer(&ip6_fib_timer
, jiffies
+ ip6_rt_gc_interval
);
687 void fib6_force_start_gc(void)
689 if (ip6_fib_timer
.expires
== 0)
690 mod_timer(&ip6_fib_timer
, jiffies
+ ip6_rt_gc_interval
);
694 * Add routing information to the routing tree.
695 * <destination addr>/<source addr>
696 * with source addr info in sub-trees
699 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
, struct nl_info
*info
)
701 struct fib6_node
*fn
, *pn
= NULL
;
704 fn
= fib6_add_1(root
, &rt
->rt6i_dst
.addr
, sizeof(struct in6_addr
),
705 rt
->rt6i_dst
.plen
, offsetof(struct rt6_info
, rt6i_dst
));
712 #ifdef CONFIG_IPV6_SUBTREES
713 if (rt
->rt6i_src
.plen
) {
714 struct fib6_node
*sn
;
716 if (fn
->subtree
== NULL
) {
717 struct fib6_node
*sfn
;
729 /* Create subtree root node */
734 sfn
->leaf
= &ip6_null_entry
;
735 atomic_inc(&ip6_null_entry
.rt6i_ref
);
736 sfn
->fn_flags
= RTN_ROOT
;
737 sfn
->fn_sernum
= fib6_new_sernum();
739 /* Now add the first leaf node to new subtree */
741 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
742 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
743 offsetof(struct rt6_info
, rt6i_src
));
746 /* If it is failed, discard just allocated
747 root, and then (in st_failure) stale node
754 /* Now link new subtree to main tree */
758 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
759 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
760 offsetof(struct rt6_info
, rt6i_src
));
766 if (fn
->leaf
== NULL
) {
768 atomic_inc(&rt
->rt6i_ref
);
774 err
= fib6_add_rt2node(fn
, rt
, info
);
778 if (!(rt
->rt6i_flags
&RTF_CACHE
))
779 fib6_prune_clones(pn
, rt
);
784 #ifdef CONFIG_IPV6_SUBTREES
786 * If fib6_add_1 has cleared the old leaf pointer in the
787 * super-tree leaf node we have to find a new one for it.
789 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
790 pn
->leaf
= fib6_find_prefix(pn
);
793 BUG_TRAP(pn
->leaf
!= NULL
);
794 pn
->leaf
= &ip6_null_entry
;
797 atomic_inc(&pn
->leaf
->rt6i_ref
);
800 dst_free(&rt
->u
.dst
);
804 #ifdef CONFIG_IPV6_SUBTREES
805 /* Subtree creation failed, probably main tree node
806 is orphan. If it is, shoot it.
809 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
810 fib6_repair_tree(fn
);
811 dst_free(&rt
->u
.dst
);
817 * Routing tree lookup
822 int offset
; /* key offset on rt6_info */
823 struct in6_addr
*addr
; /* search key */
826 static struct fib6_node
* fib6_lookup_1(struct fib6_node
*root
,
827 struct lookup_args
*args
)
829 struct fib6_node
*fn
;
832 if (unlikely(args
->offset
== 0))
842 struct fib6_node
*next
;
844 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
846 next
= dir
? fn
->right
: fn
->left
;
857 if (SUBTREE(fn
) || fn
->fn_flags
& RTN_RTINFO
) {
860 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
863 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
)) {
864 #ifdef CONFIG_IPV6_SUBTREES
866 fn
= fib6_lookup_1(fn
->subtree
, args
+ 1);
868 if (!fn
|| fn
->fn_flags
& RTN_RTINFO
)
873 if (fn
->fn_flags
& RTN_ROOT
)
882 struct fib6_node
* fib6_lookup(struct fib6_node
*root
, struct in6_addr
*daddr
,
883 struct in6_addr
*saddr
)
885 struct fib6_node
*fn
;
886 struct lookup_args args
[] = {
888 .offset
= offsetof(struct rt6_info
, rt6i_dst
),
891 #ifdef CONFIG_IPV6_SUBTREES
893 .offset
= offsetof(struct rt6_info
, rt6i_src
),
898 .offset
= 0, /* sentinel */
902 fn
= fib6_lookup_1(root
, args
);
904 if (fn
== NULL
|| fn
->fn_flags
& RTN_TL_ROOT
)
911 * Get node with specified destination prefix (and source prefix,
912 * if subtrees are used)
916 static struct fib6_node
* fib6_locate_1(struct fib6_node
*root
,
917 struct in6_addr
*addr
,
918 int plen
, int offset
)
920 struct fib6_node
*fn
;
922 for (fn
= root
; fn
; ) {
923 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
928 if (plen
< fn
->fn_bit
||
929 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
932 if (plen
== fn
->fn_bit
)
936 * We have more bits to go
938 if (addr_bit_set(addr
, fn
->fn_bit
))
946 struct fib6_node
* fib6_locate(struct fib6_node
*root
,
947 struct in6_addr
*daddr
, int dst_len
,
948 struct in6_addr
*saddr
, int src_len
)
950 struct fib6_node
*fn
;
952 fn
= fib6_locate_1(root
, daddr
, dst_len
,
953 offsetof(struct rt6_info
, rt6i_dst
));
955 #ifdef CONFIG_IPV6_SUBTREES
957 BUG_TRAP(saddr
!=NULL
);
958 if (fn
&& fn
->subtree
)
959 fn
= fib6_locate_1(fn
->subtree
, saddr
, src_len
,
960 offsetof(struct rt6_info
, rt6i_src
));
964 if (fn
&& fn
->fn_flags
&RTN_RTINFO
)
976 static struct rt6_info
* fib6_find_prefix(struct fib6_node
*fn
)
978 if (fn
->fn_flags
&RTN_ROOT
)
979 return &ip6_null_entry
;
983 return fn
->left
->leaf
;
986 return fn
->right
->leaf
;
994 * Called to trim the tree of intermediate nodes when possible. "fn"
995 * is the node we want to try and remove.
998 static struct fib6_node
* fib6_repair_tree(struct fib6_node
*fn
)
1002 struct fib6_node
*child
, *pn
;
1003 struct fib6_walker_t
*w
;
1007 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
1010 BUG_TRAP(!(fn
->fn_flags
&RTN_RTINFO
));
1011 BUG_TRAP(!(fn
->fn_flags
&RTN_TL_ROOT
));
1012 BUG_TRAP(fn
->leaf
==NULL
);
1016 if (fn
->right
) child
= fn
->right
, children
|= 1;
1017 if (fn
->left
) child
= fn
->left
, children
|= 2;
1019 if (children
== 3 || SUBTREE(fn
)
1020 #ifdef CONFIG_IPV6_SUBTREES
1021 /* Subtree root (i.e. fn) may have one child */
1022 || (children
&& fn
->fn_flags
&RTN_ROOT
)
1025 fn
->leaf
= fib6_find_prefix(fn
);
1027 if (fn
->leaf
==NULL
) {
1029 fn
->leaf
= &ip6_null_entry
;
1032 atomic_inc(&fn
->leaf
->rt6i_ref
);
1037 #ifdef CONFIG_IPV6_SUBTREES
1038 if (SUBTREE(pn
) == fn
) {
1039 BUG_TRAP(fn
->fn_flags
&RTN_ROOT
);
1043 BUG_TRAP(!(fn
->fn_flags
&RTN_ROOT
));
1045 if (pn
->right
== fn
) pn
->right
= child
;
1046 else if (pn
->left
== fn
) pn
->left
= child
;
1053 #ifdef CONFIG_IPV6_SUBTREES
1057 read_lock(&fib6_walker_lock
);
1059 if (child
== NULL
) {
1060 if (w
->root
== fn
) {
1061 w
->root
= w
->node
= NULL
;
1062 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1063 } else if (w
->node
== fn
) {
1064 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1069 if (w
->root
== fn
) {
1071 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1073 if (w
->node
== fn
) {
1076 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1077 w
->state
= w
->state
>=FWS_R
? FWS_U
: FWS_INIT
;
1079 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1080 w
->state
= w
->state
>=FWS_C
? FWS_U
: FWS_INIT
;
1085 read_unlock(&fib6_walker_lock
);
1088 if (pn
->fn_flags
&RTN_RTINFO
|| SUBTREE(pn
))
1091 rt6_release(pn
->leaf
);
1097 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1098 struct nl_info
*info
)
1100 struct fib6_walker_t
*w
;
1101 struct rt6_info
*rt
= *rtp
;
1103 RT6_TRACE("fib6_del_route\n");
1107 rt
->rt6i_node
= NULL
;
1108 rt6_stats
.fib_rt_entries
--;
1109 rt6_stats
.fib_discarded_routes
++;
1111 /* Adjust walkers */
1112 read_lock(&fib6_walker_lock
);
1114 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1115 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1116 w
->leaf
= rt
->u
.next
;
1117 if (w
->leaf
== NULL
)
1121 read_unlock(&fib6_walker_lock
);
1125 if (fn
->leaf
== NULL
&& fn
->fn_flags
&RTN_TL_ROOT
)
1126 fn
->leaf
= &ip6_null_entry
;
1128 /* If it was last route, expunge its radix tree node */
1129 if (fn
->leaf
== NULL
) {
1130 fn
->fn_flags
&= ~RTN_RTINFO
;
1131 rt6_stats
.fib_route_nodes
--;
1132 fn
= fib6_repair_tree(fn
);
1135 if (atomic_read(&rt
->rt6i_ref
) != 1) {
1136 /* This route is used as dummy address holder in some split
1137 * nodes. It is not leaked, but it still holds other resources,
1138 * which must be released in time. So, scan ascendant nodes
1139 * and replace dummy references to this route with references
1140 * to still alive ones.
1143 if (!(fn
->fn_flags
&RTN_RTINFO
) && fn
->leaf
== rt
) {
1144 fn
->leaf
= fib6_find_prefix(fn
);
1145 atomic_inc(&fn
->leaf
->rt6i_ref
);
1150 /* No more references are possible at this point. */
1151 if (atomic_read(&rt
->rt6i_ref
) != 1) BUG();
1154 inet6_rt_notify(RTM_DELROUTE
, rt
, info
);
1158 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1160 struct fib6_node
*fn
= rt
->rt6i_node
;
1161 struct rt6_info
**rtp
;
1164 if (rt
->u
.dst
.obsolete
>0) {
1169 if (fn
== NULL
|| rt
== &ip6_null_entry
)
1172 BUG_TRAP(fn
->fn_flags
&RTN_RTINFO
);
1174 if (!(rt
->rt6i_flags
&RTF_CACHE
))
1175 fib6_prune_clones(fn
, rt
);
1178 * Walk the leaf entries looking for ourself
1181 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->u
.next
) {
1183 fib6_del_route(fn
, rtp
, info
);
1191 * Tree traversal function.
1193 * Certainly, it is not interrupt safe.
1194 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1195 * It means, that we can modify tree during walking
1196 * and use this function for garbage collection, clone pruning,
1197 * cleaning tree when a device goes down etc. etc.
1199 * It guarantees that every node will be traversed,
1200 * and that it will be traversed only once.
1202 * Callback function w->func may return:
1203 * 0 -> continue walking.
1204 * positive value -> walking is suspended (used by tree dumps,
1205 * and probably by gc, if it will be split to several slices)
1206 * negative value -> terminate walking.
1208 * The function itself returns:
1209 * 0 -> walk is complete.
1210 * >0 -> walk is incomplete (i.e. suspended)
1211 * <0 -> walk is terminated by an error.
1214 static int fib6_walk_continue(struct fib6_walker_t
*w
)
1216 struct fib6_node
*fn
, *pn
;
1223 if (w
->prune
&& fn
!= w
->root
&&
1224 fn
->fn_flags
&RTN_RTINFO
&& w
->state
< FWS_C
) {
1229 #ifdef CONFIG_IPV6_SUBTREES
1232 w
->node
= SUBTREE(fn
);
1240 w
->state
= FWS_INIT
;
1246 w
->node
= fn
->right
;
1247 w
->state
= FWS_INIT
;
1253 if (w
->leaf
&& fn
->fn_flags
&RTN_RTINFO
) {
1254 int err
= w
->func(w
);
1265 #ifdef CONFIG_IPV6_SUBTREES
1266 if (SUBTREE(pn
) == fn
) {
1267 BUG_TRAP(fn
->fn_flags
&RTN_ROOT
);
1272 if (pn
->left
== fn
) {
1276 if (pn
->right
== fn
) {
1278 w
->leaf
= w
->node
->leaf
;
1288 static int fib6_walk(struct fib6_walker_t
*w
)
1292 w
->state
= FWS_INIT
;
1295 fib6_walker_link(w
);
1296 res
= fib6_walk_continue(w
);
1298 fib6_walker_unlink(w
);
1302 static int fib6_clean_node(struct fib6_walker_t
*w
)
1305 struct rt6_info
*rt
;
1306 struct fib6_cleaner_t
*c
= (struct fib6_cleaner_t
*)w
;
1308 for (rt
= w
->leaf
; rt
; rt
= rt
->u
.next
) {
1309 res
= c
->func(rt
, c
->arg
);
1312 res
= fib6_del(rt
, NULL
);
1315 printk(KERN_DEBUG
"fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt
, rt
->rt6i_node
, res
);
1328 * Convenient frontend to tree walker.
1330 * func is called on each route.
1331 * It may return -1 -> delete this route.
1332 * 0 -> continue walking
1334 * prune==1 -> only immediate children of node (certainly,
1335 * ignoring pure split nodes) will be scanned.
1338 static void fib6_clean_tree(struct fib6_node
*root
,
1339 int (*func
)(struct rt6_info
*, void *arg
),
1340 int prune
, void *arg
)
1342 struct fib6_cleaner_t c
;
1345 c
.w
.func
= fib6_clean_node
;
1353 void fib6_clean_all(int (*func
)(struct rt6_info
*, void *arg
),
1354 int prune
, void *arg
)
1356 struct fib6_table
*table
;
1357 struct hlist_node
*node
;
1361 for (h
= 0; h
< FIB_TABLE_HASHSZ
; h
++) {
1362 hlist_for_each_entry_rcu(table
, node
, &fib_table_hash
[h
],
1364 write_lock_bh(&table
->tb6_lock
);
1365 fib6_clean_tree(&table
->tb6_root
, func
, prune
, arg
);
1366 write_unlock_bh(&table
->tb6_lock
);
1372 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1374 if (rt
->rt6i_flags
& RTF_CACHE
) {
1375 RT6_TRACE("pruning clone %p\n", rt
);
1382 static void fib6_prune_clones(struct fib6_node
*fn
, struct rt6_info
*rt
)
1384 fib6_clean_tree(fn
, fib6_prune_clone
, 1, rt
);
1388 * Garbage collection
1391 static struct fib6_gc_args
1397 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1399 unsigned long now
= jiffies
;
1402 * check addrconf expiration here.
1403 * Routes are expired even if they are in use.
1405 * Also age clones. Note, that clones are aged out
1406 * only if they are not in use now.
1409 if (rt
->rt6i_flags
&RTF_EXPIRES
&& rt
->rt6i_expires
) {
1410 if (time_after(now
, rt
->rt6i_expires
)) {
1411 RT6_TRACE("expiring %p\n", rt
);
1415 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1416 if (atomic_read(&rt
->u
.dst
.__refcnt
) == 0 &&
1417 time_after_eq(now
, rt
->u
.dst
.lastuse
+ gc_args
.timeout
)) {
1418 RT6_TRACE("aging clone %p\n", rt
);
1420 } else if ((rt
->rt6i_flags
& RTF_GATEWAY
) &&
1421 (!(rt
->rt6i_nexthop
->flags
& NTF_ROUTER
))) {
1422 RT6_TRACE("purging route %p via non-router but gateway\n",
1432 static DEFINE_SPINLOCK(fib6_gc_lock
);
1434 void fib6_run_gc(unsigned long dummy
)
1436 if (dummy
!= ~0UL) {
1437 spin_lock_bh(&fib6_gc_lock
);
1438 gc_args
.timeout
= dummy
? (int)dummy
: ip6_rt_gc_interval
;
1441 if (!spin_trylock(&fib6_gc_lock
)) {
1442 mod_timer(&ip6_fib_timer
, jiffies
+ HZ
);
1446 gc_args
.timeout
= ip6_rt_gc_interval
;
1450 ndisc_dst_gc(&gc_args
.more
);
1451 fib6_clean_all(fib6_age
, 0, NULL
);
1454 mod_timer(&ip6_fib_timer
, jiffies
+ ip6_rt_gc_interval
);
1456 del_timer(&ip6_fib_timer
);
1457 ip6_fib_timer
.expires
= 0;
1459 spin_unlock_bh(&fib6_gc_lock
);
1462 void __init
fib6_init(void)
1464 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1465 sizeof(struct fib6_node
),
1466 0, SLAB_HWCACHE_ALIGN
,
1468 if (!fib6_node_kmem
)
1469 panic("cannot create fib6_nodes cache");
1474 void fib6_gc_cleanup(void)
1476 del_timer(&ip6_fib_timer
);
1477 kmem_cache_destroy(fib6_node_kmem
);