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 fib6_node
* fib6_repair_tree(struct fib6_node
*fn
);
84 static int fib6_walk(struct fib6_walker_t
*w
);
85 static int fib6_walk_continue(struct fib6_walker_t
*w
);
88 * A routing update causes an increase of the serial number on the
89 * affected subtree. This allows for cached routes to be asynchronously
90 * tested when modifications are made to the destination cache as a
91 * result of redirects, path MTU changes, etc.
94 static __u32 rt_sernum
;
96 static DEFINE_TIMER(ip6_fib_timer
, fib6_run_gc
, 0, 0);
98 static struct fib6_walker_t fib6_walker_list
= {
99 .prev
= &fib6_walker_list
,
100 .next
= &fib6_walker_list
,
103 #define FOR_WALKERS(w) for ((w)=fib6_walker_list.next; (w) != &fib6_walker_list; (w)=(w)->next)
105 static inline void fib6_walker_link(struct fib6_walker_t
*w
)
107 write_lock_bh(&fib6_walker_lock
);
108 w
->next
= fib6_walker_list
.next
;
109 w
->prev
= &fib6_walker_list
;
112 write_unlock_bh(&fib6_walker_lock
);
115 static inline void fib6_walker_unlink(struct fib6_walker_t
*w
)
117 write_lock_bh(&fib6_walker_lock
);
118 w
->next
->prev
= w
->prev
;
119 w
->prev
->next
= w
->next
;
120 w
->prev
= w
->next
= w
;
121 write_unlock_bh(&fib6_walker_lock
);
123 static __inline__ u32
fib6_new_sernum(void)
132 * Auxiliary address test functions for the radix tree.
134 * These assume a 32bit processor (although it will work on
142 static __inline__
int addr_bit_set(void *token
, int fn_bit
)
146 return htonl(1 << ((~fn_bit
)&0x1F)) & addr
[fn_bit
>>5];
149 static __inline__
struct fib6_node
* node_alloc(void)
151 struct fib6_node
*fn
;
153 if ((fn
= kmem_cache_alloc(fib6_node_kmem
, SLAB_ATOMIC
)) != NULL
)
154 memset(fn
, 0, sizeof(struct fib6_node
));
159 static __inline__
void node_free(struct fib6_node
* fn
)
161 kmem_cache_free(fib6_node_kmem
, fn
);
164 static __inline__
void rt6_release(struct rt6_info
*rt
)
166 if (atomic_dec_and_test(&rt
->rt6i_ref
))
167 dst_free(&rt
->u
.dst
);
170 static struct fib6_table fib6_main_tbl
= {
171 .tb6_id
= RT6_TABLE_MAIN
,
172 .tb6_lock
= RW_LOCK_UNLOCKED
,
174 .leaf
= &ip6_null_entry
,
175 .fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
,
179 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
180 #define FIB_TABLE_HASHSZ 256
182 #define FIB_TABLE_HASHSZ 1
184 static struct hlist_head fib_table_hash
[FIB_TABLE_HASHSZ
];
186 static void fib6_link_table(struct fib6_table
*tb
)
190 h
= tb
->tb6_id
& (FIB_TABLE_HASHSZ
- 1);
193 * No protection necessary, this is the only list mutatation
194 * operation, tables never disappear once they exist.
196 hlist_add_head_rcu(&tb
->tb6_hlist
, &fib_table_hash
[h
]);
199 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
200 static struct fib6_table fib6_local_tbl
= {
201 .tb6_id
= RT6_TABLE_LOCAL
,
202 .tb6_lock
= RW_LOCK_UNLOCKED
,
204 .leaf
= &ip6_null_entry
,
205 .fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
,
209 static struct fib6_table
*fib6_alloc_table(u32 id
)
211 struct fib6_table
*table
;
213 table
= kzalloc(sizeof(*table
), GFP_ATOMIC
);
216 table
->tb6_lock
= RW_LOCK_UNLOCKED
;
217 table
->tb6_root
.leaf
= &ip6_null_entry
;
218 table
->tb6_root
.fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
224 struct fib6_table
*fib6_new_table(u32 id
)
226 struct fib6_table
*tb
;
230 tb
= fib6_get_table(id
);
234 tb
= fib6_alloc_table(id
);
241 struct fib6_table
*fib6_get_table(u32 id
)
243 struct fib6_table
*tb
;
244 struct hlist_node
*node
;
249 h
= id
& (FIB_TABLE_HASHSZ
- 1);
251 hlist_for_each_entry_rcu(tb
, node
, &fib_table_hash
[h
], tb6_hlist
) {
252 if (tb
->tb6_id
== id
) {
262 static void __init
fib6_tables_init(void)
264 fib6_link_table(&fib6_main_tbl
);
265 fib6_link_table(&fib6_local_tbl
);
270 struct fib6_table
*fib6_new_table(u32 id
)
272 return fib6_get_table(id
);
275 struct fib6_table
*fib6_get_table(u32 id
)
277 return &fib6_main_tbl
;
280 struct dst_entry
*fib6_rule_lookup(struct flowi
*fl
, int flags
,
283 return (struct dst_entry
*) lookup(&fib6_main_tbl
, fl
, flags
);
286 static void __init
fib6_tables_init(void)
288 fib6_link_table(&fib6_main_tbl
);
293 static int fib6_dump_node(struct fib6_walker_t
*w
)
298 for (rt
= w
->leaf
; rt
; rt
= rt
->u
.next
) {
299 res
= rt6_dump_route(rt
, w
->args
);
301 /* Frame is full, suspend walking */
311 static void fib6_dump_end(struct netlink_callback
*cb
)
313 struct fib6_walker_t
*w
= (void*)cb
->args
[2];
319 cb
->done
= (void*)cb
->args
[3];
323 static int fib6_dump_done(struct netlink_callback
*cb
)
326 return cb
->done
? cb
->done(cb
) : 0;
329 static int fib6_dump_table(struct fib6_table
*table
, struct sk_buff
*skb
,
330 struct netlink_callback
*cb
)
332 struct fib6_walker_t
*w
;
335 w
= (void *)cb
->args
[2];
336 w
->root
= &table
->tb6_root
;
338 if (cb
->args
[4] == 0) {
339 read_lock_bh(&table
->tb6_lock
);
341 read_unlock_bh(&table
->tb6_lock
);
345 read_lock_bh(&table
->tb6_lock
);
346 res
= fib6_walk_continue(w
);
347 read_unlock_bh(&table
->tb6_lock
);
350 fib6_walker_unlink(w
);
353 fib6_walker_unlink(w
);
360 int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
363 unsigned int e
= 0, s_e
;
364 struct rt6_rtnl_dump_arg arg
;
365 struct fib6_walker_t
*w
;
366 struct fib6_table
*tb
;
367 struct hlist_node
*node
;
373 w
= (void *)cb
->args
[2];
377 * 1. hook callback destructor.
379 cb
->args
[3] = (long)cb
->done
;
380 cb
->done
= fib6_dump_done
;
383 * 2. allocate and initialize walker.
385 w
= kzalloc(sizeof(*w
), GFP_ATOMIC
);
388 w
->func
= fib6_dump_node
;
389 cb
->args
[2] = (long)w
;
396 for (h
= s_h
; h
< FIB_TABLE_HASHSZ
; h
++, s_e
= 0) {
398 hlist_for_each_entry(tb
, node
, &fib_table_hash
[h
], tb6_hlist
) {
401 res
= fib6_dump_table(tb
, skb
, cb
);
412 res
= res
< 0 ? res
: skb
->len
;
421 * return the appropriate node for a routing tree "add" operation
422 * by either creating and inserting or by returning an existing
426 static struct fib6_node
* fib6_add_1(struct fib6_node
*root
, void *addr
,
427 int addrlen
, int plen
,
430 struct fib6_node
*fn
, *in
, *ln
;
431 struct fib6_node
*pn
= NULL
;
435 __u32 sernum
= fib6_new_sernum();
437 RT6_TRACE("fib6_add_1\n");
439 /* insert node in tree */
444 key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
449 if (plen
< fn
->fn_bit
||
450 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
457 if (plen
== fn
->fn_bit
) {
458 /* clean up an intermediate node */
459 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
460 rt6_release(fn
->leaf
);
464 fn
->fn_sernum
= sernum
;
470 * We have more bits to go
473 /* Try to walk down on tree. */
474 fn
->fn_sernum
= sernum
;
475 dir
= addr_bit_set(addr
, fn
->fn_bit
);
477 fn
= dir
? fn
->right
: fn
->left
;
481 * We walked to the bottom of tree.
482 * Create new leaf node without children.
492 ln
->fn_sernum
= sernum
;
504 * split since we don't have a common prefix anymore or
505 * we have a less significant route.
506 * we've to insert an intermediate node on the list
507 * this new node will point to the one we need to create
513 /* find 1st bit in difference between the 2 addrs.
515 See comment in __ipv6_addr_diff: bit may be an invalid value,
516 but if it is >= plen, the value is ignored in any case.
519 bit
= __ipv6_addr_diff(addr
, &key
->addr
, addrlen
);
524 * (new leaf node)[ln] (old node)[fn]
530 if (in
== NULL
|| ln
== NULL
) {
539 * new intermediate node.
541 * be off since that an address that chooses one of
542 * the branches would not match less specific routes
543 * in the other branch
550 atomic_inc(&in
->leaf
->rt6i_ref
);
552 in
->fn_sernum
= sernum
;
554 /* update parent pointer */
565 ln
->fn_sernum
= sernum
;
567 if (addr_bit_set(addr
, bit
)) {
574 } else { /* plen <= bit */
577 * (new leaf node)[ln]
579 * (old node)[fn] NULL
591 ln
->fn_sernum
= sernum
;
598 if (addr_bit_set(&key
->addr
, plen
))
609 * Insert routing information in a node.
612 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
613 struct nlmsghdr
*nlh
, struct netlink_skb_parms
*req
)
615 struct rt6_info
*iter
= NULL
;
616 struct rt6_info
**ins
;
620 if (fn
->fn_flags
&RTN_TL_ROOT
&&
621 fn
->leaf
== &ip6_null_entry
&&
622 !(rt
->rt6i_flags
& (RTF_DEFAULT
| RTF_ADDRCONF
)) ){
628 for (iter
= fn
->leaf
; iter
; iter
=iter
->u
.next
) {
630 * Search for duplicates
633 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
635 * Same priority level
638 if (iter
->rt6i_dev
== rt
->rt6i_dev
&&
639 iter
->rt6i_idev
== rt
->rt6i_idev
&&
640 ipv6_addr_equal(&iter
->rt6i_gateway
,
641 &rt
->rt6i_gateway
)) {
642 if (!(iter
->rt6i_flags
&RTF_EXPIRES
))
644 iter
->rt6i_expires
= rt
->rt6i_expires
;
645 if (!(rt
->rt6i_flags
&RTF_EXPIRES
)) {
646 iter
->rt6i_flags
&= ~RTF_EXPIRES
;
647 iter
->rt6i_expires
= 0;
653 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
667 atomic_inc(&rt
->rt6i_ref
);
668 inet6_rt_notify(RTM_NEWROUTE
, rt
, nlh
, req
);
669 rt6_stats
.fib_rt_entries
++;
671 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
672 rt6_stats
.fib_route_nodes
++;
673 fn
->fn_flags
|= RTN_RTINFO
;
679 static __inline__
void fib6_start_gc(struct rt6_info
*rt
)
681 if (ip6_fib_timer
.expires
== 0 &&
682 (rt
->rt6i_flags
& (RTF_EXPIRES
|RTF_CACHE
)))
683 mod_timer(&ip6_fib_timer
, jiffies
+ ip6_rt_gc_interval
);
686 void fib6_force_start_gc(void)
688 if (ip6_fib_timer
.expires
== 0)
689 mod_timer(&ip6_fib_timer
, jiffies
+ ip6_rt_gc_interval
);
693 * Add routing information to the routing tree.
694 * <destination addr>/<source addr>
695 * with source addr info in sub-trees
698 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
,
699 struct nlmsghdr
*nlh
, void *_rtattr
, struct netlink_skb_parms
*req
)
701 struct fib6_node
*fn
;
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
));
710 #ifdef CONFIG_IPV6_SUBTREES
711 if (rt
->rt6i_src
.plen
) {
712 struct fib6_node
*sn
;
714 if (fn
->subtree
== NULL
) {
715 struct fib6_node
*sfn
;
727 /* Create subtree root node */
732 sfn
->leaf
= &ip6_null_entry
;
733 atomic_inc(&ip6_null_entry
.rt6i_ref
);
734 sfn
->fn_flags
= RTN_ROOT
;
735 sfn
->fn_sernum
= fib6_new_sernum();
737 /* Now add the first leaf node to new subtree */
739 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
740 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
741 offsetof(struct rt6_info
, rt6i_src
));
744 /* If it is failed, discard just allocated
745 root, and then (in st_failure) stale node
752 /* Now link new subtree to main tree */
755 if (fn
->leaf
== NULL
) {
757 atomic_inc(&rt
->rt6i_ref
);
760 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
761 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
762 offsetof(struct rt6_info
, rt6i_src
));
772 err
= fib6_add_rt2node(fn
, rt
, nlh
, req
);
776 if (!(rt
->rt6i_flags
&RTF_CACHE
))
777 fib6_prune_clones(fn
, rt
);
782 dst_free(&rt
->u
.dst
);
785 #ifdef CONFIG_IPV6_SUBTREES
786 /* Subtree creation failed, probably main tree node
787 is orphan. If it is, shoot it.
790 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
791 fib6_repair_tree(fn
);
792 dst_free(&rt
->u
.dst
);
798 * Routing tree lookup
803 int offset
; /* key offset on rt6_info */
804 struct in6_addr
*addr
; /* search key */
807 static struct fib6_node
* fib6_lookup_1(struct fib6_node
*root
,
808 struct lookup_args
*args
)
810 struct fib6_node
*fn
;
820 struct fib6_node
*next
;
822 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
824 next
= dir
? fn
->right
: fn
->left
;
834 while ((fn
->fn_flags
& RTN_ROOT
) == 0) {
835 #ifdef CONFIG_IPV6_SUBTREES
837 struct fib6_node
*st
;
838 struct lookup_args
*narg
;
843 st
= fib6_lookup_1(fn
->subtree
, narg
);
845 if (st
&& !(st
->fn_flags
& RTN_ROOT
))
851 if (fn
->fn_flags
& RTN_RTINFO
) {
854 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
857 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
))
867 struct fib6_node
* fib6_lookup(struct fib6_node
*root
, struct in6_addr
*daddr
,
868 struct in6_addr
*saddr
)
870 struct lookup_args args
[2];
871 struct fib6_node
*fn
;
873 args
[0].offset
= offsetof(struct rt6_info
, rt6i_dst
);
874 args
[0].addr
= daddr
;
876 #ifdef CONFIG_IPV6_SUBTREES
877 args
[1].offset
= offsetof(struct rt6_info
, rt6i_src
);
878 args
[1].addr
= saddr
;
881 fn
= fib6_lookup_1(root
, args
);
883 if (fn
== NULL
|| fn
->fn_flags
& RTN_TL_ROOT
)
890 * Get node with specified destination prefix (and source prefix,
891 * if subtrees are used)
895 static struct fib6_node
* fib6_locate_1(struct fib6_node
*root
,
896 struct in6_addr
*addr
,
897 int plen
, int offset
)
899 struct fib6_node
*fn
;
901 for (fn
= root
; fn
; ) {
902 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
907 if (plen
< fn
->fn_bit
||
908 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
911 if (plen
== fn
->fn_bit
)
915 * We have more bits to go
917 if (addr_bit_set(addr
, fn
->fn_bit
))
925 struct fib6_node
* fib6_locate(struct fib6_node
*root
,
926 struct in6_addr
*daddr
, int dst_len
,
927 struct in6_addr
*saddr
, int src_len
)
929 struct fib6_node
*fn
;
931 fn
= fib6_locate_1(root
, daddr
, dst_len
,
932 offsetof(struct rt6_info
, rt6i_dst
));
934 #ifdef CONFIG_IPV6_SUBTREES
936 BUG_TRAP(saddr
!=NULL
);
940 fn
= fib6_locate_1(fn
, saddr
, src_len
,
941 offsetof(struct rt6_info
, rt6i_src
));
945 if (fn
&& fn
->fn_flags
&RTN_RTINFO
)
957 static struct rt6_info
* fib6_find_prefix(struct fib6_node
*fn
)
959 if (fn
->fn_flags
&RTN_ROOT
)
960 return &ip6_null_entry
;
964 return fn
->left
->leaf
;
967 return fn
->right
->leaf
;
975 * Called to trim the tree of intermediate nodes when possible. "fn"
976 * is the node we want to try and remove.
979 static struct fib6_node
* fib6_repair_tree(struct fib6_node
*fn
)
983 struct fib6_node
*child
, *pn
;
984 struct fib6_walker_t
*w
;
988 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
991 BUG_TRAP(!(fn
->fn_flags
&RTN_RTINFO
));
992 BUG_TRAP(!(fn
->fn_flags
&RTN_TL_ROOT
));
993 BUG_TRAP(fn
->leaf
==NULL
);
997 if (fn
->right
) child
= fn
->right
, children
|= 1;
998 if (fn
->left
) child
= fn
->left
, children
|= 2;
1000 if (children
== 3 || SUBTREE(fn
)
1001 #ifdef CONFIG_IPV6_SUBTREES
1002 /* Subtree root (i.e. fn) may have one child */
1003 || (children
&& fn
->fn_flags
&RTN_ROOT
)
1006 fn
->leaf
= fib6_find_prefix(fn
);
1008 if (fn
->leaf
==NULL
) {
1010 fn
->leaf
= &ip6_null_entry
;
1013 atomic_inc(&fn
->leaf
->rt6i_ref
);
1018 #ifdef CONFIG_IPV6_SUBTREES
1019 if (SUBTREE(pn
) == fn
) {
1020 BUG_TRAP(fn
->fn_flags
&RTN_ROOT
);
1024 BUG_TRAP(!(fn
->fn_flags
&RTN_ROOT
));
1026 if (pn
->right
== fn
) pn
->right
= child
;
1027 else if (pn
->left
== fn
) pn
->left
= child
;
1034 #ifdef CONFIG_IPV6_SUBTREES
1038 read_lock(&fib6_walker_lock
);
1040 if (child
== NULL
) {
1041 if (w
->root
== fn
) {
1042 w
->root
= w
->node
= NULL
;
1043 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1044 } else if (w
->node
== fn
) {
1045 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1050 if (w
->root
== fn
) {
1052 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1054 if (w
->node
== fn
) {
1057 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1058 w
->state
= w
->state
>=FWS_R
? FWS_U
: FWS_INIT
;
1060 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1061 w
->state
= w
->state
>=FWS_C
? FWS_U
: FWS_INIT
;
1066 read_unlock(&fib6_walker_lock
);
1069 if (pn
->fn_flags
&RTN_RTINFO
|| SUBTREE(pn
))
1072 rt6_release(pn
->leaf
);
1078 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1079 struct nlmsghdr
*nlh
, void *_rtattr
, struct netlink_skb_parms
*req
)
1081 struct fib6_walker_t
*w
;
1082 struct rt6_info
*rt
= *rtp
;
1084 RT6_TRACE("fib6_del_route\n");
1088 rt
->rt6i_node
= NULL
;
1089 rt6_stats
.fib_rt_entries
--;
1090 rt6_stats
.fib_discarded_routes
++;
1092 /* Adjust walkers */
1093 read_lock(&fib6_walker_lock
);
1095 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1096 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1097 w
->leaf
= rt
->u
.next
;
1098 if (w
->leaf
== NULL
)
1102 read_unlock(&fib6_walker_lock
);
1106 if (fn
->leaf
== NULL
&& fn
->fn_flags
&RTN_TL_ROOT
)
1107 fn
->leaf
= &ip6_null_entry
;
1109 /* If it was last route, expunge its radix tree node */
1110 if (fn
->leaf
== NULL
) {
1111 fn
->fn_flags
&= ~RTN_RTINFO
;
1112 rt6_stats
.fib_route_nodes
--;
1113 fn
= fib6_repair_tree(fn
);
1116 if (atomic_read(&rt
->rt6i_ref
) != 1) {
1117 /* This route is used as dummy address holder in some split
1118 * nodes. It is not leaked, but it still holds other resources,
1119 * which must be released in time. So, scan ascendant nodes
1120 * and replace dummy references to this route with references
1121 * to still alive ones.
1124 if (!(fn
->fn_flags
&RTN_RTINFO
) && fn
->leaf
== rt
) {
1125 fn
->leaf
= fib6_find_prefix(fn
);
1126 atomic_inc(&fn
->leaf
->rt6i_ref
);
1131 /* No more references are possible at this point. */
1132 if (atomic_read(&rt
->rt6i_ref
) != 1) BUG();
1135 inet6_rt_notify(RTM_DELROUTE
, rt
, nlh
, req
);
1139 int fib6_del(struct rt6_info
*rt
, struct nlmsghdr
*nlh
, void *_rtattr
, struct netlink_skb_parms
*req
)
1141 struct fib6_node
*fn
= rt
->rt6i_node
;
1142 struct rt6_info
**rtp
;
1145 if (rt
->u
.dst
.obsolete
>0) {
1150 if (fn
== NULL
|| rt
== &ip6_null_entry
)
1153 BUG_TRAP(fn
->fn_flags
&RTN_RTINFO
);
1155 if (!(rt
->rt6i_flags
&RTF_CACHE
))
1156 fib6_prune_clones(fn
, rt
);
1159 * Walk the leaf entries looking for ourself
1162 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->u
.next
) {
1164 fib6_del_route(fn
, rtp
, nlh
, _rtattr
, req
);
1172 * Tree traversal function.
1174 * Certainly, it is not interrupt safe.
1175 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1176 * It means, that we can modify tree during walking
1177 * and use this function for garbage collection, clone pruning,
1178 * cleaning tree when a device goes down etc. etc.
1180 * It guarantees that every node will be traversed,
1181 * and that it will be traversed only once.
1183 * Callback function w->func may return:
1184 * 0 -> continue walking.
1185 * positive value -> walking is suspended (used by tree dumps,
1186 * and probably by gc, if it will be split to several slices)
1187 * negative value -> terminate walking.
1189 * The function itself returns:
1190 * 0 -> walk is complete.
1191 * >0 -> walk is incomplete (i.e. suspended)
1192 * <0 -> walk is terminated by an error.
1195 static int fib6_walk_continue(struct fib6_walker_t
*w
)
1197 struct fib6_node
*fn
, *pn
;
1204 if (w
->prune
&& fn
!= w
->root
&&
1205 fn
->fn_flags
&RTN_RTINFO
&& w
->state
< FWS_C
) {
1210 #ifdef CONFIG_IPV6_SUBTREES
1213 w
->node
= SUBTREE(fn
);
1221 w
->state
= FWS_INIT
;
1227 w
->node
= fn
->right
;
1228 w
->state
= FWS_INIT
;
1234 if (w
->leaf
&& fn
->fn_flags
&RTN_RTINFO
) {
1235 int err
= w
->func(w
);
1246 #ifdef CONFIG_IPV6_SUBTREES
1247 if (SUBTREE(pn
) == fn
) {
1248 BUG_TRAP(fn
->fn_flags
&RTN_ROOT
);
1253 if (pn
->left
== fn
) {
1257 if (pn
->right
== fn
) {
1259 w
->leaf
= w
->node
->leaf
;
1269 static int fib6_walk(struct fib6_walker_t
*w
)
1273 w
->state
= FWS_INIT
;
1276 fib6_walker_link(w
);
1277 res
= fib6_walk_continue(w
);
1279 fib6_walker_unlink(w
);
1283 static int fib6_clean_node(struct fib6_walker_t
*w
)
1286 struct rt6_info
*rt
;
1287 struct fib6_cleaner_t
*c
= (struct fib6_cleaner_t
*)w
;
1289 for (rt
= w
->leaf
; rt
; rt
= rt
->u
.next
) {
1290 res
= c
->func(rt
, c
->arg
);
1293 res
= fib6_del(rt
, NULL
, NULL
, NULL
);
1296 printk(KERN_DEBUG
"fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt
, rt
->rt6i_node
, res
);
1309 * Convenient frontend to tree walker.
1311 * func is called on each route.
1312 * It may return -1 -> delete this route.
1313 * 0 -> continue walking
1315 * prune==1 -> only immediate children of node (certainly,
1316 * ignoring pure split nodes) will be scanned.
1319 static void fib6_clean_tree(struct fib6_node
*root
,
1320 int (*func
)(struct rt6_info
*, void *arg
),
1321 int prune
, void *arg
)
1323 struct fib6_cleaner_t c
;
1326 c
.w
.func
= fib6_clean_node
;
1334 void fib6_clean_all(int (*func
)(struct rt6_info
*, void *arg
),
1335 int prune
, void *arg
)
1337 struct fib6_table
*table
;
1338 struct hlist_node
*node
;
1342 for (h
= 0; h
< FIB_TABLE_HASHSZ
; h
++) {
1343 hlist_for_each_entry_rcu(table
, node
, &fib_table_hash
[h
],
1345 write_lock_bh(&table
->tb6_lock
);
1346 fib6_clean_tree(&table
->tb6_root
, func
, prune
, arg
);
1347 write_unlock_bh(&table
->tb6_lock
);
1353 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1355 if (rt
->rt6i_flags
& RTF_CACHE
) {
1356 RT6_TRACE("pruning clone %p\n", rt
);
1363 static void fib6_prune_clones(struct fib6_node
*fn
, struct rt6_info
*rt
)
1365 fib6_clean_tree(fn
, fib6_prune_clone
, 1, rt
);
1369 * Garbage collection
1372 static struct fib6_gc_args
1378 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1380 unsigned long now
= jiffies
;
1383 * check addrconf expiration here.
1384 * Routes are expired even if they are in use.
1386 * Also age clones. Note, that clones are aged out
1387 * only if they are not in use now.
1390 if (rt
->rt6i_flags
&RTF_EXPIRES
&& rt
->rt6i_expires
) {
1391 if (time_after(now
, rt
->rt6i_expires
)) {
1392 RT6_TRACE("expiring %p\n", rt
);
1396 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1397 if (atomic_read(&rt
->u
.dst
.__refcnt
) == 0 &&
1398 time_after_eq(now
, rt
->u
.dst
.lastuse
+ gc_args
.timeout
)) {
1399 RT6_TRACE("aging clone %p\n", rt
);
1401 } else if ((rt
->rt6i_flags
& RTF_GATEWAY
) &&
1402 (!(rt
->rt6i_nexthop
->flags
& NTF_ROUTER
))) {
1403 RT6_TRACE("purging route %p via non-router but gateway\n",
1413 static DEFINE_SPINLOCK(fib6_gc_lock
);
1415 void fib6_run_gc(unsigned long dummy
)
1417 if (dummy
!= ~0UL) {
1418 spin_lock_bh(&fib6_gc_lock
);
1419 gc_args
.timeout
= dummy
? (int)dummy
: ip6_rt_gc_interval
;
1422 if (!spin_trylock(&fib6_gc_lock
)) {
1423 mod_timer(&ip6_fib_timer
, jiffies
+ HZ
);
1427 gc_args
.timeout
= ip6_rt_gc_interval
;
1431 ndisc_dst_gc(&gc_args
.more
);
1432 fib6_clean_all(fib6_age
, 0, NULL
);
1435 mod_timer(&ip6_fib_timer
, jiffies
+ ip6_rt_gc_interval
);
1437 del_timer(&ip6_fib_timer
);
1438 ip6_fib_timer
.expires
= 0;
1440 spin_unlock_bh(&fib6_gc_lock
);
1443 void __init
fib6_init(void)
1445 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1446 sizeof(struct fib6_node
),
1447 0, SLAB_HWCACHE_ALIGN
,
1449 if (!fib6_node_kmem
)
1450 panic("cannot create fib6_nodes cache");
1455 void fib6_gc_cleanup(void)
1457 del_timer(&ip6_fib_timer
);
1458 kmem_cache_destroy(fib6_node_kmem
);