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>
29 #include <linux/slab.h>
32 #include <net/ndisc.h>
33 #include <net/addrconf.h>
35 #include <net/ip6_fib.h>
36 #include <net/ip6_route.h>
41 #define RT6_TRACE(x...) printk(KERN_DEBUG x)
43 #define RT6_TRACE(x...) do { ; } while (0)
46 static struct kmem_cache
* fib6_node_kmem __read_mostly
;
50 #ifdef CONFIG_IPV6_SUBTREES
61 struct fib6_walker_t w
;
63 int (*func
)(struct rt6_info
*, void *arg
);
67 static DEFINE_RWLOCK(fib6_walker_lock
);
69 #ifdef CONFIG_IPV6_SUBTREES
70 #define FWS_INIT FWS_S
72 #define FWS_INIT FWS_L
75 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
,
77 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
);
78 static struct fib6_node
*fib6_repair_tree(struct net
*net
, struct fib6_node
*fn
);
79 static int fib6_walk(struct fib6_walker_t
*w
);
80 static int fib6_walk_continue(struct fib6_walker_t
*w
);
83 * A routing update causes an increase of the serial number on the
84 * affected subtree. This allows for cached routes to be asynchronously
85 * tested when modifications are made to the destination cache as a
86 * result of redirects, path MTU changes, etc.
89 static __u32 rt_sernum
;
91 static void fib6_gc_timer_cb(unsigned long arg
);
93 static LIST_HEAD(fib6_walkers
);
94 #define FOR_WALKERS(w) list_for_each_entry(w, &fib6_walkers, lh)
96 static inline void fib6_walker_link(struct fib6_walker_t
*w
)
98 write_lock_bh(&fib6_walker_lock
);
99 list_add(&w
->lh
, &fib6_walkers
);
100 write_unlock_bh(&fib6_walker_lock
);
103 static inline void fib6_walker_unlink(struct fib6_walker_t
*w
)
105 write_lock_bh(&fib6_walker_lock
);
107 write_unlock_bh(&fib6_walker_lock
);
109 static __inline__ u32
fib6_new_sernum(void)
118 * Auxiliary address test functions for the radix tree.
120 * These assume a 32bit processor (although it will work on
127 #if defined(__LITTLE_ENDIAN)
128 # define BITOP_BE32_SWIZZLE (0x1F & ~7)
130 # define BITOP_BE32_SWIZZLE 0
133 static __inline__ __be32
addr_bit_set(const void *token
, int fn_bit
)
135 const __be32
*addr
= token
;
138 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
139 * is optimized version of
140 * htonl(1 << ((~fn_bit)&0x1F))
141 * See include/asm-generic/bitops/le.h.
143 return (__force __be32
)(1 << ((~fn_bit
^ BITOP_BE32_SWIZZLE
) & 0x1f)) &
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
))
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 __net_init
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 flowi6
*fl6
,
260 int flags
, pol_lookup_t lookup
)
262 return (struct dst_entry
*) lookup(net
, net
->ipv6
.fib6_main_tbl
, fl6
, flags
);
265 static void __net_init
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
->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) {
325 read_lock_bh(&table
->tb6_lock
);
327 read_unlock_bh(&table
->tb6_lock
);
330 cb
->args
[5] = w
->root
->fn_sernum
;
333 if (cb
->args
[5] != w
->root
->fn_sernum
) {
334 /* Begin at the root if the tree changed */
335 cb
->args
[5] = w
->root
->fn_sernum
;
342 read_lock_bh(&table
->tb6_lock
);
343 res
= fib6_walk_continue(w
);
344 read_unlock_bh(&table
->tb6_lock
);
346 fib6_walker_unlink(w
);
354 static int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
356 struct net
*net
= sock_net(skb
->sk
);
358 unsigned int e
= 0, s_e
;
359 struct rt6_rtnl_dump_arg arg
;
360 struct fib6_walker_t
*w
;
361 struct fib6_table
*tb
;
362 struct hlist_node
*node
;
363 struct hlist_head
*head
;
369 w
= (void *)cb
->args
[2];
373 * 1. hook callback destructor.
375 cb
->args
[3] = (long)cb
->done
;
376 cb
->done
= fib6_dump_done
;
379 * 2. allocate and initialize walker.
381 w
= kzalloc(sizeof(*w
), GFP_ATOMIC
);
384 w
->func
= fib6_dump_node
;
385 cb
->args
[2] = (long)w
;
394 for (h
= s_h
; h
< FIB6_TABLE_HASHSZ
; h
++, s_e
= 0) {
396 head
= &net
->ipv6
.fib_table_hash
[h
];
397 hlist_for_each_entry_rcu(tb
, node
, head
, tb6_hlist
) {
400 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 nl_info
*info
)
615 struct rt6_info
*iter
= NULL
;
616 struct rt6_info
**ins
;
620 for (iter
= fn
->leaf
; iter
; iter
=iter
->dst
.rt6_next
) {
622 * Search for duplicates
625 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
627 * Same priority level
630 if (iter
->rt6i_dev
== rt
->rt6i_dev
&&
631 iter
->rt6i_idev
== rt
->rt6i_idev
&&
632 ipv6_addr_equal(&iter
->rt6i_gateway
,
633 &rt
->rt6i_gateway
)) {
634 if (!(iter
->rt6i_flags
&RTF_EXPIRES
))
636 iter
->rt6i_expires
= rt
->rt6i_expires
;
637 if (!(rt
->rt6i_flags
&RTF_EXPIRES
)) {
638 iter
->rt6i_flags
&= ~RTF_EXPIRES
;
639 iter
->rt6i_expires
= 0;
645 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
648 ins
= &iter
->dst
.rt6_next
;
651 /* Reset round-robin state, if necessary */
652 if (ins
== &fn
->leaf
)
659 rt
->dst
.rt6_next
= iter
;
662 atomic_inc(&rt
->rt6i_ref
);
663 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
);
664 info
->nl_net
->ipv6
.rt6_stats
->fib_rt_entries
++;
666 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
667 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
668 fn
->fn_flags
|= RTN_RTINFO
;
674 static __inline__
void fib6_start_gc(struct net
*net
, struct rt6_info
*rt
)
676 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
) &&
677 (rt
->rt6i_flags
& (RTF_EXPIRES
|RTF_CACHE
)))
678 mod_timer(&net
->ipv6
.ip6_fib_timer
,
679 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
682 void fib6_force_start_gc(struct net
*net
)
684 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
))
685 mod_timer(&net
->ipv6
.ip6_fib_timer
,
686 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
690 * Add routing information to the routing tree.
691 * <destination addr>/<source addr>
692 * with source addr info in sub-trees
695 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
, struct nl_info
*info
)
697 struct fib6_node
*fn
, *pn
= NULL
;
700 fn
= fib6_add_1(root
, &rt
->rt6i_dst
.addr
, sizeof(struct in6_addr
),
701 rt
->rt6i_dst
.plen
, offsetof(struct rt6_info
, rt6i_dst
));
708 #ifdef CONFIG_IPV6_SUBTREES
709 if (rt
->rt6i_src
.plen
) {
710 struct fib6_node
*sn
;
712 if (fn
->subtree
== NULL
) {
713 struct fib6_node
*sfn
;
725 /* Create subtree root node */
730 sfn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
731 atomic_inc(&info
->nl_net
->ipv6
.ip6_null_entry
->rt6i_ref
);
732 sfn
->fn_flags
= RTN_ROOT
;
733 sfn
->fn_sernum
= fib6_new_sernum();
735 /* Now add the first leaf node to new subtree */
737 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
738 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
739 offsetof(struct rt6_info
, rt6i_src
));
742 /* If it is failed, discard just allocated
743 root, and then (in st_failure) stale node
750 /* Now link new subtree to main tree */
754 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
755 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
756 offsetof(struct rt6_info
, rt6i_src
));
762 if (fn
->leaf
== NULL
) {
764 atomic_inc(&rt
->rt6i_ref
);
770 err
= fib6_add_rt2node(fn
, rt
, info
);
773 fib6_start_gc(info
->nl_net
, rt
);
774 if (!(rt
->rt6i_flags
&RTF_CACHE
))
775 fib6_prune_clones(info
->nl_net
, pn
, rt
);
780 #ifdef CONFIG_IPV6_SUBTREES
782 * If fib6_add_1 has cleared the old leaf pointer in the
783 * super-tree leaf node we have to find a new one for it.
785 if (pn
!= fn
&& pn
->leaf
== rt
) {
787 atomic_dec(&rt
->rt6i_ref
);
789 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
790 pn
->leaf
= fib6_find_prefix(info
->nl_net
, pn
);
793 WARN_ON(pn
->leaf
== NULL
);
794 pn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
797 atomic_inc(&pn
->leaf
->rt6i_ref
);
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(info
->nl_net
, fn
);
817 * Routing tree lookup
822 int offset
; /* key offset on rt6_info */
823 const 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 (FIB6_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
, const struct in6_addr
*daddr
,
883 const 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
, daddr
? args
: args
+ 1);
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 const 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 const struct in6_addr
*daddr
, int dst_len
,
948 const 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 WARN_ON(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 net
*net
, struct fib6_node
*fn
)
978 if (fn
->fn_flags
&RTN_ROOT
)
979 return net
->ipv6
.ip6_null_entry
;
983 return fn
->left
->leaf
;
986 return fn
->right
->leaf
;
988 fn
= FIB6_SUBTREE(fn
);
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 net
*net
,
999 struct fib6_node
*fn
)
1003 struct fib6_node
*child
, *pn
;
1004 struct fib6_walker_t
*w
;
1008 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
1011 WARN_ON(fn
->fn_flags
& RTN_RTINFO
);
1012 WARN_ON(fn
->fn_flags
& RTN_TL_ROOT
);
1013 WARN_ON(fn
->leaf
!= NULL
);
1017 if (fn
->right
) child
= fn
->right
, children
|= 1;
1018 if (fn
->left
) child
= fn
->left
, children
|= 2;
1020 if (children
== 3 || FIB6_SUBTREE(fn
)
1021 #ifdef CONFIG_IPV6_SUBTREES
1022 /* Subtree root (i.e. fn) may have one child */
1023 || (children
&& fn
->fn_flags
&RTN_ROOT
)
1026 fn
->leaf
= fib6_find_prefix(net
, fn
);
1028 if (fn
->leaf
==NULL
) {
1030 fn
->leaf
= net
->ipv6
.ip6_null_entry
;
1033 atomic_inc(&fn
->leaf
->rt6i_ref
);
1038 #ifdef CONFIG_IPV6_SUBTREES
1039 if (FIB6_SUBTREE(pn
) == fn
) {
1040 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1041 FIB6_SUBTREE(pn
) = NULL
;
1044 WARN_ON(fn
->fn_flags
& RTN_ROOT
);
1046 if (pn
->right
== fn
) pn
->right
= child
;
1047 else if (pn
->left
== fn
) pn
->left
= child
;
1055 #ifdef CONFIG_IPV6_SUBTREES
1059 read_lock(&fib6_walker_lock
);
1061 if (child
== NULL
) {
1062 if (w
->root
== fn
) {
1063 w
->root
= w
->node
= NULL
;
1064 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1065 } else if (w
->node
== fn
) {
1066 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1071 if (w
->root
== fn
) {
1073 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1075 if (w
->node
== fn
) {
1078 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1079 w
->state
= w
->state
>=FWS_R
? FWS_U
: FWS_INIT
;
1081 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1082 w
->state
= w
->state
>=FWS_C
? FWS_U
: FWS_INIT
;
1087 read_unlock(&fib6_walker_lock
);
1090 if (pn
->fn_flags
&RTN_RTINFO
|| FIB6_SUBTREE(pn
))
1093 rt6_release(pn
->leaf
);
1099 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1100 struct nl_info
*info
)
1102 struct fib6_walker_t
*w
;
1103 struct rt6_info
*rt
= *rtp
;
1104 struct net
*net
= info
->nl_net
;
1106 RT6_TRACE("fib6_del_route\n");
1109 *rtp
= rt
->dst
.rt6_next
;
1110 rt
->rt6i_node
= NULL
;
1111 net
->ipv6
.rt6_stats
->fib_rt_entries
--;
1112 net
->ipv6
.rt6_stats
->fib_discarded_routes
++;
1114 /* Reset round-robin state, if necessary */
1115 if (fn
->rr_ptr
== rt
)
1118 /* Adjust walkers */
1119 read_lock(&fib6_walker_lock
);
1121 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1122 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1123 w
->leaf
= rt
->dst
.rt6_next
;
1124 if (w
->leaf
== NULL
)
1128 read_unlock(&fib6_walker_lock
);
1130 rt
->dst
.rt6_next
= NULL
;
1132 /* If it was last route, expunge its radix tree node */
1133 if (fn
->leaf
== NULL
) {
1134 fn
->fn_flags
&= ~RTN_RTINFO
;
1135 net
->ipv6
.rt6_stats
->fib_route_nodes
--;
1136 fn
= fib6_repair_tree(net
, fn
);
1139 if (atomic_read(&rt
->rt6i_ref
) != 1) {
1140 /* This route is used as dummy address holder in some split
1141 * nodes. It is not leaked, but it still holds other resources,
1142 * which must be released in time. So, scan ascendant nodes
1143 * and replace dummy references to this route with references
1144 * to still alive ones.
1147 if (!(fn
->fn_flags
&RTN_RTINFO
) && fn
->leaf
== rt
) {
1148 fn
->leaf
= fib6_find_prefix(net
, fn
);
1149 atomic_inc(&fn
->leaf
->rt6i_ref
);
1154 /* No more references are possible at this point. */
1155 BUG_ON(atomic_read(&rt
->rt6i_ref
) != 1);
1158 inet6_rt_notify(RTM_DELROUTE
, rt
, info
);
1162 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1164 struct net
*net
= info
->nl_net
;
1165 struct fib6_node
*fn
= rt
->rt6i_node
;
1166 struct rt6_info
**rtp
;
1169 if (rt
->dst
.obsolete
>0) {
1170 WARN_ON(fn
!= NULL
);
1174 if (fn
== NULL
|| rt
== net
->ipv6
.ip6_null_entry
)
1177 WARN_ON(!(fn
->fn_flags
& RTN_RTINFO
));
1179 if (!(rt
->rt6i_flags
&RTF_CACHE
)) {
1180 struct fib6_node
*pn
= fn
;
1181 #ifdef CONFIG_IPV6_SUBTREES
1182 /* clones of this route might be in another subtree */
1183 if (rt
->rt6i_src
.plen
) {
1184 while (!(pn
->fn_flags
&RTN_ROOT
))
1189 fib6_prune_clones(info
->nl_net
, pn
, rt
);
1193 * Walk the leaf entries looking for ourself
1196 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->dst
.rt6_next
) {
1198 fib6_del_route(fn
, rtp
, info
);
1206 * Tree traversal function.
1208 * Certainly, it is not interrupt safe.
1209 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1210 * It means, that we can modify tree during walking
1211 * and use this function for garbage collection, clone pruning,
1212 * cleaning tree when a device goes down etc. etc.
1214 * It guarantees that every node will be traversed,
1215 * and that it will be traversed only once.
1217 * Callback function w->func may return:
1218 * 0 -> continue walking.
1219 * positive value -> walking is suspended (used by tree dumps,
1220 * and probably by gc, if it will be split to several slices)
1221 * negative value -> terminate walking.
1223 * The function itself returns:
1224 * 0 -> walk is complete.
1225 * >0 -> walk is incomplete (i.e. suspended)
1226 * <0 -> walk is terminated by an error.
1229 static int fib6_walk_continue(struct fib6_walker_t
*w
)
1231 struct fib6_node
*fn
, *pn
;
1238 if (w
->prune
&& fn
!= w
->root
&&
1239 fn
->fn_flags
&RTN_RTINFO
&& w
->state
< FWS_C
) {
1244 #ifdef CONFIG_IPV6_SUBTREES
1246 if (FIB6_SUBTREE(fn
)) {
1247 w
->node
= FIB6_SUBTREE(fn
);
1255 w
->state
= FWS_INIT
;
1261 w
->node
= fn
->right
;
1262 w
->state
= FWS_INIT
;
1268 if (w
->leaf
&& fn
->fn_flags
&RTN_RTINFO
) {
1271 if (w
->count
< w
->skip
) {
1289 #ifdef CONFIG_IPV6_SUBTREES
1290 if (FIB6_SUBTREE(pn
) == fn
) {
1291 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1296 if (pn
->left
== fn
) {
1300 if (pn
->right
== fn
) {
1302 w
->leaf
= w
->node
->leaf
;
1312 static int fib6_walk(struct fib6_walker_t
*w
)
1316 w
->state
= FWS_INIT
;
1319 fib6_walker_link(w
);
1320 res
= fib6_walk_continue(w
);
1322 fib6_walker_unlink(w
);
1326 static int fib6_clean_node(struct fib6_walker_t
*w
)
1329 struct rt6_info
*rt
;
1330 struct fib6_cleaner_t
*c
= container_of(w
, struct fib6_cleaner_t
, w
);
1331 struct nl_info info
= {
1335 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1336 res
= c
->func(rt
, c
->arg
);
1339 res
= fib6_del(rt
, &info
);
1342 printk(KERN_DEBUG
"fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt
, rt
->rt6i_node
, res
);
1355 * Convenient frontend to tree walker.
1357 * func is called on each route.
1358 * It may return -1 -> delete this route.
1359 * 0 -> continue walking
1361 * prune==1 -> only immediate children of node (certainly,
1362 * ignoring pure split nodes) will be scanned.
1365 static void fib6_clean_tree(struct net
*net
, struct fib6_node
*root
,
1366 int (*func
)(struct rt6_info
*, void *arg
),
1367 int prune
, void *arg
)
1369 struct fib6_cleaner_t c
;
1372 c
.w
.func
= fib6_clean_node
;
1383 void fib6_clean_all(struct net
*net
, int (*func
)(struct rt6_info
*, void *arg
),
1384 int prune
, void *arg
)
1386 struct fib6_table
*table
;
1387 struct hlist_node
*node
;
1388 struct hlist_head
*head
;
1392 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1393 head
= &net
->ipv6
.fib_table_hash
[h
];
1394 hlist_for_each_entry_rcu(table
, node
, head
, tb6_hlist
) {
1395 write_lock_bh(&table
->tb6_lock
);
1396 fib6_clean_tree(net
, &table
->tb6_root
,
1398 write_unlock_bh(&table
->tb6_lock
);
1404 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1406 if (rt
->rt6i_flags
& RTF_CACHE
) {
1407 RT6_TRACE("pruning clone %p\n", rt
);
1414 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
,
1415 struct rt6_info
*rt
)
1417 fib6_clean_tree(net
, fn
, fib6_prune_clone
, 1, rt
);
1421 * Garbage collection
1424 static struct fib6_gc_args
1430 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1432 unsigned long now
= jiffies
;
1435 * check addrconf expiration here.
1436 * Routes are expired even if they are in use.
1438 * Also age clones. Note, that clones are aged out
1439 * only if they are not in use now.
1442 if (rt
->rt6i_flags
&RTF_EXPIRES
&& rt
->rt6i_expires
) {
1443 if (time_after(now
, rt
->rt6i_expires
)) {
1444 RT6_TRACE("expiring %p\n", rt
);
1448 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1449 if (atomic_read(&rt
->dst
.__refcnt
) == 0 &&
1450 time_after_eq(now
, rt
->dst
.lastuse
+ gc_args
.timeout
)) {
1451 RT6_TRACE("aging clone %p\n", rt
);
1453 } else if ((rt
->rt6i_flags
& RTF_GATEWAY
) &&
1454 (!(dst_get_neighbour_raw(&rt
->dst
)->flags
& NTF_ROUTER
))) {
1455 RT6_TRACE("purging route %p via non-router but gateway\n",
1465 static DEFINE_SPINLOCK(fib6_gc_lock
);
1467 void fib6_run_gc(unsigned long expires
, struct net
*net
)
1469 if (expires
!= ~0UL) {
1470 spin_lock_bh(&fib6_gc_lock
);
1471 gc_args
.timeout
= expires
? (int)expires
:
1472 net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1474 if (!spin_trylock_bh(&fib6_gc_lock
)) {
1475 mod_timer(&net
->ipv6
.ip6_fib_timer
, jiffies
+ HZ
);
1478 gc_args
.timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1481 gc_args
.more
= icmp6_dst_gc();
1483 fib6_clean_all(net
, fib6_age
, 0, NULL
);
1486 mod_timer(&net
->ipv6
.ip6_fib_timer
,
1487 round_jiffies(jiffies
1488 + net
->ipv6
.sysctl
.ip6_rt_gc_interval
));
1490 del_timer(&net
->ipv6
.ip6_fib_timer
);
1491 spin_unlock_bh(&fib6_gc_lock
);
1494 static void fib6_gc_timer_cb(unsigned long arg
)
1496 fib6_run_gc(0, (struct net
*)arg
);
1499 static int __net_init
fib6_net_init(struct net
*net
)
1501 size_t size
= sizeof(struct hlist_head
) * FIB6_TABLE_HASHSZ
;
1503 setup_timer(&net
->ipv6
.ip6_fib_timer
, fib6_gc_timer_cb
, (unsigned long)net
);
1505 net
->ipv6
.rt6_stats
= kzalloc(sizeof(*net
->ipv6
.rt6_stats
), GFP_KERNEL
);
1506 if (!net
->ipv6
.rt6_stats
)
1509 /* Avoid false sharing : Use at least a full cache line */
1510 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1512 net
->ipv6
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1513 if (!net
->ipv6
.fib_table_hash
)
1516 net
->ipv6
.fib6_main_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_main_tbl
),
1518 if (!net
->ipv6
.fib6_main_tbl
)
1519 goto out_fib_table_hash
;
1521 net
->ipv6
.fib6_main_tbl
->tb6_id
= RT6_TABLE_MAIN
;
1522 net
->ipv6
.fib6_main_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1523 net
->ipv6
.fib6_main_tbl
->tb6_root
.fn_flags
=
1524 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1526 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1527 net
->ipv6
.fib6_local_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_local_tbl
),
1529 if (!net
->ipv6
.fib6_local_tbl
)
1530 goto out_fib6_main_tbl
;
1531 net
->ipv6
.fib6_local_tbl
->tb6_id
= RT6_TABLE_LOCAL
;
1532 net
->ipv6
.fib6_local_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1533 net
->ipv6
.fib6_local_tbl
->tb6_root
.fn_flags
=
1534 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1536 fib6_tables_init(net
);
1540 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1542 kfree(net
->ipv6
.fib6_main_tbl
);
1545 kfree(net
->ipv6
.fib_table_hash
);
1547 kfree(net
->ipv6
.rt6_stats
);
1552 static void fib6_net_exit(struct net
*net
)
1554 rt6_ifdown(net
, NULL
);
1555 del_timer_sync(&net
->ipv6
.ip6_fib_timer
);
1557 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1558 kfree(net
->ipv6
.fib6_local_tbl
);
1560 kfree(net
->ipv6
.fib6_main_tbl
);
1561 kfree(net
->ipv6
.fib_table_hash
);
1562 kfree(net
->ipv6
.rt6_stats
);
1565 static struct pernet_operations fib6_net_ops
= {
1566 .init
= fib6_net_init
,
1567 .exit
= fib6_net_exit
,
1570 int __init
fib6_init(void)
1574 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1575 sizeof(struct fib6_node
),
1576 0, SLAB_HWCACHE_ALIGN
,
1578 if (!fib6_node_kmem
)
1581 ret
= register_pernet_subsys(&fib6_net_ops
);
1583 goto out_kmem_cache_create
;
1585 ret
= __rtnl_register(PF_INET6
, RTM_GETROUTE
, NULL
, inet6_dump_fib
,
1588 goto out_unregister_subsys
;
1592 out_unregister_subsys
:
1593 unregister_pernet_subsys(&fib6_net_ops
);
1594 out_kmem_cache_create
:
1595 kmem_cache_destroy(fib6_node_kmem
);
1599 void fib6_gc_cleanup(void)
1601 unregister_pernet_subsys(&fib6_net_ops
);
1602 kmem_cache_destroy(fib6_node_kmem
);