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 <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 static struct kmem_cache
* fib6_node_kmem __read_mostly
;
54 #ifdef CONFIG_IPV6_SUBTREES
65 struct fib6_walker_t w
;
67 int (*func
)(struct rt6_info
*, void *arg
);
71 static DEFINE_RWLOCK(fib6_walker_lock
);
73 #ifdef CONFIG_IPV6_SUBTREES
74 #define FWS_INIT FWS_S
76 #define FWS_INIT FWS_L
79 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
,
81 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
);
82 static struct fib6_node
*fib6_repair_tree(struct net
*net
, struct fib6_node
*fn
);
83 static int fib6_walk(struct fib6_walker_t
*w
);
84 static int fib6_walk_continue(struct fib6_walker_t
*w
);
87 * A routing update causes an increase of the serial number on the
88 * affected subtree. This allows for cached routes to be asynchronously
89 * tested when modifications are made to the destination cache as a
90 * result of redirects, path MTU changes, etc.
93 static __u32 rt_sernum
;
95 static void fib6_gc_timer_cb(unsigned long arg
);
97 static LIST_HEAD(fib6_walkers
);
98 #define FOR_WALKERS(w) list_for_each_entry(w, &fib6_walkers, lh)
100 static inline void fib6_walker_link(struct fib6_walker_t
*w
)
102 write_lock_bh(&fib6_walker_lock
);
103 list_add(&w
->lh
, &fib6_walkers
);
104 write_unlock_bh(&fib6_walker_lock
);
107 static inline void fib6_walker_unlink(struct fib6_walker_t
*w
)
109 write_lock_bh(&fib6_walker_lock
);
111 write_unlock_bh(&fib6_walker_lock
);
113 static __inline__ u32
fib6_new_sernum(void)
122 * Auxiliary address test functions for the radix tree.
124 * These assume a 32bit processor (although it will work on
131 #if defined(__LITTLE_ENDIAN)
132 # define BITOP_BE32_SWIZZLE (0x1F & ~7)
134 # define BITOP_BE32_SWIZZLE 0
137 static __inline__ __be32
addr_bit_set(const void *token
, int fn_bit
)
139 const __be32
*addr
= token
;
142 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
143 * is optimized version of
144 * htonl(1 << ((~fn_bit)&0x1F))
145 * See include/asm-generic/bitops/le.h.
147 return (__force __be32
)(1 << ((~fn_bit
^ BITOP_BE32_SWIZZLE
) & 0x1f)) &
151 static __inline__
struct fib6_node
* node_alloc(void)
153 struct fib6_node
*fn
;
155 fn
= kmem_cache_zalloc(fib6_node_kmem
, GFP_ATOMIC
);
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
))
171 static void fib6_link_table(struct net
*net
, struct fib6_table
*tb
)
176 * Initialize table lock at a single place to give lockdep a key,
177 * tables aren't visible prior to being linked to the list.
179 rwlock_init(&tb
->tb6_lock
);
181 h
= tb
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1);
184 * No protection necessary, this is the only list mutatation
185 * operation, tables never disappear once they exist.
187 hlist_add_head_rcu(&tb
->tb6_hlist
, &net
->ipv6
.fib_table_hash
[h
]);
190 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
192 static struct fib6_table
*fib6_alloc_table(struct net
*net
, u32 id
)
194 struct fib6_table
*table
;
196 table
= kzalloc(sizeof(*table
), GFP_ATOMIC
);
199 table
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
200 table
->tb6_root
.fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
206 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
208 struct fib6_table
*tb
;
212 tb
= fib6_get_table(net
, id
);
216 tb
= fib6_alloc_table(net
, id
);
218 fib6_link_table(net
, tb
);
223 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
225 struct fib6_table
*tb
;
226 struct hlist_head
*head
;
227 struct hlist_node
*node
;
232 h
= id
& (FIB6_TABLE_HASHSZ
- 1);
234 head
= &net
->ipv6
.fib_table_hash
[h
];
235 hlist_for_each_entry_rcu(tb
, node
, head
, tb6_hlist
) {
236 if (tb
->tb6_id
== id
) {
246 static void __net_init
fib6_tables_init(struct net
*net
)
248 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
249 fib6_link_table(net
, net
->ipv6
.fib6_local_tbl
);
253 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
255 return fib6_get_table(net
, id
);
258 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
260 return net
->ipv6
.fib6_main_tbl
;
263 struct dst_entry
*fib6_rule_lookup(struct net
*net
, struct flowi6
*fl6
,
264 int flags
, pol_lookup_t lookup
)
266 return (struct dst_entry
*) lookup(net
, net
->ipv6
.fib6_main_tbl
, fl6
, flags
);
269 static void __net_init
fib6_tables_init(struct net
*net
)
271 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
276 static int fib6_dump_node(struct fib6_walker_t
*w
)
281 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
282 res
= rt6_dump_route(rt
, w
->args
);
284 /* Frame is full, suspend walking */
294 static void fib6_dump_end(struct netlink_callback
*cb
)
296 struct fib6_walker_t
*w
= (void*)cb
->args
[2];
301 fib6_walker_unlink(w
);
306 cb
->done
= (void*)cb
->args
[3];
310 static int fib6_dump_done(struct netlink_callback
*cb
)
313 return cb
->done
? cb
->done(cb
) : 0;
316 static int fib6_dump_table(struct fib6_table
*table
, struct sk_buff
*skb
,
317 struct netlink_callback
*cb
)
319 struct fib6_walker_t
*w
;
322 w
= (void *)cb
->args
[2];
323 w
->root
= &table
->tb6_root
;
325 if (cb
->args
[4] == 0) {
329 read_lock_bh(&table
->tb6_lock
);
331 read_unlock_bh(&table
->tb6_lock
);
334 cb
->args
[5] = w
->root
->fn_sernum
;
337 if (cb
->args
[5] != w
->root
->fn_sernum
) {
338 /* Begin at the root if the tree changed */
339 cb
->args
[5] = w
->root
->fn_sernum
;
346 read_lock_bh(&table
->tb6_lock
);
347 res
= fib6_walk_continue(w
);
348 read_unlock_bh(&table
->tb6_lock
);
350 fib6_walker_unlink(w
);
358 static int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
360 struct net
*net
= sock_net(skb
->sk
);
362 unsigned int e
= 0, s_e
;
363 struct rt6_rtnl_dump_arg arg
;
364 struct fib6_walker_t
*w
;
365 struct fib6_table
*tb
;
366 struct hlist_node
*node
;
367 struct hlist_head
*head
;
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
;
398 for (h
= s_h
; h
< FIB6_TABLE_HASHSZ
; h
++, s_e
= 0) {
400 head
= &net
->ipv6
.fib_table_hash
[h
];
401 hlist_for_each_entry_rcu(tb
, node
, head
, tb6_hlist
) {
404 res
= fib6_dump_table(tb
, skb
, cb
);
416 res
= res
< 0 ? res
: skb
->len
;
425 * return the appropriate node for a routing tree "add" operation
426 * by either creating and inserting or by returning an existing
430 static struct fib6_node
* fib6_add_1(struct fib6_node
*root
, void *addr
,
431 int addrlen
, int plen
,
434 struct fib6_node
*fn
, *in
, *ln
;
435 struct fib6_node
*pn
= NULL
;
439 __u32 sernum
= fib6_new_sernum();
441 RT6_TRACE("fib6_add_1\n");
443 /* insert node in tree */
448 key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
453 if (plen
< fn
->fn_bit
||
454 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
461 if (plen
== fn
->fn_bit
) {
462 /* clean up an intermediate node */
463 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
464 rt6_release(fn
->leaf
);
468 fn
->fn_sernum
= sernum
;
474 * We have more bits to go
477 /* Try to walk down on tree. */
478 fn
->fn_sernum
= sernum
;
479 dir
= addr_bit_set(addr
, fn
->fn_bit
);
481 fn
= dir
? fn
->right
: fn
->left
;
485 * We walked to the bottom of tree.
486 * Create new leaf node without children.
496 ln
->fn_sernum
= sernum
;
508 * split since we don't have a common prefix anymore or
509 * we have a less significant route.
510 * we've to insert an intermediate node on the list
511 * this new node will point to the one we need to create
517 /* find 1st bit in difference between the 2 addrs.
519 See comment in __ipv6_addr_diff: bit may be an invalid value,
520 but if it is >= plen, the value is ignored in any case.
523 bit
= __ipv6_addr_diff(addr
, &key
->addr
, addrlen
);
528 * (new leaf node)[ln] (old node)[fn]
534 if (in
== NULL
|| ln
== NULL
) {
543 * new intermediate node.
545 * be off since that an address that chooses one of
546 * the branches would not match less specific routes
547 * in the other branch
554 atomic_inc(&in
->leaf
->rt6i_ref
);
556 in
->fn_sernum
= sernum
;
558 /* update parent pointer */
569 ln
->fn_sernum
= sernum
;
571 if (addr_bit_set(addr
, bit
)) {
578 } else { /* plen <= bit */
581 * (new leaf node)[ln]
583 * (old node)[fn] NULL
595 ln
->fn_sernum
= sernum
;
602 if (addr_bit_set(&key
->addr
, plen
))
613 * Insert routing information in a node.
616 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
617 struct nl_info
*info
)
619 struct rt6_info
*iter
= NULL
;
620 struct rt6_info
**ins
;
624 for (iter
= fn
->leaf
; iter
; iter
=iter
->dst
.rt6_next
) {
626 * Search for duplicates
629 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
631 * Same priority level
634 if (iter
->rt6i_dev
== rt
->rt6i_dev
&&
635 iter
->rt6i_idev
== rt
->rt6i_idev
&&
636 ipv6_addr_equal(&iter
->rt6i_gateway
,
637 &rt
->rt6i_gateway
)) {
638 if (!(iter
->rt6i_flags
&RTF_EXPIRES
))
640 iter
->rt6i_expires
= rt
->rt6i_expires
;
641 if (!(rt
->rt6i_flags
&RTF_EXPIRES
)) {
642 iter
->rt6i_flags
&= ~RTF_EXPIRES
;
643 iter
->rt6i_expires
= 0;
649 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
652 ins
= &iter
->dst
.rt6_next
;
655 /* Reset round-robin state, if necessary */
656 if (ins
== &fn
->leaf
)
663 rt
->dst
.rt6_next
= iter
;
666 atomic_inc(&rt
->rt6i_ref
);
667 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
);
668 info
->nl_net
->ipv6
.rt6_stats
->fib_rt_entries
++;
670 if ((fn
->fn_flags
& RTN_RTINFO
) == 0) {
671 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
672 fn
->fn_flags
|= RTN_RTINFO
;
678 static __inline__
void fib6_start_gc(struct net
*net
, struct rt6_info
*rt
)
680 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
) &&
681 (rt
->rt6i_flags
& (RTF_EXPIRES
|RTF_CACHE
)))
682 mod_timer(&net
->ipv6
.ip6_fib_timer
,
683 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
686 void fib6_force_start_gc(struct net
*net
)
688 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
))
689 mod_timer(&net
->ipv6
.ip6_fib_timer
,
690 jiffies
+ net
->ipv6
.sysctl
.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
= info
->nl_net
->ipv6
.ip6_null_entry
;
735 atomic_inc(&info
->nl_net
->ipv6
.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
);
777 fib6_start_gc(info
->nl_net
, rt
);
778 if (!(rt
->rt6i_flags
&RTF_CACHE
))
779 fib6_prune_clones(info
->nl_net
, 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
== rt
) {
791 atomic_dec(&rt
->rt6i_ref
);
793 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
794 pn
->leaf
= fib6_find_prefix(info
->nl_net
, pn
);
797 WARN_ON(pn
->leaf
== NULL
);
798 pn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
801 atomic_inc(&pn
->leaf
->rt6i_ref
);
808 #ifdef CONFIG_IPV6_SUBTREES
809 /* Subtree creation failed, probably main tree node
810 is orphan. If it is, shoot it.
813 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
814 fib6_repair_tree(info
->nl_net
, fn
);
821 * Routing tree lookup
826 int offset
; /* key offset on rt6_info */
827 const struct in6_addr
*addr
; /* search key */
830 static struct fib6_node
* fib6_lookup_1(struct fib6_node
*root
,
831 struct lookup_args
*args
)
833 struct fib6_node
*fn
;
836 if (unlikely(args
->offset
== 0))
846 struct fib6_node
*next
;
848 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
850 next
= dir
? fn
->right
: fn
->left
;
861 if (FIB6_SUBTREE(fn
) || fn
->fn_flags
& RTN_RTINFO
) {
864 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
867 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
)) {
868 #ifdef CONFIG_IPV6_SUBTREES
870 fn
= fib6_lookup_1(fn
->subtree
, args
+ 1);
872 if (!fn
|| fn
->fn_flags
& RTN_RTINFO
)
877 if (fn
->fn_flags
& RTN_ROOT
)
886 struct fib6_node
* fib6_lookup(struct fib6_node
*root
, const struct in6_addr
*daddr
,
887 const struct in6_addr
*saddr
)
889 struct fib6_node
*fn
;
890 struct lookup_args args
[] = {
892 .offset
= offsetof(struct rt6_info
, rt6i_dst
),
895 #ifdef CONFIG_IPV6_SUBTREES
897 .offset
= offsetof(struct rt6_info
, rt6i_src
),
902 .offset
= 0, /* sentinel */
906 fn
= fib6_lookup_1(root
, daddr
? args
: args
+ 1);
908 if (fn
== NULL
|| fn
->fn_flags
& RTN_TL_ROOT
)
915 * Get node with specified destination prefix (and source prefix,
916 * if subtrees are used)
920 static struct fib6_node
* fib6_locate_1(struct fib6_node
*root
,
921 const struct in6_addr
*addr
,
922 int plen
, int offset
)
924 struct fib6_node
*fn
;
926 for (fn
= root
; fn
; ) {
927 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
932 if (plen
< fn
->fn_bit
||
933 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
936 if (plen
== fn
->fn_bit
)
940 * We have more bits to go
942 if (addr_bit_set(addr
, fn
->fn_bit
))
950 struct fib6_node
* fib6_locate(struct fib6_node
*root
,
951 const struct in6_addr
*daddr
, int dst_len
,
952 const struct in6_addr
*saddr
, int src_len
)
954 struct fib6_node
*fn
;
956 fn
= fib6_locate_1(root
, daddr
, dst_len
,
957 offsetof(struct rt6_info
, rt6i_dst
));
959 #ifdef CONFIG_IPV6_SUBTREES
961 WARN_ON(saddr
== NULL
);
962 if (fn
&& fn
->subtree
)
963 fn
= fib6_locate_1(fn
->subtree
, saddr
, src_len
,
964 offsetof(struct rt6_info
, rt6i_src
));
968 if (fn
&& fn
->fn_flags
&RTN_RTINFO
)
980 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
)
982 if (fn
->fn_flags
&RTN_ROOT
)
983 return net
->ipv6
.ip6_null_entry
;
987 return fn
->left
->leaf
;
990 return fn
->right
->leaf
;
992 fn
= FIB6_SUBTREE(fn
);
998 * Called to trim the tree of intermediate nodes when possible. "fn"
999 * is the node we want to try and remove.
1002 static struct fib6_node
*fib6_repair_tree(struct net
*net
,
1003 struct fib6_node
*fn
)
1007 struct fib6_node
*child
, *pn
;
1008 struct fib6_walker_t
*w
;
1012 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
1015 WARN_ON(fn
->fn_flags
& RTN_RTINFO
);
1016 WARN_ON(fn
->fn_flags
& RTN_TL_ROOT
);
1017 WARN_ON(fn
->leaf
!= NULL
);
1021 if (fn
->right
) child
= fn
->right
, children
|= 1;
1022 if (fn
->left
) child
= fn
->left
, children
|= 2;
1024 if (children
== 3 || FIB6_SUBTREE(fn
)
1025 #ifdef CONFIG_IPV6_SUBTREES
1026 /* Subtree root (i.e. fn) may have one child */
1027 || (children
&& fn
->fn_flags
&RTN_ROOT
)
1030 fn
->leaf
= fib6_find_prefix(net
, fn
);
1032 if (fn
->leaf
==NULL
) {
1034 fn
->leaf
= net
->ipv6
.ip6_null_entry
;
1037 atomic_inc(&fn
->leaf
->rt6i_ref
);
1042 #ifdef CONFIG_IPV6_SUBTREES
1043 if (FIB6_SUBTREE(pn
) == fn
) {
1044 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1045 FIB6_SUBTREE(pn
) = NULL
;
1048 WARN_ON(fn
->fn_flags
& RTN_ROOT
);
1050 if (pn
->right
== fn
) pn
->right
= child
;
1051 else if (pn
->left
== fn
) pn
->left
= child
;
1059 #ifdef CONFIG_IPV6_SUBTREES
1063 read_lock(&fib6_walker_lock
);
1065 if (child
== NULL
) {
1066 if (w
->root
== fn
) {
1067 w
->root
= w
->node
= NULL
;
1068 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1069 } else if (w
->node
== fn
) {
1070 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1075 if (w
->root
== fn
) {
1077 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1079 if (w
->node
== fn
) {
1082 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1083 w
->state
= w
->state
>=FWS_R
? FWS_U
: FWS_INIT
;
1085 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1086 w
->state
= w
->state
>=FWS_C
? FWS_U
: FWS_INIT
;
1091 read_unlock(&fib6_walker_lock
);
1094 if (pn
->fn_flags
&RTN_RTINFO
|| FIB6_SUBTREE(pn
))
1097 rt6_release(pn
->leaf
);
1103 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1104 struct nl_info
*info
)
1106 struct fib6_walker_t
*w
;
1107 struct rt6_info
*rt
= *rtp
;
1108 struct net
*net
= info
->nl_net
;
1110 RT6_TRACE("fib6_del_route\n");
1113 *rtp
= rt
->dst
.rt6_next
;
1114 rt
->rt6i_node
= NULL
;
1115 net
->ipv6
.rt6_stats
->fib_rt_entries
--;
1116 net
->ipv6
.rt6_stats
->fib_discarded_routes
++;
1118 /* Reset round-robin state, if necessary */
1119 if (fn
->rr_ptr
== rt
)
1122 /* Adjust walkers */
1123 read_lock(&fib6_walker_lock
);
1125 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1126 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1127 w
->leaf
= rt
->dst
.rt6_next
;
1128 if (w
->leaf
== NULL
)
1132 read_unlock(&fib6_walker_lock
);
1134 rt
->dst
.rt6_next
= NULL
;
1136 /* If it was last route, expunge its radix tree node */
1137 if (fn
->leaf
== NULL
) {
1138 fn
->fn_flags
&= ~RTN_RTINFO
;
1139 net
->ipv6
.rt6_stats
->fib_route_nodes
--;
1140 fn
= fib6_repair_tree(net
, fn
);
1143 if (atomic_read(&rt
->rt6i_ref
) != 1) {
1144 /* This route is used as dummy address holder in some split
1145 * nodes. It is not leaked, but it still holds other resources,
1146 * which must be released in time. So, scan ascendant nodes
1147 * and replace dummy references to this route with references
1148 * to still alive ones.
1151 if (!(fn
->fn_flags
&RTN_RTINFO
) && fn
->leaf
== rt
) {
1152 fn
->leaf
= fib6_find_prefix(net
, fn
);
1153 atomic_inc(&fn
->leaf
->rt6i_ref
);
1158 /* No more references are possible at this point. */
1159 BUG_ON(atomic_read(&rt
->rt6i_ref
) != 1);
1162 inet6_rt_notify(RTM_DELROUTE
, rt
, info
);
1166 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1168 struct net
*net
= info
->nl_net
;
1169 struct fib6_node
*fn
= rt
->rt6i_node
;
1170 struct rt6_info
**rtp
;
1173 if (rt
->dst
.obsolete
>0) {
1174 WARN_ON(fn
!= NULL
);
1178 if (fn
== NULL
|| rt
== net
->ipv6
.ip6_null_entry
)
1181 WARN_ON(!(fn
->fn_flags
& RTN_RTINFO
));
1183 if (!(rt
->rt6i_flags
&RTF_CACHE
)) {
1184 struct fib6_node
*pn
= fn
;
1185 #ifdef CONFIG_IPV6_SUBTREES
1186 /* clones of this route might be in another subtree */
1187 if (rt
->rt6i_src
.plen
) {
1188 while (!(pn
->fn_flags
&RTN_ROOT
))
1193 fib6_prune_clones(info
->nl_net
, pn
, rt
);
1197 * Walk the leaf entries looking for ourself
1200 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->dst
.rt6_next
) {
1202 fib6_del_route(fn
, rtp
, info
);
1210 * Tree traversal function.
1212 * Certainly, it is not interrupt safe.
1213 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1214 * It means, that we can modify tree during walking
1215 * and use this function for garbage collection, clone pruning,
1216 * cleaning tree when a device goes down etc. etc.
1218 * It guarantees that every node will be traversed,
1219 * and that it will be traversed only once.
1221 * Callback function w->func may return:
1222 * 0 -> continue walking.
1223 * positive value -> walking is suspended (used by tree dumps,
1224 * and probably by gc, if it will be split to several slices)
1225 * negative value -> terminate walking.
1227 * The function itself returns:
1228 * 0 -> walk is complete.
1229 * >0 -> walk is incomplete (i.e. suspended)
1230 * <0 -> walk is terminated by an error.
1233 static int fib6_walk_continue(struct fib6_walker_t
*w
)
1235 struct fib6_node
*fn
, *pn
;
1242 if (w
->prune
&& fn
!= w
->root
&&
1243 fn
->fn_flags
&RTN_RTINFO
&& w
->state
< FWS_C
) {
1248 #ifdef CONFIG_IPV6_SUBTREES
1250 if (FIB6_SUBTREE(fn
)) {
1251 w
->node
= FIB6_SUBTREE(fn
);
1259 w
->state
= FWS_INIT
;
1265 w
->node
= fn
->right
;
1266 w
->state
= FWS_INIT
;
1272 if (w
->leaf
&& fn
->fn_flags
&RTN_RTINFO
) {
1275 if (w
->count
< w
->skip
) {
1293 #ifdef CONFIG_IPV6_SUBTREES
1294 if (FIB6_SUBTREE(pn
) == fn
) {
1295 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1300 if (pn
->left
== fn
) {
1304 if (pn
->right
== fn
) {
1306 w
->leaf
= w
->node
->leaf
;
1316 static int fib6_walk(struct fib6_walker_t
*w
)
1320 w
->state
= FWS_INIT
;
1323 fib6_walker_link(w
);
1324 res
= fib6_walk_continue(w
);
1326 fib6_walker_unlink(w
);
1330 static int fib6_clean_node(struct fib6_walker_t
*w
)
1333 struct rt6_info
*rt
;
1334 struct fib6_cleaner_t
*c
= container_of(w
, struct fib6_cleaner_t
, w
);
1335 struct nl_info info
= {
1339 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1340 res
= c
->func(rt
, c
->arg
);
1343 res
= fib6_del(rt
, &info
);
1346 printk(KERN_DEBUG
"fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt
, rt
->rt6i_node
, res
);
1359 * Convenient frontend to tree walker.
1361 * func is called on each route.
1362 * It may return -1 -> delete this route.
1363 * 0 -> continue walking
1365 * prune==1 -> only immediate children of node (certainly,
1366 * ignoring pure split nodes) will be scanned.
1369 static void fib6_clean_tree(struct net
*net
, struct fib6_node
*root
,
1370 int (*func
)(struct rt6_info
*, void *arg
),
1371 int prune
, void *arg
)
1373 struct fib6_cleaner_t c
;
1376 c
.w
.func
= fib6_clean_node
;
1387 void fib6_clean_all(struct net
*net
, int (*func
)(struct rt6_info
*, void *arg
),
1388 int prune
, void *arg
)
1390 struct fib6_table
*table
;
1391 struct hlist_node
*node
;
1392 struct hlist_head
*head
;
1396 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1397 head
= &net
->ipv6
.fib_table_hash
[h
];
1398 hlist_for_each_entry_rcu(table
, node
, head
, tb6_hlist
) {
1399 write_lock_bh(&table
->tb6_lock
);
1400 fib6_clean_tree(net
, &table
->tb6_root
,
1402 write_unlock_bh(&table
->tb6_lock
);
1408 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1410 if (rt
->rt6i_flags
& RTF_CACHE
) {
1411 RT6_TRACE("pruning clone %p\n", rt
);
1418 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
,
1419 struct rt6_info
*rt
)
1421 fib6_clean_tree(net
, fn
, fib6_prune_clone
, 1, rt
);
1425 * Garbage collection
1428 static struct fib6_gc_args
1434 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1436 unsigned long now
= jiffies
;
1439 * check addrconf expiration here.
1440 * Routes are expired even if they are in use.
1442 * Also age clones. Note, that clones are aged out
1443 * only if they are not in use now.
1446 if (rt
->rt6i_flags
&RTF_EXPIRES
&& rt
->rt6i_expires
) {
1447 if (time_after(now
, rt
->rt6i_expires
)) {
1448 RT6_TRACE("expiring %p\n", rt
);
1452 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1453 if (atomic_read(&rt
->dst
.__refcnt
) == 0 &&
1454 time_after_eq(now
, rt
->dst
.lastuse
+ gc_args
.timeout
)) {
1455 RT6_TRACE("aging clone %p\n", rt
);
1457 } else if ((rt
->rt6i_flags
& RTF_GATEWAY
) &&
1458 (!(dst_get_neighbour_raw(&rt
->dst
)->flags
& NTF_ROUTER
))) {
1459 RT6_TRACE("purging route %p via non-router but gateway\n",
1469 static DEFINE_SPINLOCK(fib6_gc_lock
);
1471 void fib6_run_gc(unsigned long expires
, struct net
*net
)
1473 if (expires
!= ~0UL) {
1474 spin_lock_bh(&fib6_gc_lock
);
1475 gc_args
.timeout
= expires
? (int)expires
:
1476 net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1478 if (!spin_trylock_bh(&fib6_gc_lock
)) {
1479 mod_timer(&net
->ipv6
.ip6_fib_timer
, jiffies
+ HZ
);
1482 gc_args
.timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1485 gc_args
.more
= icmp6_dst_gc();
1487 fib6_clean_all(net
, fib6_age
, 0, NULL
);
1490 mod_timer(&net
->ipv6
.ip6_fib_timer
,
1491 round_jiffies(jiffies
1492 + net
->ipv6
.sysctl
.ip6_rt_gc_interval
));
1494 del_timer(&net
->ipv6
.ip6_fib_timer
);
1495 spin_unlock_bh(&fib6_gc_lock
);
1498 static void fib6_gc_timer_cb(unsigned long arg
)
1500 fib6_run_gc(0, (struct net
*)arg
);
1503 static int __net_init
fib6_net_init(struct net
*net
)
1505 size_t size
= sizeof(struct hlist_head
) * FIB6_TABLE_HASHSZ
;
1507 setup_timer(&net
->ipv6
.ip6_fib_timer
, fib6_gc_timer_cb
, (unsigned long)net
);
1509 net
->ipv6
.rt6_stats
= kzalloc(sizeof(*net
->ipv6
.rt6_stats
), GFP_KERNEL
);
1510 if (!net
->ipv6
.rt6_stats
)
1513 /* Avoid false sharing : Use at least a full cache line */
1514 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1516 net
->ipv6
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1517 if (!net
->ipv6
.fib_table_hash
)
1520 net
->ipv6
.fib6_main_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_main_tbl
),
1522 if (!net
->ipv6
.fib6_main_tbl
)
1523 goto out_fib_table_hash
;
1525 net
->ipv6
.fib6_main_tbl
->tb6_id
= RT6_TABLE_MAIN
;
1526 net
->ipv6
.fib6_main_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1527 net
->ipv6
.fib6_main_tbl
->tb6_root
.fn_flags
=
1528 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1530 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1531 net
->ipv6
.fib6_local_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_local_tbl
),
1533 if (!net
->ipv6
.fib6_local_tbl
)
1534 goto out_fib6_main_tbl
;
1535 net
->ipv6
.fib6_local_tbl
->tb6_id
= RT6_TABLE_LOCAL
;
1536 net
->ipv6
.fib6_local_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1537 net
->ipv6
.fib6_local_tbl
->tb6_root
.fn_flags
=
1538 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1540 fib6_tables_init(net
);
1544 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1546 kfree(net
->ipv6
.fib6_main_tbl
);
1549 kfree(net
->ipv6
.fib_table_hash
);
1551 kfree(net
->ipv6
.rt6_stats
);
1556 static void fib6_net_exit(struct net
*net
)
1558 rt6_ifdown(net
, NULL
);
1559 del_timer_sync(&net
->ipv6
.ip6_fib_timer
);
1561 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1562 kfree(net
->ipv6
.fib6_local_tbl
);
1564 kfree(net
->ipv6
.fib6_main_tbl
);
1565 kfree(net
->ipv6
.fib_table_hash
);
1566 kfree(net
->ipv6
.rt6_stats
);
1569 static struct pernet_operations fib6_net_ops
= {
1570 .init
= fib6_net_init
,
1571 .exit
= fib6_net_exit
,
1574 int __init
fib6_init(void)
1578 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1579 sizeof(struct fib6_node
),
1580 0, SLAB_HWCACHE_ALIGN
,
1582 if (!fib6_node_kmem
)
1585 ret
= register_pernet_subsys(&fib6_net_ops
);
1587 goto out_kmem_cache_create
;
1589 ret
= __rtnl_register(PF_INET6
, RTM_GETROUTE
, NULL
, inet6_dump_fib
,
1592 goto out_unregister_subsys
;
1596 out_unregister_subsys
:
1597 unregister_pernet_subsys(&fib6_net_ops
);
1598 out_kmem_cache_create
:
1599 kmem_cache_destroy(fib6_node_kmem
);
1603 void fib6_gc_cleanup(void)
1605 unregister_pernet_subsys(&fib6_net_ops
);
1606 kmem_cache_destroy(fib6_node_kmem
);