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
,
428 int offset
, int allow_create
,
429 int replace_required
)
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
)) {
453 if (replace_required
) {
454 pr_warn("IPv6: Can't replace route, "
456 return ERR_PTR(-ENOENT
);
458 pr_warn("IPv6: NLM_F_CREATE should be set "
459 "when creating new route\n");
468 if (plen
== fn
->fn_bit
) {
469 /* clean up an intermediate node */
470 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
471 rt6_release(fn
->leaf
);
475 fn
->fn_sernum
= sernum
;
481 * We have more bits to go
484 /* Try to walk down on tree. */
485 fn
->fn_sernum
= sernum
;
486 dir
= addr_bit_set(addr
, fn
->fn_bit
);
488 fn
= dir
? fn
->right
: fn
->left
;
492 /* We should not create new node because
493 * NLM_F_REPLACE was specified without NLM_F_CREATE
494 * I assume it is safe to require NLM_F_CREATE when
495 * REPLACE flag is used! Later we may want to remove the
496 * check for replace_required, because according
497 * to netlink specification, NLM_F_CREATE
498 * MUST be specified if new route is created.
499 * That would keep IPv6 consistent with IPv4
501 if (replace_required
) {
502 pr_warn("IPv6: Can't replace route, no match found\n");
503 return ERR_PTR(-ENOENT
);
505 pr_warn("IPv6: NLM_F_CREATE should be set "
506 "when creating new route\n");
509 * We walked to the bottom of tree.
510 * Create new leaf node without children.
520 ln
->fn_sernum
= sernum
;
532 * split since we don't have a common prefix anymore or
533 * we have a less significant route.
534 * we've to insert an intermediate node on the list
535 * this new node will point to the one we need to create
541 /* find 1st bit in difference between the 2 addrs.
543 See comment in __ipv6_addr_diff: bit may be an invalid value,
544 but if it is >= plen, the value is ignored in any case.
547 bit
= __ipv6_addr_diff(addr
, &key
->addr
, addrlen
);
552 * (new leaf node)[ln] (old node)[fn]
567 * new intermediate node.
569 * be off since that an address that chooses one of
570 * the branches would not match less specific routes
571 * in the other branch
578 atomic_inc(&in
->leaf
->rt6i_ref
);
580 in
->fn_sernum
= sernum
;
582 /* update parent pointer */
593 ln
->fn_sernum
= sernum
;
595 if (addr_bit_set(addr
, bit
)) {
602 } else { /* plen <= bit */
605 * (new leaf node)[ln]
607 * (old node)[fn] NULL
619 ln
->fn_sernum
= sernum
;
626 if (addr_bit_set(&key
->addr
, plen
))
637 * Insert routing information in a node.
640 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
641 struct nl_info
*info
)
643 struct rt6_info
*iter
= NULL
;
644 struct rt6_info
**ins
;
645 int replace
= (info
->nlh
&&
646 (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
));
647 int add
= (!info
->nlh
||
648 (info
->nlh
->nlmsg_flags
& NLM_F_CREATE
));
653 for (iter
= fn
->leaf
; iter
; iter
= iter
->dst
.rt6_next
) {
655 * Search for duplicates
658 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
660 * Same priority level
663 (info
->nlh
->nlmsg_flags
& NLM_F_EXCL
))
670 if (iter
->dst
.dev
== rt
->dst
.dev
&&
671 iter
->rt6i_idev
== rt
->rt6i_idev
&&
672 ipv6_addr_equal(&iter
->rt6i_gateway
,
673 &rt
->rt6i_gateway
)) {
674 if (!(iter
->rt6i_flags
& RTF_EXPIRES
))
676 iter
->dst
.expires
= rt
->dst
.expires
;
677 if (!(rt
->rt6i_flags
& RTF_EXPIRES
)) {
678 iter
->rt6i_flags
&= ~RTF_EXPIRES
;
679 iter
->dst
.expires
= 0;
685 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
688 ins
= &iter
->dst
.rt6_next
;
691 /* Reset round-robin state, if necessary */
692 if (ins
== &fn
->leaf
)
700 pr_warn("IPv6: NLM_F_CREATE should be set when creating new route\n");
703 rt
->dst
.rt6_next
= iter
;
706 atomic_inc(&rt
->rt6i_ref
);
707 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
);
708 info
->nl_net
->ipv6
.rt6_stats
->fib_rt_entries
++;
710 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
711 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
712 fn
->fn_flags
|= RTN_RTINFO
;
719 pr_warn("IPv6: NLM_F_REPLACE set, but no existing node found!\n");
724 rt
->dst
.rt6_next
= iter
->dst
.rt6_next
;
725 atomic_inc(&rt
->rt6i_ref
);
726 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
);
728 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
729 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
730 fn
->fn_flags
|= RTN_RTINFO
;
737 static __inline__
void fib6_start_gc(struct net
*net
, struct rt6_info
*rt
)
739 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
) &&
740 (rt
->rt6i_flags
& (RTF_EXPIRES
| RTF_CACHE
)))
741 mod_timer(&net
->ipv6
.ip6_fib_timer
,
742 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
745 void fib6_force_start_gc(struct net
*net
)
747 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
))
748 mod_timer(&net
->ipv6
.ip6_fib_timer
,
749 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
753 * Add routing information to the routing tree.
754 * <destination addr>/<source addr>
755 * with source addr info in sub-trees
758 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
, struct nl_info
*info
)
760 struct fib6_node
*fn
, *pn
= NULL
;
762 int allow_create
= 1;
763 int replace_required
= 0;
766 if (!(info
->nlh
->nlmsg_flags
& NLM_F_CREATE
))
768 if (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
)
769 replace_required
= 1;
771 if (!allow_create
&& !replace_required
)
772 pr_warn("IPv6: RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
774 fn
= fib6_add_1(root
, &rt
->rt6i_dst
.addr
, sizeof(struct in6_addr
),
775 rt
->rt6i_dst
.plen
, offsetof(struct rt6_info
, rt6i_dst
),
776 allow_create
, replace_required
);
788 #ifdef CONFIG_IPV6_SUBTREES
789 if (rt
->rt6i_src
.plen
) {
790 struct fib6_node
*sn
;
793 struct fib6_node
*sfn
;
805 /* Create subtree root node */
810 sfn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
811 atomic_inc(&info
->nl_net
->ipv6
.ip6_null_entry
->rt6i_ref
);
812 sfn
->fn_flags
= RTN_ROOT
;
813 sfn
->fn_sernum
= fib6_new_sernum();
815 /* Now add the first leaf node to new subtree */
817 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
818 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
819 offsetof(struct rt6_info
, rt6i_src
),
820 allow_create
, replace_required
);
823 /* If it is failed, discard just allocated
824 root, and then (in st_failure) stale node
831 /* Now link new subtree to main tree */
835 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
836 sizeof(struct in6_addr
), rt
->rt6i_src
.plen
,
837 offsetof(struct rt6_info
, rt6i_src
),
838 allow_create
, replace_required
);
850 atomic_inc(&rt
->rt6i_ref
);
856 err
= fib6_add_rt2node(fn
, rt
, info
);
858 fib6_start_gc(info
->nl_net
, rt
);
859 if (!(rt
->rt6i_flags
& RTF_CACHE
))
860 fib6_prune_clones(info
->nl_net
, pn
, rt
);
865 #ifdef CONFIG_IPV6_SUBTREES
867 * If fib6_add_1 has cleared the old leaf pointer in the
868 * super-tree leaf node we have to find a new one for it.
870 if (pn
!= fn
&& pn
->leaf
== rt
) {
872 atomic_dec(&rt
->rt6i_ref
);
874 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
875 pn
->leaf
= fib6_find_prefix(info
->nl_net
, pn
);
878 WARN_ON(pn
->leaf
== NULL
);
879 pn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
882 atomic_inc(&pn
->leaf
->rt6i_ref
);
889 #ifdef CONFIG_IPV6_SUBTREES
890 /* Subtree creation failed, probably main tree node
891 is orphan. If it is, shoot it.
894 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
895 fib6_repair_tree(info
->nl_net
, fn
);
902 * Routing tree lookup
907 int offset
; /* key offset on rt6_info */
908 const struct in6_addr
*addr
; /* search key */
911 static struct fib6_node
* fib6_lookup_1(struct fib6_node
*root
,
912 struct lookup_args
*args
)
914 struct fib6_node
*fn
;
917 if (unlikely(args
->offset
== 0))
927 struct fib6_node
*next
;
929 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
931 next
= dir
? fn
->right
: fn
->left
;
941 if (FIB6_SUBTREE(fn
) || fn
->fn_flags
& RTN_RTINFO
) {
944 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
947 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
)) {
948 #ifdef CONFIG_IPV6_SUBTREES
950 fn
= fib6_lookup_1(fn
->subtree
, args
+ 1);
952 if (!fn
|| fn
->fn_flags
& RTN_RTINFO
)
957 if (fn
->fn_flags
& RTN_ROOT
)
966 struct fib6_node
* fib6_lookup(struct fib6_node
*root
, const struct in6_addr
*daddr
,
967 const struct in6_addr
*saddr
)
969 struct fib6_node
*fn
;
970 struct lookup_args args
[] = {
972 .offset
= offsetof(struct rt6_info
, rt6i_dst
),
975 #ifdef CONFIG_IPV6_SUBTREES
977 .offset
= offsetof(struct rt6_info
, rt6i_src
),
982 .offset
= 0, /* sentinel */
986 fn
= fib6_lookup_1(root
, daddr
? args
: args
+ 1);
987 if (!fn
|| fn
->fn_flags
& RTN_TL_ROOT
)
994 * Get node with specified destination prefix (and source prefix,
995 * if subtrees are used)
999 static struct fib6_node
* fib6_locate_1(struct fib6_node
*root
,
1000 const struct in6_addr
*addr
,
1001 int plen
, int offset
)
1003 struct fib6_node
*fn
;
1005 for (fn
= root
; fn
; ) {
1006 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
1011 if (plen
< fn
->fn_bit
||
1012 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
1015 if (plen
== fn
->fn_bit
)
1019 * We have more bits to go
1021 if (addr_bit_set(addr
, fn
->fn_bit
))
1029 struct fib6_node
* fib6_locate(struct fib6_node
*root
,
1030 const struct in6_addr
*daddr
, int dst_len
,
1031 const struct in6_addr
*saddr
, int src_len
)
1033 struct fib6_node
*fn
;
1035 fn
= fib6_locate_1(root
, daddr
, dst_len
,
1036 offsetof(struct rt6_info
, rt6i_dst
));
1038 #ifdef CONFIG_IPV6_SUBTREES
1040 WARN_ON(saddr
== NULL
);
1041 if (fn
&& fn
->subtree
)
1042 fn
= fib6_locate_1(fn
->subtree
, saddr
, src_len
,
1043 offsetof(struct rt6_info
, rt6i_src
));
1047 if (fn
&& fn
->fn_flags
& RTN_RTINFO
)
1059 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
)
1061 if (fn
->fn_flags
& RTN_ROOT
)
1062 return net
->ipv6
.ip6_null_entry
;
1066 return fn
->left
->leaf
;
1068 return fn
->right
->leaf
;
1070 fn
= FIB6_SUBTREE(fn
);
1076 * Called to trim the tree of intermediate nodes when possible. "fn"
1077 * is the node we want to try and remove.
1080 static struct fib6_node
*fib6_repair_tree(struct net
*net
,
1081 struct fib6_node
*fn
)
1085 struct fib6_node
*child
, *pn
;
1086 struct fib6_walker_t
*w
;
1090 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
1093 WARN_ON(fn
->fn_flags
& RTN_RTINFO
);
1094 WARN_ON(fn
->fn_flags
& RTN_TL_ROOT
);
1095 WARN_ON(fn
->leaf
!= NULL
);
1099 if (fn
->right
) child
= fn
->right
, children
|= 1;
1100 if (fn
->left
) child
= fn
->left
, children
|= 2;
1102 if (children
== 3 || FIB6_SUBTREE(fn
)
1103 #ifdef CONFIG_IPV6_SUBTREES
1104 /* Subtree root (i.e. fn) may have one child */
1105 || (children
&& fn
->fn_flags
& RTN_ROOT
)
1108 fn
->leaf
= fib6_find_prefix(net
, fn
);
1112 fn
->leaf
= net
->ipv6
.ip6_null_entry
;
1115 atomic_inc(&fn
->leaf
->rt6i_ref
);
1120 #ifdef CONFIG_IPV6_SUBTREES
1121 if (FIB6_SUBTREE(pn
) == fn
) {
1122 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1123 FIB6_SUBTREE(pn
) = NULL
;
1126 WARN_ON(fn
->fn_flags
& RTN_ROOT
);
1128 if (pn
->right
== fn
) pn
->right
= child
;
1129 else if (pn
->left
== fn
) pn
->left
= child
;
1137 #ifdef CONFIG_IPV6_SUBTREES
1141 read_lock(&fib6_walker_lock
);
1144 if (w
->root
== fn
) {
1145 w
->root
= w
->node
= NULL
;
1146 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1147 } else if (w
->node
== fn
) {
1148 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1153 if (w
->root
== fn
) {
1155 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1157 if (w
->node
== fn
) {
1160 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1161 w
->state
= w
->state
>=FWS_R
? FWS_U
: FWS_INIT
;
1163 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1164 w
->state
= w
->state
>=FWS_C
? FWS_U
: FWS_INIT
;
1169 read_unlock(&fib6_walker_lock
);
1172 if (pn
->fn_flags
& RTN_RTINFO
|| FIB6_SUBTREE(pn
))
1175 rt6_release(pn
->leaf
);
1181 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1182 struct nl_info
*info
)
1184 struct fib6_walker_t
*w
;
1185 struct rt6_info
*rt
= *rtp
;
1186 struct net
*net
= info
->nl_net
;
1188 RT6_TRACE("fib6_del_route\n");
1191 *rtp
= rt
->dst
.rt6_next
;
1192 rt
->rt6i_node
= NULL
;
1193 net
->ipv6
.rt6_stats
->fib_rt_entries
--;
1194 net
->ipv6
.rt6_stats
->fib_discarded_routes
++;
1196 /* Reset round-robin state, if necessary */
1197 if (fn
->rr_ptr
== rt
)
1200 /* Adjust walkers */
1201 read_lock(&fib6_walker_lock
);
1203 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1204 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1205 w
->leaf
= rt
->dst
.rt6_next
;
1210 read_unlock(&fib6_walker_lock
);
1212 rt
->dst
.rt6_next
= NULL
;
1214 /* If it was last route, expunge its radix tree node */
1216 fn
->fn_flags
&= ~RTN_RTINFO
;
1217 net
->ipv6
.rt6_stats
->fib_route_nodes
--;
1218 fn
= fib6_repair_tree(net
, fn
);
1221 if (atomic_read(&rt
->rt6i_ref
) != 1) {
1222 /* This route is used as dummy address holder in some split
1223 * nodes. It is not leaked, but it still holds other resources,
1224 * which must be released in time. So, scan ascendant nodes
1225 * and replace dummy references to this route with references
1226 * to still alive ones.
1229 if (!(fn
->fn_flags
& RTN_RTINFO
) && fn
->leaf
== rt
) {
1230 fn
->leaf
= fib6_find_prefix(net
, fn
);
1231 atomic_inc(&fn
->leaf
->rt6i_ref
);
1236 /* No more references are possible at this point. */
1237 BUG_ON(atomic_read(&rt
->rt6i_ref
) != 1);
1240 inet6_rt_notify(RTM_DELROUTE
, rt
, info
);
1244 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1246 struct net
*net
= info
->nl_net
;
1247 struct fib6_node
*fn
= rt
->rt6i_node
;
1248 struct rt6_info
**rtp
;
1251 if (rt
->dst
.obsolete
>0) {
1252 WARN_ON(fn
!= NULL
);
1256 if (!fn
|| rt
== net
->ipv6
.ip6_null_entry
)
1259 WARN_ON(!(fn
->fn_flags
& RTN_RTINFO
));
1261 if (!(rt
->rt6i_flags
& RTF_CACHE
)) {
1262 struct fib6_node
*pn
= fn
;
1263 #ifdef CONFIG_IPV6_SUBTREES
1264 /* clones of this route might be in another subtree */
1265 if (rt
->rt6i_src
.plen
) {
1266 while (!(pn
->fn_flags
& RTN_ROOT
))
1271 fib6_prune_clones(info
->nl_net
, pn
, rt
);
1275 * Walk the leaf entries looking for ourself
1278 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->dst
.rt6_next
) {
1280 fib6_del_route(fn
, rtp
, info
);
1288 * Tree traversal function.
1290 * Certainly, it is not interrupt safe.
1291 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1292 * It means, that we can modify tree during walking
1293 * and use this function for garbage collection, clone pruning,
1294 * cleaning tree when a device goes down etc. etc.
1296 * It guarantees that every node will be traversed,
1297 * and that it will be traversed only once.
1299 * Callback function w->func may return:
1300 * 0 -> continue walking.
1301 * positive value -> walking is suspended (used by tree dumps,
1302 * and probably by gc, if it will be split to several slices)
1303 * negative value -> terminate walking.
1305 * The function itself returns:
1306 * 0 -> walk is complete.
1307 * >0 -> walk is incomplete (i.e. suspended)
1308 * <0 -> walk is terminated by an error.
1311 static int fib6_walk_continue(struct fib6_walker_t
*w
)
1313 struct fib6_node
*fn
, *pn
;
1320 if (w
->prune
&& fn
!= w
->root
&&
1321 fn
->fn_flags
& RTN_RTINFO
&& w
->state
< FWS_C
) {
1326 #ifdef CONFIG_IPV6_SUBTREES
1328 if (FIB6_SUBTREE(fn
)) {
1329 w
->node
= FIB6_SUBTREE(fn
);
1337 w
->state
= FWS_INIT
;
1343 w
->node
= fn
->right
;
1344 w
->state
= FWS_INIT
;
1350 if (w
->leaf
&& fn
->fn_flags
& RTN_RTINFO
) {
1353 if (w
->count
< w
->skip
) {
1371 #ifdef CONFIG_IPV6_SUBTREES
1372 if (FIB6_SUBTREE(pn
) == fn
) {
1373 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1378 if (pn
->left
== fn
) {
1382 if (pn
->right
== fn
) {
1384 w
->leaf
= w
->node
->leaf
;
1394 static int fib6_walk(struct fib6_walker_t
*w
)
1398 w
->state
= FWS_INIT
;
1401 fib6_walker_link(w
);
1402 res
= fib6_walk_continue(w
);
1404 fib6_walker_unlink(w
);
1408 static int fib6_clean_node(struct fib6_walker_t
*w
)
1411 struct rt6_info
*rt
;
1412 struct fib6_cleaner_t
*c
= container_of(w
, struct fib6_cleaner_t
, w
);
1413 struct nl_info info
= {
1417 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1418 res
= c
->func(rt
, c
->arg
);
1421 res
= fib6_del(rt
, &info
);
1424 printk(KERN_DEBUG
"fib6_clean_node: del failed: rt=%p@%p err=%d\n", rt
, rt
->rt6i_node
, res
);
1437 * Convenient frontend to tree walker.
1439 * func is called on each route.
1440 * It may return -1 -> delete this route.
1441 * 0 -> continue walking
1443 * prune==1 -> only immediate children of node (certainly,
1444 * ignoring pure split nodes) will be scanned.
1447 static void fib6_clean_tree(struct net
*net
, struct fib6_node
*root
,
1448 int (*func
)(struct rt6_info
*, void *arg
),
1449 int prune
, void *arg
)
1451 struct fib6_cleaner_t c
;
1454 c
.w
.func
= fib6_clean_node
;
1465 void fib6_clean_all_ro(struct net
*net
, int (*func
)(struct rt6_info
*, void *arg
),
1466 int prune
, void *arg
)
1468 struct fib6_table
*table
;
1469 struct hlist_node
*node
;
1470 struct hlist_head
*head
;
1474 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1475 head
= &net
->ipv6
.fib_table_hash
[h
];
1476 hlist_for_each_entry_rcu(table
, node
, head
, tb6_hlist
) {
1477 read_lock_bh(&table
->tb6_lock
);
1478 fib6_clean_tree(net
, &table
->tb6_root
,
1480 read_unlock_bh(&table
->tb6_lock
);
1485 void fib6_clean_all(struct net
*net
, int (*func
)(struct rt6_info
*, void *arg
),
1486 int prune
, void *arg
)
1488 struct fib6_table
*table
;
1489 struct hlist_node
*node
;
1490 struct hlist_head
*head
;
1494 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1495 head
= &net
->ipv6
.fib_table_hash
[h
];
1496 hlist_for_each_entry_rcu(table
, node
, head
, tb6_hlist
) {
1497 write_lock_bh(&table
->tb6_lock
);
1498 fib6_clean_tree(net
, &table
->tb6_root
,
1500 write_unlock_bh(&table
->tb6_lock
);
1506 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1508 if (rt
->rt6i_flags
& RTF_CACHE
) {
1509 RT6_TRACE("pruning clone %p\n", rt
);
1516 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
,
1517 struct rt6_info
*rt
)
1519 fib6_clean_tree(net
, fn
, fib6_prune_clone
, 1, rt
);
1523 * Garbage collection
1526 static struct fib6_gc_args
1532 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1534 unsigned long now
= jiffies
;
1537 * check addrconf expiration here.
1538 * Routes are expired even if they are in use.
1540 * Also age clones. Note, that clones are aged out
1541 * only if they are not in use now.
1544 if (rt
->rt6i_flags
& RTF_EXPIRES
&& rt
->dst
.expires
) {
1545 if (time_after(now
, rt
->dst
.expires
)) {
1546 RT6_TRACE("expiring %p\n", rt
);
1550 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1551 if (atomic_read(&rt
->dst
.__refcnt
) == 0 &&
1552 time_after_eq(now
, rt
->dst
.lastuse
+ gc_args
.timeout
)) {
1553 RT6_TRACE("aging clone %p\n", rt
);
1555 } else if ((rt
->rt6i_flags
& RTF_GATEWAY
) &&
1556 (!(dst_get_neighbour_noref_raw(&rt
->dst
)->flags
& NTF_ROUTER
))) {
1557 RT6_TRACE("purging route %p via non-router but gateway\n",
1567 static DEFINE_SPINLOCK(fib6_gc_lock
);
1569 void fib6_run_gc(unsigned long expires
, struct net
*net
)
1571 if (expires
!= ~0UL) {
1572 spin_lock_bh(&fib6_gc_lock
);
1573 gc_args
.timeout
= expires
? (int)expires
:
1574 net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1576 if (!spin_trylock_bh(&fib6_gc_lock
)) {
1577 mod_timer(&net
->ipv6
.ip6_fib_timer
, jiffies
+ HZ
);
1580 gc_args
.timeout
= net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1583 gc_args
.more
= icmp6_dst_gc();
1585 fib6_clean_all(net
, fib6_age
, 0, NULL
);
1588 mod_timer(&net
->ipv6
.ip6_fib_timer
,
1589 round_jiffies(jiffies
1590 + net
->ipv6
.sysctl
.ip6_rt_gc_interval
));
1592 del_timer(&net
->ipv6
.ip6_fib_timer
);
1593 spin_unlock_bh(&fib6_gc_lock
);
1596 static void fib6_gc_timer_cb(unsigned long arg
)
1598 fib6_run_gc(0, (struct net
*)arg
);
1601 static int __net_init
fib6_net_init(struct net
*net
)
1603 size_t size
= sizeof(struct hlist_head
) * FIB6_TABLE_HASHSZ
;
1605 setup_timer(&net
->ipv6
.ip6_fib_timer
, fib6_gc_timer_cb
, (unsigned long)net
);
1607 net
->ipv6
.rt6_stats
= kzalloc(sizeof(*net
->ipv6
.rt6_stats
), GFP_KERNEL
);
1608 if (!net
->ipv6
.rt6_stats
)
1611 /* Avoid false sharing : Use at least a full cache line */
1612 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1614 net
->ipv6
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1615 if (!net
->ipv6
.fib_table_hash
)
1618 net
->ipv6
.fib6_main_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_main_tbl
),
1620 if (!net
->ipv6
.fib6_main_tbl
)
1621 goto out_fib_table_hash
;
1623 net
->ipv6
.fib6_main_tbl
->tb6_id
= RT6_TABLE_MAIN
;
1624 net
->ipv6
.fib6_main_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1625 net
->ipv6
.fib6_main_tbl
->tb6_root
.fn_flags
=
1626 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1628 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1629 net
->ipv6
.fib6_local_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_local_tbl
),
1631 if (!net
->ipv6
.fib6_local_tbl
)
1632 goto out_fib6_main_tbl
;
1633 net
->ipv6
.fib6_local_tbl
->tb6_id
= RT6_TABLE_LOCAL
;
1634 net
->ipv6
.fib6_local_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1635 net
->ipv6
.fib6_local_tbl
->tb6_root
.fn_flags
=
1636 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1638 fib6_tables_init(net
);
1642 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1644 kfree(net
->ipv6
.fib6_main_tbl
);
1647 kfree(net
->ipv6
.fib_table_hash
);
1649 kfree(net
->ipv6
.rt6_stats
);
1654 static void fib6_net_exit(struct net
*net
)
1656 rt6_ifdown(net
, NULL
);
1657 del_timer_sync(&net
->ipv6
.ip6_fib_timer
);
1659 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1660 kfree(net
->ipv6
.fib6_local_tbl
);
1662 kfree(net
->ipv6
.fib6_main_tbl
);
1663 kfree(net
->ipv6
.fib_table_hash
);
1664 kfree(net
->ipv6
.rt6_stats
);
1667 static struct pernet_operations fib6_net_ops
= {
1668 .init
= fib6_net_init
,
1669 .exit
= fib6_net_exit
,
1672 int __init
fib6_init(void)
1676 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1677 sizeof(struct fib6_node
),
1678 0, SLAB_HWCACHE_ALIGN
,
1680 if (!fib6_node_kmem
)
1683 ret
= register_pernet_subsys(&fib6_net_ops
);
1685 goto out_kmem_cache_create
;
1687 ret
= __rtnl_register(PF_INET6
, RTM_GETROUTE
, NULL
, inet6_dump_fib
,
1690 goto out_unregister_subsys
;
1694 out_unregister_subsys
:
1695 unregister_pernet_subsys(&fib6_net_ops
);
1696 out_kmem_cache_create
:
1697 kmem_cache_destroy(fib6_node_kmem
);
1701 void fib6_gc_cleanup(void)
1703 unregister_pernet_subsys(&fib6_net_ops
);
1704 kmem_cache_destroy(fib6_node_kmem
);