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.
14 * Yuji SEKIYA @USAGI: Support default route on router node;
15 * remove ip6_null_entry from the top of
17 * Ville Nuorvala: Fixed routing subtrees.
20 #define pr_fmt(fmt) "IPv6: " fmt
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>
30 #include <linux/slab.h>
33 #include <net/ndisc.h>
34 #include <net/addrconf.h>
36 #include <net/ip6_fib.h>
37 #include <net/ip6_route.h>
42 #define RT6_TRACE(x...) pr_debug(x)
44 #define RT6_TRACE(x...) do { ; } while (0)
47 static struct kmem_cache
*fib6_node_kmem __read_mostly
;
52 int (*func
)(struct rt6_info
*, void *arg
);
57 static DEFINE_RWLOCK(fib6_walker_lock
);
59 #ifdef CONFIG_IPV6_SUBTREES
60 #define FWS_INIT FWS_S
62 #define FWS_INIT FWS_L
65 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
);
66 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
);
67 static struct fib6_node
*fib6_repair_tree(struct net
*net
, struct fib6_node
*fn
);
68 static int fib6_walk(struct fib6_walker
*w
);
69 static int fib6_walk_continue(struct fib6_walker
*w
);
72 * A routing update causes an increase of the serial number on the
73 * affected subtree. This allows for cached routes to be asynchronously
74 * tested when modifications are made to the destination cache as a
75 * result of redirects, path MTU changes, etc.
78 static void fib6_gc_timer_cb(unsigned long arg
);
80 static LIST_HEAD(fib6_walkers
);
81 #define FOR_WALKERS(w) list_for_each_entry(w, &fib6_walkers, lh)
83 static void fib6_walker_link(struct fib6_walker
*w
)
85 write_lock_bh(&fib6_walker_lock
);
86 list_add(&w
->lh
, &fib6_walkers
);
87 write_unlock_bh(&fib6_walker_lock
);
90 static void fib6_walker_unlink(struct fib6_walker
*w
)
92 write_lock_bh(&fib6_walker_lock
);
94 write_unlock_bh(&fib6_walker_lock
);
97 static int fib6_new_sernum(struct net
*net
)
102 old
= atomic_read(&net
->ipv6
.fib6_sernum
);
103 new = old
< INT_MAX
? old
+ 1 : 1;
104 } while (atomic_cmpxchg(&net
->ipv6
.fib6_sernum
,
110 FIB6_NO_SERNUM_CHANGE
= 0,
114 * Auxiliary address test functions for the radix tree.
116 * These assume a 32bit processor (although it will work on
123 #if defined(__LITTLE_ENDIAN)
124 # define BITOP_BE32_SWIZZLE (0x1F & ~7)
126 # define BITOP_BE32_SWIZZLE 0
129 static __be32
addr_bit_set(const void *token
, int fn_bit
)
131 const __be32
*addr
= token
;
134 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
135 * is optimized version of
136 * htonl(1 << ((~fn_bit)&0x1F))
137 * See include/asm-generic/bitops/le.h.
139 return (__force __be32
)(1 << ((~fn_bit
^ BITOP_BE32_SWIZZLE
) & 0x1f)) &
143 static struct fib6_node
*node_alloc(void)
145 struct fib6_node
*fn
;
147 fn
= kmem_cache_zalloc(fib6_node_kmem
, GFP_ATOMIC
);
152 static void node_free(struct fib6_node
*fn
)
154 kmem_cache_free(fib6_node_kmem
, fn
);
157 static void rt6_release(struct rt6_info
*rt
)
159 if (atomic_dec_and_test(&rt
->rt6i_ref
))
163 static void fib6_link_table(struct net
*net
, struct fib6_table
*tb
)
168 * Initialize table lock at a single place to give lockdep a key,
169 * tables aren't visible prior to being linked to the list.
171 rwlock_init(&tb
->tb6_lock
);
173 h
= tb
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1);
176 * No protection necessary, this is the only list mutatation
177 * operation, tables never disappear once they exist.
179 hlist_add_head_rcu(&tb
->tb6_hlist
, &net
->ipv6
.fib_table_hash
[h
]);
182 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
184 static struct fib6_table
*fib6_alloc_table(struct net
*net
, u32 id
)
186 struct fib6_table
*table
;
188 table
= kzalloc(sizeof(*table
), GFP_ATOMIC
);
191 table
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
192 table
->tb6_root
.fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
193 inet_peer_base_init(&table
->tb6_peers
);
199 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
201 struct fib6_table
*tb
;
205 tb
= fib6_get_table(net
, id
);
209 tb
= fib6_alloc_table(net
, id
);
211 fib6_link_table(net
, tb
);
216 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
218 struct fib6_table
*tb
;
219 struct hlist_head
*head
;
224 h
= id
& (FIB6_TABLE_HASHSZ
- 1);
226 head
= &net
->ipv6
.fib_table_hash
[h
];
227 hlist_for_each_entry_rcu(tb
, head
, tb6_hlist
) {
228 if (tb
->tb6_id
== id
) {
238 static void __net_init
fib6_tables_init(struct net
*net
)
240 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
241 fib6_link_table(net
, net
->ipv6
.fib6_local_tbl
);
245 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
247 return fib6_get_table(net
, id
);
250 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
252 return net
->ipv6
.fib6_main_tbl
;
255 struct dst_entry
*fib6_rule_lookup(struct net
*net
, struct flowi6
*fl6
,
256 int flags
, pol_lookup_t lookup
)
258 return (struct dst_entry
*) lookup(net
, net
->ipv6
.fib6_main_tbl
, fl6
, flags
);
261 static void __net_init
fib6_tables_init(struct net
*net
)
263 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
268 static int fib6_dump_node(struct fib6_walker
*w
)
273 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
274 res
= rt6_dump_route(rt
, w
->args
);
276 /* Frame is full, suspend walking */
286 static void fib6_dump_end(struct netlink_callback
*cb
)
288 struct fib6_walker
*w
= (void *)cb
->args
[2];
293 fib6_walker_unlink(w
);
298 cb
->done
= (void *)cb
->args
[3];
302 static int fib6_dump_done(struct netlink_callback
*cb
)
305 return cb
->done
? cb
->done(cb
) : 0;
308 static int fib6_dump_table(struct fib6_table
*table
, struct sk_buff
*skb
,
309 struct netlink_callback
*cb
)
311 struct fib6_walker
*w
;
314 w
= (void *)cb
->args
[2];
315 w
->root
= &table
->tb6_root
;
317 if (cb
->args
[4] == 0) {
321 read_lock_bh(&table
->tb6_lock
);
323 read_unlock_bh(&table
->tb6_lock
);
326 cb
->args
[5] = w
->root
->fn_sernum
;
329 if (cb
->args
[5] != w
->root
->fn_sernum
) {
330 /* Begin at the root if the tree changed */
331 cb
->args
[5] = w
->root
->fn_sernum
;
338 read_lock_bh(&table
->tb6_lock
);
339 res
= fib6_walk_continue(w
);
340 read_unlock_bh(&table
->tb6_lock
);
342 fib6_walker_unlink(w
);
350 static int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
352 struct net
*net
= sock_net(skb
->sk
);
354 unsigned int e
= 0, s_e
;
355 struct rt6_rtnl_dump_arg arg
;
356 struct fib6_walker
*w
;
357 struct fib6_table
*tb
;
358 struct hlist_head
*head
;
364 w
= (void *)cb
->args
[2];
368 * 1. hook callback destructor.
370 cb
->args
[3] = (long)cb
->done
;
371 cb
->done
= fib6_dump_done
;
374 * 2. allocate and initialize walker.
376 w
= kzalloc(sizeof(*w
), GFP_ATOMIC
);
379 w
->func
= fib6_dump_node
;
380 cb
->args
[2] = (long)w
;
389 for (h
= s_h
; h
< FIB6_TABLE_HASHSZ
; h
++, s_e
= 0) {
391 head
= &net
->ipv6
.fib_table_hash
[h
];
392 hlist_for_each_entry_rcu(tb
, head
, tb6_hlist
) {
395 res
= fib6_dump_table(tb
, skb
, cb
);
407 res
= res
< 0 ? res
: skb
->len
;
416 * return the appropriate node for a routing tree "add" operation
417 * by either creating and inserting or by returning an existing
421 static struct fib6_node
*fib6_add_1(struct fib6_node
*root
,
422 struct in6_addr
*addr
, int plen
,
423 int offset
, int allow_create
,
424 int replace_required
, int sernum
)
426 struct fib6_node
*fn
, *in
, *ln
;
427 struct fib6_node
*pn
= NULL
;
432 RT6_TRACE("fib6_add_1\n");
434 /* insert node in tree */
439 key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
444 if (plen
< fn
->fn_bit
||
445 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
)) {
447 if (replace_required
) {
448 pr_warn("Can't replace route, no match found\n");
449 return ERR_PTR(-ENOENT
);
451 pr_warn("NLM_F_CREATE should be set when creating new route\n");
460 if (plen
== fn
->fn_bit
) {
461 /* clean up an intermediate node */
462 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
463 rt6_release(fn
->leaf
);
467 fn
->fn_sernum
= sernum
;
473 * We have more bits to go
476 /* Try to walk down on tree. */
477 fn
->fn_sernum
= sernum
;
478 dir
= addr_bit_set(addr
, fn
->fn_bit
);
480 fn
= dir
? fn
->right
: fn
->left
;
484 /* We should not create new node because
485 * NLM_F_REPLACE was specified without NLM_F_CREATE
486 * I assume it is safe to require NLM_F_CREATE when
487 * REPLACE flag is used! Later we may want to remove the
488 * check for replace_required, because according
489 * to netlink specification, NLM_F_CREATE
490 * MUST be specified if new route is created.
491 * That would keep IPv6 consistent with IPv4
493 if (replace_required
) {
494 pr_warn("Can't replace route, no match found\n");
495 return ERR_PTR(-ENOENT
);
497 pr_warn("NLM_F_CREATE should be set when creating new route\n");
500 * We walked to the bottom of tree.
501 * Create new leaf node without children.
507 return ERR_PTR(-ENOMEM
);
511 ln
->fn_sernum
= sernum
;
523 * split since we don't have a common prefix anymore or
524 * we have a less significant route.
525 * we've to insert an intermediate node on the list
526 * this new node will point to the one we need to create
532 /* find 1st bit in difference between the 2 addrs.
534 See comment in __ipv6_addr_diff: bit may be an invalid value,
535 but if it is >= plen, the value is ignored in any case.
538 bit
= __ipv6_addr_diff(addr
, &key
->addr
, sizeof(*addr
));
543 * (new leaf node)[ln] (old node)[fn]
554 return ERR_PTR(-ENOMEM
);
558 * new intermediate node.
560 * be off since that an address that chooses one of
561 * the branches would not match less specific routes
562 * in the other branch
569 atomic_inc(&in
->leaf
->rt6i_ref
);
571 in
->fn_sernum
= sernum
;
573 /* update parent pointer */
584 ln
->fn_sernum
= sernum
;
586 if (addr_bit_set(addr
, bit
)) {
593 } else { /* plen <= bit */
596 * (new leaf node)[ln]
598 * (old node)[fn] NULL
604 return ERR_PTR(-ENOMEM
);
610 ln
->fn_sernum
= sernum
;
617 if (addr_bit_set(&key
->addr
, plen
))
627 static bool rt6_qualify_for_ecmp(struct rt6_info
*rt
)
629 return (rt
->rt6i_flags
& (RTF_GATEWAY
|RTF_ADDRCONF
|RTF_DYNAMIC
)) ==
633 static int fib6_commit_metrics(struct dst_entry
*dst
,
634 struct nlattr
*mx
, int mx_len
)
640 if (dst
->flags
& DST_HOST
) {
641 mp
= dst_metrics_write_ptr(dst
);
643 mp
= kzalloc(sizeof(u32
) * RTAX_MAX
, GFP_ATOMIC
);
646 dst_init_metrics(dst
, mp
, 0);
649 nla_for_each_attr(nla
, mx
, mx_len
, remaining
) {
650 int type
= nla_type(nla
);
656 mp
[type
- 1] = nla_get_u32(nla
);
663 * Insert routing information in a node.
666 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
667 struct nl_info
*info
, struct nlattr
*mx
, int mx_len
)
669 struct rt6_info
*iter
= NULL
;
670 struct rt6_info
**ins
;
671 int replace
= (info
->nlh
&&
672 (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
));
673 int add
= (!info
->nlh
||
674 (info
->nlh
->nlmsg_flags
& NLM_F_CREATE
));
676 bool rt_can_ecmp
= rt6_qualify_for_ecmp(rt
);
681 for (iter
= fn
->leaf
; iter
; iter
= iter
->dst
.rt6_next
) {
683 * Search for duplicates
686 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
688 * Same priority level
691 (info
->nlh
->nlmsg_flags
& NLM_F_EXCL
))
698 if (iter
->dst
.dev
== rt
->dst
.dev
&&
699 iter
->rt6i_idev
== rt
->rt6i_idev
&&
700 ipv6_addr_equal(&iter
->rt6i_gateway
,
701 &rt
->rt6i_gateway
)) {
702 if (rt
->rt6i_nsiblings
)
703 rt
->rt6i_nsiblings
= 0;
704 if (!(iter
->rt6i_flags
& RTF_EXPIRES
))
706 if (!(rt
->rt6i_flags
& RTF_EXPIRES
))
707 rt6_clean_expires(iter
);
709 rt6_set_expires(iter
, rt
->dst
.expires
);
712 /* If we have the same destination and the same metric,
713 * but not the same gateway, then the route we try to
714 * add is sibling to this route, increment our counter
715 * of siblings, and later we will add our route to the
717 * Only static routes (which don't have flag
718 * RTF_EXPIRES) are used for ECMPv6.
720 * To avoid long list, we only had siblings if the
721 * route have a gateway.
724 rt6_qualify_for_ecmp(iter
))
725 rt
->rt6i_nsiblings
++;
728 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
731 ins
= &iter
->dst
.rt6_next
;
734 /* Reset round-robin state, if necessary */
735 if (ins
== &fn
->leaf
)
738 /* Link this route to others same route. */
739 if (rt
->rt6i_nsiblings
) {
740 unsigned int rt6i_nsiblings
;
741 struct rt6_info
*sibling
, *temp_sibling
;
743 /* Find the first route that have the same metric */
746 if (sibling
->rt6i_metric
== rt
->rt6i_metric
&&
747 rt6_qualify_for_ecmp(sibling
)) {
748 list_add_tail(&rt
->rt6i_siblings
,
749 &sibling
->rt6i_siblings
);
752 sibling
= sibling
->dst
.rt6_next
;
754 /* For each sibling in the list, increment the counter of
755 * siblings. BUG() if counters does not match, list of siblings
759 list_for_each_entry_safe(sibling
, temp_sibling
,
760 &rt
->rt6i_siblings
, rt6i_siblings
) {
761 sibling
->rt6i_nsiblings
++;
762 BUG_ON(sibling
->rt6i_nsiblings
!= rt
->rt6i_nsiblings
);
765 BUG_ON(rt6i_nsiblings
!= rt
->rt6i_nsiblings
);
773 pr_warn("NLM_F_CREATE should be set when creating new route\n");
777 err
= fib6_commit_metrics(&rt
->dst
, mx
, mx_len
);
781 rt
->dst
.rt6_next
= iter
;
784 atomic_inc(&rt
->rt6i_ref
);
785 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
);
786 info
->nl_net
->ipv6
.rt6_stats
->fib_rt_entries
++;
788 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
789 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
790 fn
->fn_flags
|= RTN_RTINFO
;
797 pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
801 err
= fib6_commit_metrics(&rt
->dst
, mx
, mx_len
);
807 rt
->dst
.rt6_next
= iter
->dst
.rt6_next
;
808 atomic_inc(&rt
->rt6i_ref
);
809 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
);
811 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
812 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
813 fn
->fn_flags
|= RTN_RTINFO
;
820 static void fib6_start_gc(struct net
*net
, struct rt6_info
*rt
)
822 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
) &&
823 (rt
->rt6i_flags
& (RTF_EXPIRES
| RTF_CACHE
)))
824 mod_timer(&net
->ipv6
.ip6_fib_timer
,
825 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
828 void fib6_force_start_gc(struct net
*net
)
830 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
))
831 mod_timer(&net
->ipv6
.ip6_fib_timer
,
832 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
836 * Add routing information to the routing tree.
837 * <destination addr>/<source addr>
838 * with source addr info in sub-trees
841 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
, struct nl_info
*info
,
842 struct nlattr
*mx
, int mx_len
)
844 struct fib6_node
*fn
, *pn
= NULL
;
846 int allow_create
= 1;
847 int replace_required
= 0;
848 int sernum
= fib6_new_sernum(info
->nl_net
);
851 if (!(info
->nlh
->nlmsg_flags
& NLM_F_CREATE
))
853 if (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
)
854 replace_required
= 1;
856 if (!allow_create
&& !replace_required
)
857 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
859 fn
= fib6_add_1(root
, &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
,
860 offsetof(struct rt6_info
, rt6i_dst
), allow_create
,
861 replace_required
, sernum
);
870 #ifdef CONFIG_IPV6_SUBTREES
871 if (rt
->rt6i_src
.plen
) {
872 struct fib6_node
*sn
;
875 struct fib6_node
*sfn
;
887 /* Create subtree root node */
892 sfn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
893 atomic_inc(&info
->nl_net
->ipv6
.ip6_null_entry
->rt6i_ref
);
894 sfn
->fn_flags
= RTN_ROOT
;
895 sfn
->fn_sernum
= sernum
;
897 /* Now add the first leaf node to new subtree */
899 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
901 offsetof(struct rt6_info
, rt6i_src
),
902 allow_create
, replace_required
, sernum
);
905 /* If it is failed, discard just allocated
906 root, and then (in st_failure) stale node
914 /* Now link new subtree to main tree */
918 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
920 offsetof(struct rt6_info
, rt6i_src
),
921 allow_create
, replace_required
, sernum
);
931 atomic_inc(&rt
->rt6i_ref
);
937 err
= fib6_add_rt2node(fn
, rt
, info
, mx
, mx_len
);
939 fib6_start_gc(info
->nl_net
, rt
);
940 if (!(rt
->rt6i_flags
& RTF_CACHE
))
941 fib6_prune_clones(info
->nl_net
, pn
);
946 #ifdef CONFIG_IPV6_SUBTREES
948 * If fib6_add_1 has cleared the old leaf pointer in the
949 * super-tree leaf node we have to find a new one for it.
951 if (pn
!= fn
&& pn
->leaf
== rt
) {
953 atomic_dec(&rt
->rt6i_ref
);
955 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
956 pn
->leaf
= fib6_find_prefix(info
->nl_net
, pn
);
959 WARN_ON(pn
->leaf
== NULL
);
960 pn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
963 atomic_inc(&pn
->leaf
->rt6i_ref
);
970 #ifdef CONFIG_IPV6_SUBTREES
971 /* Subtree creation failed, probably main tree node
972 is orphan. If it is, shoot it.
975 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
976 fib6_repair_tree(info
->nl_net
, fn
);
983 * Routing tree lookup
988 int offset
; /* key offset on rt6_info */
989 const struct in6_addr
*addr
; /* search key */
992 static struct fib6_node
*fib6_lookup_1(struct fib6_node
*root
,
993 struct lookup_args
*args
)
995 struct fib6_node
*fn
;
998 if (unlikely(args
->offset
== 0))
1008 struct fib6_node
*next
;
1010 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
1012 next
= dir
? fn
->right
: fn
->left
;
1022 if (FIB6_SUBTREE(fn
) || fn
->fn_flags
& RTN_RTINFO
) {
1025 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
1028 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
)) {
1029 #ifdef CONFIG_IPV6_SUBTREES
1031 struct fib6_node
*sfn
;
1032 sfn
= fib6_lookup_1(fn
->subtree
,
1039 if (fn
->fn_flags
& RTN_RTINFO
)
1043 #ifdef CONFIG_IPV6_SUBTREES
1046 if (fn
->fn_flags
& RTN_ROOT
)
1055 struct fib6_node
*fib6_lookup(struct fib6_node
*root
, const struct in6_addr
*daddr
,
1056 const struct in6_addr
*saddr
)
1058 struct fib6_node
*fn
;
1059 struct lookup_args args
[] = {
1061 .offset
= offsetof(struct rt6_info
, rt6i_dst
),
1064 #ifdef CONFIG_IPV6_SUBTREES
1066 .offset
= offsetof(struct rt6_info
, rt6i_src
),
1071 .offset
= 0, /* sentinel */
1075 fn
= fib6_lookup_1(root
, daddr
? args
: args
+ 1);
1076 if (!fn
|| fn
->fn_flags
& RTN_TL_ROOT
)
1083 * Get node with specified destination prefix (and source prefix,
1084 * if subtrees are used)
1088 static struct fib6_node
*fib6_locate_1(struct fib6_node
*root
,
1089 const struct in6_addr
*addr
,
1090 int plen
, int offset
)
1092 struct fib6_node
*fn
;
1094 for (fn
= root
; fn
; ) {
1095 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
1100 if (plen
< fn
->fn_bit
||
1101 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
1104 if (plen
== fn
->fn_bit
)
1108 * We have more bits to go
1110 if (addr_bit_set(addr
, fn
->fn_bit
))
1118 struct fib6_node
*fib6_locate(struct fib6_node
*root
,
1119 const struct in6_addr
*daddr
, int dst_len
,
1120 const struct in6_addr
*saddr
, int src_len
)
1122 struct fib6_node
*fn
;
1124 fn
= fib6_locate_1(root
, daddr
, dst_len
,
1125 offsetof(struct rt6_info
, rt6i_dst
));
1127 #ifdef CONFIG_IPV6_SUBTREES
1129 WARN_ON(saddr
== NULL
);
1130 if (fn
&& fn
->subtree
)
1131 fn
= fib6_locate_1(fn
->subtree
, saddr
, src_len
,
1132 offsetof(struct rt6_info
, rt6i_src
));
1136 if (fn
&& fn
->fn_flags
& RTN_RTINFO
)
1148 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
)
1150 if (fn
->fn_flags
& RTN_ROOT
)
1151 return net
->ipv6
.ip6_null_entry
;
1155 return fn
->left
->leaf
;
1157 return fn
->right
->leaf
;
1159 fn
= FIB6_SUBTREE(fn
);
1165 * Called to trim the tree of intermediate nodes when possible. "fn"
1166 * is the node we want to try and remove.
1169 static struct fib6_node
*fib6_repair_tree(struct net
*net
,
1170 struct fib6_node
*fn
)
1174 struct fib6_node
*child
, *pn
;
1175 struct fib6_walker
*w
;
1179 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
1182 WARN_ON(fn
->fn_flags
& RTN_RTINFO
);
1183 WARN_ON(fn
->fn_flags
& RTN_TL_ROOT
);
1184 WARN_ON(fn
->leaf
!= NULL
);
1189 child
= fn
->right
, children
|= 1;
1191 child
= fn
->left
, children
|= 2;
1193 if (children
== 3 || FIB6_SUBTREE(fn
)
1194 #ifdef CONFIG_IPV6_SUBTREES
1195 /* Subtree root (i.e. fn) may have one child */
1196 || (children
&& fn
->fn_flags
& RTN_ROOT
)
1199 fn
->leaf
= fib6_find_prefix(net
, fn
);
1203 fn
->leaf
= net
->ipv6
.ip6_null_entry
;
1206 atomic_inc(&fn
->leaf
->rt6i_ref
);
1211 #ifdef CONFIG_IPV6_SUBTREES
1212 if (FIB6_SUBTREE(pn
) == fn
) {
1213 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1214 FIB6_SUBTREE(pn
) = NULL
;
1217 WARN_ON(fn
->fn_flags
& RTN_ROOT
);
1219 if (pn
->right
== fn
)
1221 else if (pn
->left
== fn
)
1230 #ifdef CONFIG_IPV6_SUBTREES
1234 read_lock(&fib6_walker_lock
);
1237 if (w
->root
== fn
) {
1238 w
->root
= w
->node
= NULL
;
1239 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1240 } else if (w
->node
== fn
) {
1241 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1246 if (w
->root
== fn
) {
1248 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1250 if (w
->node
== fn
) {
1253 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1254 w
->state
= w
->state
>= FWS_R
? FWS_U
: FWS_INIT
;
1256 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1257 w
->state
= w
->state
>= FWS_C
? FWS_U
: FWS_INIT
;
1262 read_unlock(&fib6_walker_lock
);
1265 if (pn
->fn_flags
& RTN_RTINFO
|| FIB6_SUBTREE(pn
))
1268 rt6_release(pn
->leaf
);
1274 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1275 struct nl_info
*info
)
1277 struct fib6_walker
*w
;
1278 struct rt6_info
*rt
= *rtp
;
1279 struct net
*net
= info
->nl_net
;
1281 RT6_TRACE("fib6_del_route\n");
1284 *rtp
= rt
->dst
.rt6_next
;
1285 rt
->rt6i_node
= NULL
;
1286 net
->ipv6
.rt6_stats
->fib_rt_entries
--;
1287 net
->ipv6
.rt6_stats
->fib_discarded_routes
++;
1289 /* Reset round-robin state, if necessary */
1290 if (fn
->rr_ptr
== rt
)
1293 /* Remove this entry from other siblings */
1294 if (rt
->rt6i_nsiblings
) {
1295 struct rt6_info
*sibling
, *next_sibling
;
1297 list_for_each_entry_safe(sibling
, next_sibling
,
1298 &rt
->rt6i_siblings
, rt6i_siblings
)
1299 sibling
->rt6i_nsiblings
--;
1300 rt
->rt6i_nsiblings
= 0;
1301 list_del_init(&rt
->rt6i_siblings
);
1304 /* Adjust walkers */
1305 read_lock(&fib6_walker_lock
);
1307 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1308 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1309 w
->leaf
= rt
->dst
.rt6_next
;
1314 read_unlock(&fib6_walker_lock
);
1316 rt
->dst
.rt6_next
= NULL
;
1318 /* If it was last route, expunge its radix tree node */
1320 fn
->fn_flags
&= ~RTN_RTINFO
;
1321 net
->ipv6
.rt6_stats
->fib_route_nodes
--;
1322 fn
= fib6_repair_tree(net
, fn
);
1325 if (atomic_read(&rt
->rt6i_ref
) != 1) {
1326 /* This route is used as dummy address holder in some split
1327 * nodes. It is not leaked, but it still holds other resources,
1328 * which must be released in time. So, scan ascendant nodes
1329 * and replace dummy references to this route with references
1330 * to still alive ones.
1333 if (!(fn
->fn_flags
& RTN_RTINFO
) && fn
->leaf
== rt
) {
1334 fn
->leaf
= fib6_find_prefix(net
, fn
);
1335 atomic_inc(&fn
->leaf
->rt6i_ref
);
1340 /* No more references are possible at this point. */
1341 BUG_ON(atomic_read(&rt
->rt6i_ref
) != 1);
1344 inet6_rt_notify(RTM_DELROUTE
, rt
, info
);
1348 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1350 struct net
*net
= info
->nl_net
;
1351 struct fib6_node
*fn
= rt
->rt6i_node
;
1352 struct rt6_info
**rtp
;
1355 if (rt
->dst
.obsolete
> 0) {
1356 WARN_ON(fn
!= NULL
);
1360 if (!fn
|| rt
== net
->ipv6
.ip6_null_entry
)
1363 WARN_ON(!(fn
->fn_flags
& RTN_RTINFO
));
1365 if (!(rt
->rt6i_flags
& RTF_CACHE
)) {
1366 struct fib6_node
*pn
= fn
;
1367 #ifdef CONFIG_IPV6_SUBTREES
1368 /* clones of this route might be in another subtree */
1369 if (rt
->rt6i_src
.plen
) {
1370 while (!(pn
->fn_flags
& RTN_ROOT
))
1375 fib6_prune_clones(info
->nl_net
, pn
);
1379 * Walk the leaf entries looking for ourself
1382 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->dst
.rt6_next
) {
1384 fib6_del_route(fn
, rtp
, info
);
1392 * Tree traversal function.
1394 * Certainly, it is not interrupt safe.
1395 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1396 * It means, that we can modify tree during walking
1397 * and use this function for garbage collection, clone pruning,
1398 * cleaning tree when a device goes down etc. etc.
1400 * It guarantees that every node will be traversed,
1401 * and that it will be traversed only once.
1403 * Callback function w->func may return:
1404 * 0 -> continue walking.
1405 * positive value -> walking is suspended (used by tree dumps,
1406 * and probably by gc, if it will be split to several slices)
1407 * negative value -> terminate walking.
1409 * The function itself returns:
1410 * 0 -> walk is complete.
1411 * >0 -> walk is incomplete (i.e. suspended)
1412 * <0 -> walk is terminated by an error.
1415 static int fib6_walk_continue(struct fib6_walker
*w
)
1417 struct fib6_node
*fn
, *pn
;
1424 if (w
->prune
&& fn
!= w
->root
&&
1425 fn
->fn_flags
& RTN_RTINFO
&& w
->state
< FWS_C
) {
1430 #ifdef CONFIG_IPV6_SUBTREES
1432 if (FIB6_SUBTREE(fn
)) {
1433 w
->node
= FIB6_SUBTREE(fn
);
1441 w
->state
= FWS_INIT
;
1447 w
->node
= fn
->right
;
1448 w
->state
= FWS_INIT
;
1454 if (w
->leaf
&& fn
->fn_flags
& RTN_RTINFO
) {
1476 #ifdef CONFIG_IPV6_SUBTREES
1477 if (FIB6_SUBTREE(pn
) == fn
) {
1478 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1483 if (pn
->left
== fn
) {
1487 if (pn
->right
== fn
) {
1489 w
->leaf
= w
->node
->leaf
;
1499 static int fib6_walk(struct fib6_walker
*w
)
1503 w
->state
= FWS_INIT
;
1506 fib6_walker_link(w
);
1507 res
= fib6_walk_continue(w
);
1509 fib6_walker_unlink(w
);
1513 static int fib6_clean_node(struct fib6_walker
*w
)
1516 struct rt6_info
*rt
;
1517 struct fib6_cleaner
*c
= container_of(w
, struct fib6_cleaner
, w
);
1518 struct nl_info info
= {
1522 if (c
->sernum
!= FIB6_NO_SERNUM_CHANGE
&&
1523 w
->node
->fn_sernum
!= c
->sernum
)
1524 w
->node
->fn_sernum
= c
->sernum
;
1527 WARN_ON_ONCE(c
->sernum
== FIB6_NO_SERNUM_CHANGE
);
1532 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1533 res
= c
->func(rt
, c
->arg
);
1536 res
= fib6_del(rt
, &info
);
1539 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
1540 __func__
, rt
, rt
->rt6i_node
, res
);
1553 * Convenient frontend to tree walker.
1555 * func is called on each route.
1556 * It may return -1 -> delete this route.
1557 * 0 -> continue walking
1559 * prune==1 -> only immediate children of node (certainly,
1560 * ignoring pure split nodes) will be scanned.
1563 static void fib6_clean_tree(struct net
*net
, struct fib6_node
*root
,
1564 int (*func
)(struct rt6_info
*, void *arg
),
1565 bool prune
, int sernum
, void *arg
)
1567 struct fib6_cleaner c
;
1570 c
.w
.func
= fib6_clean_node
;
1582 static void __fib6_clean_all(struct net
*net
,
1583 int (*func
)(struct rt6_info
*, void *),
1584 int sernum
, void *arg
)
1586 struct fib6_table
*table
;
1587 struct hlist_head
*head
;
1591 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1592 head
= &net
->ipv6
.fib_table_hash
[h
];
1593 hlist_for_each_entry_rcu(table
, head
, tb6_hlist
) {
1594 write_lock_bh(&table
->tb6_lock
);
1595 fib6_clean_tree(net
, &table
->tb6_root
,
1596 func
, false, sernum
, arg
);
1597 write_unlock_bh(&table
->tb6_lock
);
1603 void fib6_clean_all(struct net
*net
, int (*func
)(struct rt6_info
*, void *),
1606 __fib6_clean_all(net
, func
, FIB6_NO_SERNUM_CHANGE
, arg
);
1609 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1611 if (rt
->rt6i_flags
& RTF_CACHE
) {
1612 RT6_TRACE("pruning clone %p\n", rt
);
1619 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
)
1621 fib6_clean_tree(net
, fn
, fib6_prune_clone
, true,
1622 FIB6_NO_SERNUM_CHANGE
, NULL
);
1625 static void fib6_flush_trees(struct net
*net
)
1627 int new_sernum
= fib6_new_sernum(net
);
1629 __fib6_clean_all(net
, NULL
, new_sernum
, NULL
);
1633 * Garbage collection
1636 static struct fib6_gc_args
1642 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1644 unsigned long now
= jiffies
;
1647 * check addrconf expiration here.
1648 * Routes are expired even if they are in use.
1650 * Also age clones. Note, that clones are aged out
1651 * only if they are not in use now.
1654 if (rt
->rt6i_flags
& RTF_EXPIRES
&& rt
->dst
.expires
) {
1655 if (time_after(now
, rt
->dst
.expires
)) {
1656 RT6_TRACE("expiring %p\n", rt
);
1660 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1661 if (atomic_read(&rt
->dst
.__refcnt
) == 0 &&
1662 time_after_eq(now
, rt
->dst
.lastuse
+ gc_args
.timeout
)) {
1663 RT6_TRACE("aging clone %p\n", rt
);
1665 } else if (rt
->rt6i_flags
& RTF_GATEWAY
) {
1666 struct neighbour
*neigh
;
1667 __u8 neigh_flags
= 0;
1669 neigh
= dst_neigh_lookup(&rt
->dst
, &rt
->rt6i_gateway
);
1671 neigh_flags
= neigh
->flags
;
1672 neigh_release(neigh
);
1674 if (!(neigh_flags
& NTF_ROUTER
)) {
1675 RT6_TRACE("purging route %p via non-router but gateway\n",
1686 static DEFINE_SPINLOCK(fib6_gc_lock
);
1688 void fib6_run_gc(unsigned long expires
, struct net
*net
, bool force
)
1693 spin_lock_bh(&fib6_gc_lock
);
1694 } else if (!spin_trylock_bh(&fib6_gc_lock
)) {
1695 mod_timer(&net
->ipv6
.ip6_fib_timer
, jiffies
+ HZ
);
1698 gc_args
.timeout
= expires
? (int)expires
:
1699 net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1701 gc_args
.more
= icmp6_dst_gc();
1703 fib6_clean_all(net
, fib6_age
, NULL
);
1705 net
->ipv6
.ip6_rt_last_gc
= now
;
1708 mod_timer(&net
->ipv6
.ip6_fib_timer
,
1710 + net
->ipv6
.sysctl
.ip6_rt_gc_interval
));
1712 del_timer(&net
->ipv6
.ip6_fib_timer
);
1713 spin_unlock_bh(&fib6_gc_lock
);
1716 static void fib6_gc_timer_cb(unsigned long arg
)
1718 fib6_run_gc(0, (struct net
*)arg
, true);
1721 static int __net_init
fib6_net_init(struct net
*net
)
1723 size_t size
= sizeof(struct hlist_head
) * FIB6_TABLE_HASHSZ
;
1725 setup_timer(&net
->ipv6
.ip6_fib_timer
, fib6_gc_timer_cb
, (unsigned long)net
);
1727 net
->ipv6
.rt6_stats
= kzalloc(sizeof(*net
->ipv6
.rt6_stats
), GFP_KERNEL
);
1728 if (!net
->ipv6
.rt6_stats
)
1731 /* Avoid false sharing : Use at least a full cache line */
1732 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1734 net
->ipv6
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1735 if (!net
->ipv6
.fib_table_hash
)
1738 net
->ipv6
.fib6_main_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_main_tbl
),
1740 if (!net
->ipv6
.fib6_main_tbl
)
1741 goto out_fib_table_hash
;
1743 net
->ipv6
.fib6_main_tbl
->tb6_id
= RT6_TABLE_MAIN
;
1744 net
->ipv6
.fib6_main_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1745 net
->ipv6
.fib6_main_tbl
->tb6_root
.fn_flags
=
1746 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1747 inet_peer_base_init(&net
->ipv6
.fib6_main_tbl
->tb6_peers
);
1749 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1750 net
->ipv6
.fib6_local_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_local_tbl
),
1752 if (!net
->ipv6
.fib6_local_tbl
)
1753 goto out_fib6_main_tbl
;
1754 net
->ipv6
.fib6_local_tbl
->tb6_id
= RT6_TABLE_LOCAL
;
1755 net
->ipv6
.fib6_local_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1756 net
->ipv6
.fib6_local_tbl
->tb6_root
.fn_flags
=
1757 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1758 inet_peer_base_init(&net
->ipv6
.fib6_local_tbl
->tb6_peers
);
1760 fib6_tables_init(net
);
1764 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1766 kfree(net
->ipv6
.fib6_main_tbl
);
1769 kfree(net
->ipv6
.fib_table_hash
);
1771 kfree(net
->ipv6
.rt6_stats
);
1776 static void fib6_net_exit(struct net
*net
)
1778 rt6_ifdown(net
, NULL
);
1779 del_timer_sync(&net
->ipv6
.ip6_fib_timer
);
1781 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1782 inetpeer_invalidate_tree(&net
->ipv6
.fib6_local_tbl
->tb6_peers
);
1783 kfree(net
->ipv6
.fib6_local_tbl
);
1785 inetpeer_invalidate_tree(&net
->ipv6
.fib6_main_tbl
->tb6_peers
);
1786 kfree(net
->ipv6
.fib6_main_tbl
);
1787 kfree(net
->ipv6
.fib_table_hash
);
1788 kfree(net
->ipv6
.rt6_stats
);
1791 static struct pernet_operations fib6_net_ops
= {
1792 .init
= fib6_net_init
,
1793 .exit
= fib6_net_exit
,
1796 int __init
fib6_init(void)
1800 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1801 sizeof(struct fib6_node
),
1802 0, SLAB_HWCACHE_ALIGN
,
1804 if (!fib6_node_kmem
)
1807 ret
= register_pernet_subsys(&fib6_net_ops
);
1809 goto out_kmem_cache_create
;
1811 ret
= __rtnl_register(PF_INET6
, RTM_GETROUTE
, NULL
, inet6_dump_fib
,
1814 goto out_unregister_subsys
;
1816 __fib6_flush_trees
= fib6_flush_trees
;
1820 out_unregister_subsys
:
1821 unregister_pernet_subsys(&fib6_net_ops
);
1822 out_kmem_cache_create
:
1823 kmem_cache_destroy(fib6_node_kmem
);
1827 void fib6_gc_cleanup(void)
1829 unregister_pernet_subsys(&fib6_net_ops
);
1830 kmem_cache_destroy(fib6_node_kmem
);
1833 #ifdef CONFIG_PROC_FS
1835 struct ipv6_route_iter
{
1836 struct seq_net_private p
;
1837 struct fib6_walker w
;
1839 struct fib6_table
*tbl
;
1843 static int ipv6_route_seq_show(struct seq_file
*seq
, void *v
)
1845 struct rt6_info
*rt
= v
;
1846 struct ipv6_route_iter
*iter
= seq
->private;
1848 seq_printf(seq
, "%pi6 %02x ", &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
);
1850 #ifdef CONFIG_IPV6_SUBTREES
1851 seq_printf(seq
, "%pi6 %02x ", &rt
->rt6i_src
.addr
, rt
->rt6i_src
.plen
);
1853 seq_puts(seq
, "00000000000000000000000000000000 00 ");
1855 if (rt
->rt6i_flags
& RTF_GATEWAY
)
1856 seq_printf(seq
, "%pi6", &rt
->rt6i_gateway
);
1858 seq_puts(seq
, "00000000000000000000000000000000");
1860 seq_printf(seq
, " %08x %08x %08x %08x %8s\n",
1861 rt
->rt6i_metric
, atomic_read(&rt
->dst
.__refcnt
),
1862 rt
->dst
.__use
, rt
->rt6i_flags
,
1863 rt
->dst
.dev
? rt
->dst
.dev
->name
: "");
1864 iter
->w
.leaf
= NULL
;
1868 static int ipv6_route_yield(struct fib6_walker
*w
)
1870 struct ipv6_route_iter
*iter
= w
->args
;
1876 iter
->w
.leaf
= iter
->w
.leaf
->dst
.rt6_next
;
1878 if (!iter
->skip
&& iter
->w
.leaf
)
1880 } while (iter
->w
.leaf
);
1885 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter
*iter
)
1887 memset(&iter
->w
, 0, sizeof(iter
->w
));
1888 iter
->w
.func
= ipv6_route_yield
;
1889 iter
->w
.root
= &iter
->tbl
->tb6_root
;
1890 iter
->w
.state
= FWS_INIT
;
1891 iter
->w
.node
= iter
->w
.root
;
1892 iter
->w
.args
= iter
;
1893 iter
->sernum
= iter
->w
.root
->fn_sernum
;
1894 INIT_LIST_HEAD(&iter
->w
.lh
);
1895 fib6_walker_link(&iter
->w
);
1898 static struct fib6_table
*ipv6_route_seq_next_table(struct fib6_table
*tbl
,
1902 struct hlist_node
*node
;
1905 h
= (tbl
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1)) + 1;
1906 node
= rcu_dereference_bh(hlist_next_rcu(&tbl
->tb6_hlist
));
1912 while (!node
&& h
< FIB6_TABLE_HASHSZ
) {
1913 node
= rcu_dereference_bh(
1914 hlist_first_rcu(&net
->ipv6
.fib_table_hash
[h
++]));
1916 return hlist_entry_safe(node
, struct fib6_table
, tb6_hlist
);
1919 static void ipv6_route_check_sernum(struct ipv6_route_iter
*iter
)
1921 if (iter
->sernum
!= iter
->w
.root
->fn_sernum
) {
1922 iter
->sernum
= iter
->w
.root
->fn_sernum
;
1923 iter
->w
.state
= FWS_INIT
;
1924 iter
->w
.node
= iter
->w
.root
;
1925 WARN_ON(iter
->w
.skip
);
1926 iter
->w
.skip
= iter
->w
.count
;
1930 static void *ipv6_route_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
1934 struct net
*net
= seq_file_net(seq
);
1935 struct ipv6_route_iter
*iter
= seq
->private;
1940 n
= ((struct rt6_info
*)v
)->dst
.rt6_next
;
1947 ipv6_route_check_sernum(iter
);
1948 read_lock(&iter
->tbl
->tb6_lock
);
1949 r
= fib6_walk_continue(&iter
->w
);
1950 read_unlock(&iter
->tbl
->tb6_lock
);
1954 return iter
->w
.leaf
;
1956 fib6_walker_unlink(&iter
->w
);
1959 fib6_walker_unlink(&iter
->w
);
1961 iter
->tbl
= ipv6_route_seq_next_table(iter
->tbl
, net
);
1965 ipv6_route_seq_setup_walk(iter
);
1969 static void *ipv6_route_seq_start(struct seq_file
*seq
, loff_t
*pos
)
1972 struct net
*net
= seq_file_net(seq
);
1973 struct ipv6_route_iter
*iter
= seq
->private;
1976 iter
->tbl
= ipv6_route_seq_next_table(NULL
, net
);
1980 ipv6_route_seq_setup_walk(iter
);
1981 return ipv6_route_seq_next(seq
, NULL
, pos
);
1987 static bool ipv6_route_iter_active(struct ipv6_route_iter
*iter
)
1989 struct fib6_walker
*w
= &iter
->w
;
1990 return w
->node
&& !(w
->state
== FWS_U
&& w
->node
== w
->root
);
1993 static void ipv6_route_seq_stop(struct seq_file
*seq
, void *v
)
1996 struct ipv6_route_iter
*iter
= seq
->private;
1998 if (ipv6_route_iter_active(iter
))
1999 fib6_walker_unlink(&iter
->w
);
2001 rcu_read_unlock_bh();
2004 static const struct seq_operations ipv6_route_seq_ops
= {
2005 .start
= ipv6_route_seq_start
,
2006 .next
= ipv6_route_seq_next
,
2007 .stop
= ipv6_route_seq_stop
,
2008 .show
= ipv6_route_seq_show
2011 int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2013 return seq_open_net(inode
, file
, &ipv6_route_seq_ops
,
2014 sizeof(struct ipv6_route_iter
));
2017 #endif /* CONFIG_PROC_FS */