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>
35 #include <net/lwtunnel.h>
37 #include <net/ip6_fib.h>
38 #include <net/ip6_route.h>
43 #define RT6_TRACE(x...) pr_debug(x)
45 #define RT6_TRACE(x...) do { ; } while (0)
48 static struct kmem_cache
*fib6_node_kmem __read_mostly
;
53 int (*func
)(struct rt6_info
*, void *arg
);
58 #ifdef CONFIG_IPV6_SUBTREES
59 #define FWS_INIT FWS_S
61 #define FWS_INIT FWS_L
64 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
);
65 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
);
66 static struct fib6_node
*fib6_repair_tree(struct net
*net
, struct fib6_node
*fn
);
67 static int fib6_walk(struct net
*net
, struct fib6_walker
*w
);
68 static int fib6_walk_continue(struct fib6_walker
*w
);
71 * A routing update causes an increase of the serial number on the
72 * affected subtree. This allows for cached routes to be asynchronously
73 * tested when modifications are made to the destination cache as a
74 * result of redirects, path MTU changes, etc.
77 static void fib6_gc_timer_cb(unsigned long arg
);
79 #define FOR_WALKERS(net, w) \
80 list_for_each_entry(w, &(net)->ipv6.fib6_walkers, lh)
82 static void fib6_walker_link(struct net
*net
, struct fib6_walker
*w
)
84 write_lock_bh(&net
->ipv6
.fib6_walker_lock
);
85 list_add(&w
->lh
, &net
->ipv6
.fib6_walkers
);
86 write_unlock_bh(&net
->ipv6
.fib6_walker_lock
);
89 static void fib6_walker_unlink(struct net
*net
, struct fib6_walker
*w
)
91 write_lock_bh(&net
->ipv6
.fib6_walker_lock
);
93 write_unlock_bh(&net
->ipv6
.fib6_walker_lock
);
96 static int fib6_new_sernum(struct net
*net
)
101 old
= atomic_read(&net
->ipv6
.fib6_sernum
);
102 new = old
< INT_MAX
? old
+ 1 : 1;
103 } while (atomic_cmpxchg(&net
->ipv6
.fib6_sernum
,
109 FIB6_NO_SERNUM_CHANGE
= 0,
113 * Auxiliary address test functions for the radix tree.
115 * These assume a 32bit processor (although it will work on
122 #if defined(__LITTLE_ENDIAN)
123 # define BITOP_BE32_SWIZZLE (0x1F & ~7)
125 # define BITOP_BE32_SWIZZLE 0
128 static __be32
addr_bit_set(const void *token
, int fn_bit
)
130 const __be32
*addr
= token
;
133 * 1 << ((~fn_bit ^ BITOP_BE32_SWIZZLE) & 0x1f)
134 * is optimized version of
135 * htonl(1 << ((~fn_bit)&0x1F))
136 * See include/asm-generic/bitops/le.h.
138 return (__force __be32
)(1 << ((~fn_bit
^ BITOP_BE32_SWIZZLE
) & 0x1f)) &
142 static struct fib6_node
*node_alloc(void)
144 struct fib6_node
*fn
;
146 fn
= kmem_cache_zalloc(fib6_node_kmem
, GFP_ATOMIC
);
151 static void node_free(struct fib6_node
*fn
)
153 kmem_cache_free(fib6_node_kmem
, fn
);
156 static void rt6_rcu_free(struct rt6_info
*rt
)
158 call_rcu(&rt
->dst
.rcu_head
, dst_rcu_free
);
161 static void rt6_free_pcpu(struct rt6_info
*non_pcpu_rt
)
165 if (!non_pcpu_rt
->rt6i_pcpu
)
168 for_each_possible_cpu(cpu
) {
169 struct rt6_info
**ppcpu_rt
;
170 struct rt6_info
*pcpu_rt
;
172 ppcpu_rt
= per_cpu_ptr(non_pcpu_rt
->rt6i_pcpu
, cpu
);
175 rt6_rcu_free(pcpu_rt
);
180 non_pcpu_rt
->rt6i_pcpu
= NULL
;
183 static void rt6_release(struct rt6_info
*rt
)
185 if (atomic_dec_and_test(&rt
->rt6i_ref
)) {
191 static void fib6_link_table(struct net
*net
, struct fib6_table
*tb
)
196 * Initialize table lock at a single place to give lockdep a key,
197 * tables aren't visible prior to being linked to the list.
199 rwlock_init(&tb
->tb6_lock
);
201 h
= tb
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1);
204 * No protection necessary, this is the only list mutatation
205 * operation, tables never disappear once they exist.
207 hlist_add_head_rcu(&tb
->tb6_hlist
, &net
->ipv6
.fib_table_hash
[h
]);
210 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
212 static struct fib6_table
*fib6_alloc_table(struct net
*net
, u32 id
)
214 struct fib6_table
*table
;
216 table
= kzalloc(sizeof(*table
), GFP_ATOMIC
);
219 table
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
220 table
->tb6_root
.fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
221 inet_peer_base_init(&table
->tb6_peers
);
227 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
229 struct fib6_table
*tb
;
233 tb
= fib6_get_table(net
, id
);
237 tb
= fib6_alloc_table(net
, id
);
239 fib6_link_table(net
, tb
);
244 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
246 struct fib6_table
*tb
;
247 struct hlist_head
*head
;
252 h
= id
& (FIB6_TABLE_HASHSZ
- 1);
254 head
= &net
->ipv6
.fib_table_hash
[h
];
255 hlist_for_each_entry_rcu(tb
, head
, tb6_hlist
) {
256 if (tb
->tb6_id
== id
) {
265 EXPORT_SYMBOL_GPL(fib6_get_table
);
267 static void __net_init
fib6_tables_init(struct net
*net
)
269 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
270 fib6_link_table(net
, net
->ipv6
.fib6_local_tbl
);
274 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
276 return fib6_get_table(net
, id
);
279 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
281 return net
->ipv6
.fib6_main_tbl
;
284 struct dst_entry
*fib6_rule_lookup(struct net
*net
, struct flowi6
*fl6
,
285 int flags
, pol_lookup_t lookup
)
289 rt
= lookup(net
, net
->ipv6
.fib6_main_tbl
, fl6
, flags
);
290 if (rt
->rt6i_flags
& RTF_REJECT
&&
291 rt
->dst
.error
== -EAGAIN
) {
293 rt
= net
->ipv6
.ip6_null_entry
;
300 static void __net_init
fib6_tables_init(struct net
*net
)
302 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
307 static int fib6_dump_node(struct fib6_walker
*w
)
312 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
313 res
= rt6_dump_route(rt
, w
->args
);
315 /* Frame is full, suspend walking */
324 static void fib6_dump_end(struct netlink_callback
*cb
)
326 struct net
*net
= sock_net(cb
->skb
->sk
);
327 struct fib6_walker
*w
= (void *)cb
->args
[2];
332 fib6_walker_unlink(net
, w
);
337 cb
->done
= (void *)cb
->args
[3];
341 static int fib6_dump_done(struct netlink_callback
*cb
)
344 return cb
->done
? cb
->done(cb
) : 0;
347 static int fib6_dump_table(struct fib6_table
*table
, struct sk_buff
*skb
,
348 struct netlink_callback
*cb
)
350 struct net
*net
= sock_net(skb
->sk
);
351 struct fib6_walker
*w
;
354 w
= (void *)cb
->args
[2];
355 w
->root
= &table
->tb6_root
;
357 if (cb
->args
[4] == 0) {
361 read_lock_bh(&table
->tb6_lock
);
362 res
= fib6_walk(net
, w
);
363 read_unlock_bh(&table
->tb6_lock
);
366 cb
->args
[5] = w
->root
->fn_sernum
;
369 if (cb
->args
[5] != w
->root
->fn_sernum
) {
370 /* Begin at the root if the tree changed */
371 cb
->args
[5] = w
->root
->fn_sernum
;
378 read_lock_bh(&table
->tb6_lock
);
379 res
= fib6_walk_continue(w
);
380 read_unlock_bh(&table
->tb6_lock
);
382 fib6_walker_unlink(net
, w
);
390 static int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
392 struct net
*net
= sock_net(skb
->sk
);
394 unsigned int e
= 0, s_e
;
395 struct rt6_rtnl_dump_arg arg
;
396 struct fib6_walker
*w
;
397 struct fib6_table
*tb
;
398 struct hlist_head
*head
;
404 w
= (void *)cb
->args
[2];
408 * 1. hook callback destructor.
410 cb
->args
[3] = (long)cb
->done
;
411 cb
->done
= fib6_dump_done
;
414 * 2. allocate and initialize walker.
416 w
= kzalloc(sizeof(*w
), GFP_ATOMIC
);
419 w
->func
= fib6_dump_node
;
420 cb
->args
[2] = (long)w
;
429 for (h
= s_h
; h
< FIB6_TABLE_HASHSZ
; h
++, s_e
= 0) {
431 head
= &net
->ipv6
.fib_table_hash
[h
];
432 hlist_for_each_entry_rcu(tb
, head
, tb6_hlist
) {
435 res
= fib6_dump_table(tb
, skb
, cb
);
447 res
= res
< 0 ? res
: skb
->len
;
456 * return the appropriate node for a routing tree "add" operation
457 * by either creating and inserting or by returning an existing
461 static struct fib6_node
*fib6_add_1(struct fib6_node
*root
,
462 struct in6_addr
*addr
, int plen
,
463 int offset
, int allow_create
,
464 int replace_required
, int sernum
)
466 struct fib6_node
*fn
, *in
, *ln
;
467 struct fib6_node
*pn
= NULL
;
472 RT6_TRACE("fib6_add_1\n");
474 /* insert node in tree */
479 key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
484 if (plen
< fn
->fn_bit
||
485 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
)) {
487 if (replace_required
) {
488 pr_warn("Can't replace route, no match found\n");
489 return ERR_PTR(-ENOENT
);
491 pr_warn("NLM_F_CREATE should be set when creating new route\n");
500 if (plen
== fn
->fn_bit
) {
501 /* clean up an intermediate node */
502 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
503 rt6_release(fn
->leaf
);
507 fn
->fn_sernum
= sernum
;
513 * We have more bits to go
516 /* Try to walk down on tree. */
517 fn
->fn_sernum
= sernum
;
518 dir
= addr_bit_set(addr
, fn
->fn_bit
);
520 fn
= dir
? fn
->right
: fn
->left
;
524 /* We should not create new node because
525 * NLM_F_REPLACE was specified without NLM_F_CREATE
526 * I assume it is safe to require NLM_F_CREATE when
527 * REPLACE flag is used! Later we may want to remove the
528 * check for replace_required, because according
529 * to netlink specification, NLM_F_CREATE
530 * MUST be specified if new route is created.
531 * That would keep IPv6 consistent with IPv4
533 if (replace_required
) {
534 pr_warn("Can't replace route, no match found\n");
535 return ERR_PTR(-ENOENT
);
537 pr_warn("NLM_F_CREATE should be set when creating new route\n");
540 * We walked to the bottom of tree.
541 * Create new leaf node without children.
547 return ERR_PTR(-ENOMEM
);
551 ln
->fn_sernum
= sernum
;
563 * split since we don't have a common prefix anymore or
564 * we have a less significant route.
565 * we've to insert an intermediate node on the list
566 * this new node will point to the one we need to create
572 /* find 1st bit in difference between the 2 addrs.
574 See comment in __ipv6_addr_diff: bit may be an invalid value,
575 but if it is >= plen, the value is ignored in any case.
578 bit
= __ipv6_addr_diff(addr
, &key
->addr
, sizeof(*addr
));
583 * (new leaf node)[ln] (old node)[fn]
594 return ERR_PTR(-ENOMEM
);
598 * new intermediate node.
600 * be off since that an address that chooses one of
601 * the branches would not match less specific routes
602 * in the other branch
609 atomic_inc(&in
->leaf
->rt6i_ref
);
611 in
->fn_sernum
= sernum
;
613 /* update parent pointer */
624 ln
->fn_sernum
= sernum
;
626 if (addr_bit_set(addr
, bit
)) {
633 } else { /* plen <= bit */
636 * (new leaf node)[ln]
638 * (old node)[fn] NULL
644 return ERR_PTR(-ENOMEM
);
650 ln
->fn_sernum
= sernum
;
657 if (addr_bit_set(&key
->addr
, plen
))
667 static bool rt6_qualify_for_ecmp(struct rt6_info
*rt
)
669 return (rt
->rt6i_flags
& (RTF_GATEWAY
|RTF_ADDRCONF
|RTF_DYNAMIC
)) ==
673 static void fib6_copy_metrics(u32
*mp
, const struct mx6_config
*mxc
)
677 for (i
= 0; i
< RTAX_MAX
; i
++) {
678 if (test_bit(i
, mxc
->mx_valid
))
683 static int fib6_commit_metrics(struct dst_entry
*dst
, struct mx6_config
*mxc
)
688 if (dst
->flags
& DST_HOST
) {
689 u32
*mp
= dst_metrics_write_ptr(dst
);
694 fib6_copy_metrics(mp
, mxc
);
696 dst_init_metrics(dst
, mxc
->mx
, false);
698 /* We've stolen mx now. */
705 static void fib6_purge_rt(struct rt6_info
*rt
, struct fib6_node
*fn
,
708 if (atomic_read(&rt
->rt6i_ref
) != 1) {
709 /* This route is used as dummy address holder in some split
710 * nodes. It is not leaked, but it still holds other resources,
711 * which must be released in time. So, scan ascendant nodes
712 * and replace dummy references to this route with references
713 * to still alive ones.
716 if (!(fn
->fn_flags
& RTN_RTINFO
) && fn
->leaf
== rt
) {
717 fn
->leaf
= fib6_find_prefix(net
, fn
);
718 atomic_inc(&fn
->leaf
->rt6i_ref
);
723 /* No more references are possible at this point. */
724 BUG_ON(atomic_read(&rt
->rt6i_ref
) != 1);
729 * Insert routing information in a node.
732 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
733 struct nl_info
*info
, struct mx6_config
*mxc
)
735 struct rt6_info
*iter
= NULL
;
736 struct rt6_info
**ins
;
737 struct rt6_info
**fallback_ins
= NULL
;
738 int replace
= (info
->nlh
&&
739 (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
));
740 int add
= (!info
->nlh
||
741 (info
->nlh
->nlmsg_flags
& NLM_F_CREATE
));
743 bool rt_can_ecmp
= rt6_qualify_for_ecmp(rt
);
748 for (iter
= fn
->leaf
; iter
; iter
= iter
->dst
.rt6_next
) {
750 * Search for duplicates
753 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
755 * Same priority level
758 (info
->nlh
->nlmsg_flags
& NLM_F_EXCL
))
761 if (rt_can_ecmp
== rt6_qualify_for_ecmp(iter
)) {
766 fallback_ins
= fallback_ins
?: ins
;
770 if (iter
->dst
.dev
== rt
->dst
.dev
&&
771 iter
->rt6i_idev
== rt
->rt6i_idev
&&
772 ipv6_addr_equal(&iter
->rt6i_gateway
,
773 &rt
->rt6i_gateway
)) {
774 if (rt
->rt6i_nsiblings
)
775 rt
->rt6i_nsiblings
= 0;
776 if (!(iter
->rt6i_flags
& RTF_EXPIRES
))
778 if (!(rt
->rt6i_flags
& RTF_EXPIRES
))
779 rt6_clean_expires(iter
);
781 rt6_set_expires(iter
, rt
->dst
.expires
);
782 iter
->rt6i_pmtu
= rt
->rt6i_pmtu
;
785 /* If we have the same destination and the same metric,
786 * but not the same gateway, then the route we try to
787 * add is sibling to this route, increment our counter
788 * of siblings, and later we will add our route to the
790 * Only static routes (which don't have flag
791 * RTF_EXPIRES) are used for ECMPv6.
793 * To avoid long list, we only had siblings if the
794 * route have a gateway.
797 rt6_qualify_for_ecmp(iter
))
798 rt
->rt6i_nsiblings
++;
801 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
805 ins
= &iter
->dst
.rt6_next
;
808 if (fallback_ins
&& !found
) {
809 /* No ECMP-able route found, replace first non-ECMP one */
815 /* Reset round-robin state, if necessary */
816 if (ins
== &fn
->leaf
)
819 /* Link this route to others same route. */
820 if (rt
->rt6i_nsiblings
) {
821 unsigned int rt6i_nsiblings
;
822 struct rt6_info
*sibling
, *temp_sibling
;
824 /* Find the first route that have the same metric */
827 if (sibling
->rt6i_metric
== rt
->rt6i_metric
&&
828 rt6_qualify_for_ecmp(sibling
)) {
829 list_add_tail(&rt
->rt6i_siblings
,
830 &sibling
->rt6i_siblings
);
833 sibling
= sibling
->dst
.rt6_next
;
835 /* For each sibling in the list, increment the counter of
836 * siblings. BUG() if counters does not match, list of siblings
840 list_for_each_entry_safe(sibling
, temp_sibling
,
841 &rt
->rt6i_siblings
, rt6i_siblings
) {
842 sibling
->rt6i_nsiblings
++;
843 BUG_ON(sibling
->rt6i_nsiblings
!= rt
->rt6i_nsiblings
);
846 BUG_ON(rt6i_nsiblings
!= rt
->rt6i_nsiblings
);
854 pr_warn("NLM_F_CREATE should be set when creating new route\n");
857 err
= fib6_commit_metrics(&rt
->dst
, mxc
);
861 rt
->dst
.rt6_next
= iter
;
864 atomic_inc(&rt
->rt6i_ref
);
865 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
, 0);
866 info
->nl_net
->ipv6
.rt6_stats
->fib_rt_entries
++;
868 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
869 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
870 fn
->fn_flags
|= RTN_RTINFO
;
879 pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
883 err
= fib6_commit_metrics(&rt
->dst
, mxc
);
889 rt
->dst
.rt6_next
= iter
->dst
.rt6_next
;
890 atomic_inc(&rt
->rt6i_ref
);
891 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
, NLM_F_REPLACE
);
892 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
893 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
894 fn
->fn_flags
|= RTN_RTINFO
;
896 nsiblings
= iter
->rt6i_nsiblings
;
897 fib6_purge_rt(iter
, fn
, info
->nl_net
);
901 /* Replacing an ECMP route, remove all siblings */
902 ins
= &rt
->dst
.rt6_next
;
905 if (rt6_qualify_for_ecmp(iter
)) {
906 *ins
= iter
->dst
.rt6_next
;
907 fib6_purge_rt(iter
, fn
, info
->nl_net
);
911 ins
= &iter
->dst
.rt6_next
;
915 WARN_ON(nsiblings
!= 0);
922 static void fib6_start_gc(struct net
*net
, struct rt6_info
*rt
)
924 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
) &&
925 (rt
->rt6i_flags
& (RTF_EXPIRES
| RTF_CACHE
)))
926 mod_timer(&net
->ipv6
.ip6_fib_timer
,
927 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
930 void fib6_force_start_gc(struct net
*net
)
932 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
))
933 mod_timer(&net
->ipv6
.ip6_fib_timer
,
934 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
938 * Add routing information to the routing tree.
939 * <destination addr>/<source addr>
940 * with source addr info in sub-trees
943 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
,
944 struct nl_info
*info
, struct mx6_config
*mxc
)
946 struct fib6_node
*fn
, *pn
= NULL
;
948 int allow_create
= 1;
949 int replace_required
= 0;
950 int sernum
= fib6_new_sernum(info
->nl_net
);
952 if (WARN_ON_ONCE((rt
->dst
.flags
& DST_NOCACHE
) &&
953 !atomic_read(&rt
->dst
.__refcnt
)))
957 if (!(info
->nlh
->nlmsg_flags
& NLM_F_CREATE
))
959 if (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
)
960 replace_required
= 1;
962 if (!allow_create
&& !replace_required
)
963 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
965 fn
= fib6_add_1(root
, &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
,
966 offsetof(struct rt6_info
, rt6i_dst
), allow_create
,
967 replace_required
, sernum
);
976 #ifdef CONFIG_IPV6_SUBTREES
977 if (rt
->rt6i_src
.plen
) {
978 struct fib6_node
*sn
;
981 struct fib6_node
*sfn
;
993 /* Create subtree root node */
998 sfn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
999 atomic_inc(&info
->nl_net
->ipv6
.ip6_null_entry
->rt6i_ref
);
1000 sfn
->fn_flags
= RTN_ROOT
;
1001 sfn
->fn_sernum
= sernum
;
1003 /* Now add the first leaf node to new subtree */
1005 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
1007 offsetof(struct rt6_info
, rt6i_src
),
1008 allow_create
, replace_required
, sernum
);
1011 /* If it is failed, discard just allocated
1012 root, and then (in st_failure) stale node
1020 /* Now link new subtree to main tree */
1024 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
1026 offsetof(struct rt6_info
, rt6i_src
),
1027 allow_create
, replace_required
, sernum
);
1037 atomic_inc(&rt
->rt6i_ref
);
1043 err
= fib6_add_rt2node(fn
, rt
, info
, mxc
);
1045 fib6_start_gc(info
->nl_net
, rt
);
1046 if (!(rt
->rt6i_flags
& RTF_CACHE
))
1047 fib6_prune_clones(info
->nl_net
, pn
);
1048 rt
->dst
.flags
&= ~DST_NOCACHE
;
1053 #ifdef CONFIG_IPV6_SUBTREES
1055 * If fib6_add_1 has cleared the old leaf pointer in the
1056 * super-tree leaf node we have to find a new one for it.
1058 if (pn
!= fn
&& pn
->leaf
== rt
) {
1060 atomic_dec(&rt
->rt6i_ref
);
1062 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
1063 pn
->leaf
= fib6_find_prefix(info
->nl_net
, pn
);
1066 WARN_ON(pn
->leaf
== NULL
);
1067 pn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
1070 atomic_inc(&pn
->leaf
->rt6i_ref
);
1073 if (!(rt
->dst
.flags
& DST_NOCACHE
))
1078 #ifdef CONFIG_IPV6_SUBTREES
1079 /* Subtree creation failed, probably main tree node
1080 is orphan. If it is, shoot it.
1083 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
1084 fib6_repair_tree(info
->nl_net
, fn
);
1085 if (!(rt
->dst
.flags
& DST_NOCACHE
))
1092 * Routing tree lookup
1096 struct lookup_args
{
1097 int offset
; /* key offset on rt6_info */
1098 const struct in6_addr
*addr
; /* search key */
1101 static struct fib6_node
*fib6_lookup_1(struct fib6_node
*root
,
1102 struct lookup_args
*args
)
1104 struct fib6_node
*fn
;
1107 if (unlikely(args
->offset
== 0))
1117 struct fib6_node
*next
;
1119 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
1121 next
= dir
? fn
->right
: fn
->left
;
1131 if (FIB6_SUBTREE(fn
) || fn
->fn_flags
& RTN_RTINFO
) {
1134 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
1137 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
)) {
1138 #ifdef CONFIG_IPV6_SUBTREES
1140 struct fib6_node
*sfn
;
1141 sfn
= fib6_lookup_1(fn
->subtree
,
1148 if (fn
->fn_flags
& RTN_RTINFO
)
1152 #ifdef CONFIG_IPV6_SUBTREES
1155 if (fn
->fn_flags
& RTN_ROOT
)
1164 struct fib6_node
*fib6_lookup(struct fib6_node
*root
, const struct in6_addr
*daddr
,
1165 const struct in6_addr
*saddr
)
1167 struct fib6_node
*fn
;
1168 struct lookup_args args
[] = {
1170 .offset
= offsetof(struct rt6_info
, rt6i_dst
),
1173 #ifdef CONFIG_IPV6_SUBTREES
1175 .offset
= offsetof(struct rt6_info
, rt6i_src
),
1180 .offset
= 0, /* sentinel */
1184 fn
= fib6_lookup_1(root
, daddr
? args
: args
+ 1);
1185 if (!fn
|| fn
->fn_flags
& RTN_TL_ROOT
)
1192 * Get node with specified destination prefix (and source prefix,
1193 * if subtrees are used)
1197 static struct fib6_node
*fib6_locate_1(struct fib6_node
*root
,
1198 const struct in6_addr
*addr
,
1199 int plen
, int offset
)
1201 struct fib6_node
*fn
;
1203 for (fn
= root
; fn
; ) {
1204 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
1209 if (plen
< fn
->fn_bit
||
1210 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
1213 if (plen
== fn
->fn_bit
)
1217 * We have more bits to go
1219 if (addr_bit_set(addr
, fn
->fn_bit
))
1227 struct fib6_node
*fib6_locate(struct fib6_node
*root
,
1228 const struct in6_addr
*daddr
, int dst_len
,
1229 const struct in6_addr
*saddr
, int src_len
)
1231 struct fib6_node
*fn
;
1233 fn
= fib6_locate_1(root
, daddr
, dst_len
,
1234 offsetof(struct rt6_info
, rt6i_dst
));
1236 #ifdef CONFIG_IPV6_SUBTREES
1238 WARN_ON(saddr
== NULL
);
1239 if (fn
&& fn
->subtree
)
1240 fn
= fib6_locate_1(fn
->subtree
, saddr
, src_len
,
1241 offsetof(struct rt6_info
, rt6i_src
));
1245 if (fn
&& fn
->fn_flags
& RTN_RTINFO
)
1257 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
)
1259 if (fn
->fn_flags
& RTN_ROOT
)
1260 return net
->ipv6
.ip6_null_entry
;
1264 return fn
->left
->leaf
;
1266 return fn
->right
->leaf
;
1268 fn
= FIB6_SUBTREE(fn
);
1274 * Called to trim the tree of intermediate nodes when possible. "fn"
1275 * is the node we want to try and remove.
1278 static struct fib6_node
*fib6_repair_tree(struct net
*net
,
1279 struct fib6_node
*fn
)
1283 struct fib6_node
*child
, *pn
;
1284 struct fib6_walker
*w
;
1288 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
1291 WARN_ON(fn
->fn_flags
& RTN_RTINFO
);
1292 WARN_ON(fn
->fn_flags
& RTN_TL_ROOT
);
1298 child
= fn
->right
, children
|= 1;
1300 child
= fn
->left
, children
|= 2;
1302 if (children
== 3 || FIB6_SUBTREE(fn
)
1303 #ifdef CONFIG_IPV6_SUBTREES
1304 /* Subtree root (i.e. fn) may have one child */
1305 || (children
&& fn
->fn_flags
& RTN_ROOT
)
1308 fn
->leaf
= fib6_find_prefix(net
, fn
);
1312 fn
->leaf
= net
->ipv6
.ip6_null_entry
;
1315 atomic_inc(&fn
->leaf
->rt6i_ref
);
1320 #ifdef CONFIG_IPV6_SUBTREES
1321 if (FIB6_SUBTREE(pn
) == fn
) {
1322 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1323 FIB6_SUBTREE(pn
) = NULL
;
1326 WARN_ON(fn
->fn_flags
& RTN_ROOT
);
1328 if (pn
->right
== fn
)
1330 else if (pn
->left
== fn
)
1339 #ifdef CONFIG_IPV6_SUBTREES
1343 read_lock(&net
->ipv6
.fib6_walker_lock
);
1344 FOR_WALKERS(net
, w
) {
1346 if (w
->root
== fn
) {
1347 w
->root
= w
->node
= NULL
;
1348 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1349 } else if (w
->node
== fn
) {
1350 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1355 if (w
->root
== fn
) {
1357 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1359 if (w
->node
== fn
) {
1362 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1363 w
->state
= w
->state
>= FWS_R
? FWS_U
: FWS_INIT
;
1365 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1366 w
->state
= w
->state
>= FWS_C
? FWS_U
: FWS_INIT
;
1371 read_unlock(&net
->ipv6
.fib6_walker_lock
);
1374 if (pn
->fn_flags
& RTN_RTINFO
|| FIB6_SUBTREE(pn
))
1377 rt6_release(pn
->leaf
);
1383 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1384 struct nl_info
*info
)
1386 struct fib6_walker
*w
;
1387 struct rt6_info
*rt
= *rtp
;
1388 struct net
*net
= info
->nl_net
;
1390 RT6_TRACE("fib6_del_route\n");
1393 *rtp
= rt
->dst
.rt6_next
;
1394 rt
->rt6i_node
= NULL
;
1395 net
->ipv6
.rt6_stats
->fib_rt_entries
--;
1396 net
->ipv6
.rt6_stats
->fib_discarded_routes
++;
1398 /* Reset round-robin state, if necessary */
1399 if (fn
->rr_ptr
== rt
)
1402 /* Remove this entry from other siblings */
1403 if (rt
->rt6i_nsiblings
) {
1404 struct rt6_info
*sibling
, *next_sibling
;
1406 list_for_each_entry_safe(sibling
, next_sibling
,
1407 &rt
->rt6i_siblings
, rt6i_siblings
)
1408 sibling
->rt6i_nsiblings
--;
1409 rt
->rt6i_nsiblings
= 0;
1410 list_del_init(&rt
->rt6i_siblings
);
1413 /* Adjust walkers */
1414 read_lock(&net
->ipv6
.fib6_walker_lock
);
1415 FOR_WALKERS(net
, w
) {
1416 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1417 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1418 w
->leaf
= rt
->dst
.rt6_next
;
1423 read_unlock(&net
->ipv6
.fib6_walker_lock
);
1425 rt
->dst
.rt6_next
= NULL
;
1427 /* If it was last route, expunge its radix tree node */
1429 fn
->fn_flags
&= ~RTN_RTINFO
;
1430 net
->ipv6
.rt6_stats
->fib_route_nodes
--;
1431 fn
= fib6_repair_tree(net
, fn
);
1434 fib6_purge_rt(rt
, fn
, net
);
1436 inet6_rt_notify(RTM_DELROUTE
, rt
, info
, 0);
1440 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1442 struct net
*net
= info
->nl_net
;
1443 struct fib6_node
*fn
= rt
->rt6i_node
;
1444 struct rt6_info
**rtp
;
1447 if (rt
->dst
.obsolete
> 0) {
1452 if (!fn
|| rt
== net
->ipv6
.ip6_null_entry
)
1455 WARN_ON(!(fn
->fn_flags
& RTN_RTINFO
));
1457 if (!(rt
->rt6i_flags
& RTF_CACHE
)) {
1458 struct fib6_node
*pn
= fn
;
1459 #ifdef CONFIG_IPV6_SUBTREES
1460 /* clones of this route might be in another subtree */
1461 if (rt
->rt6i_src
.plen
) {
1462 while (!(pn
->fn_flags
& RTN_ROOT
))
1467 fib6_prune_clones(info
->nl_net
, pn
);
1471 * Walk the leaf entries looking for ourself
1474 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->dst
.rt6_next
) {
1476 fib6_del_route(fn
, rtp
, info
);
1484 * Tree traversal function.
1486 * Certainly, it is not interrupt safe.
1487 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1488 * It means, that we can modify tree during walking
1489 * and use this function for garbage collection, clone pruning,
1490 * cleaning tree when a device goes down etc. etc.
1492 * It guarantees that every node will be traversed,
1493 * and that it will be traversed only once.
1495 * Callback function w->func may return:
1496 * 0 -> continue walking.
1497 * positive value -> walking is suspended (used by tree dumps,
1498 * and probably by gc, if it will be split to several slices)
1499 * negative value -> terminate walking.
1501 * The function itself returns:
1502 * 0 -> walk is complete.
1503 * >0 -> walk is incomplete (i.e. suspended)
1504 * <0 -> walk is terminated by an error.
1507 static int fib6_walk_continue(struct fib6_walker
*w
)
1509 struct fib6_node
*fn
, *pn
;
1516 if (w
->prune
&& fn
!= w
->root
&&
1517 fn
->fn_flags
& RTN_RTINFO
&& w
->state
< FWS_C
) {
1522 #ifdef CONFIG_IPV6_SUBTREES
1524 if (FIB6_SUBTREE(fn
)) {
1525 w
->node
= FIB6_SUBTREE(fn
);
1533 w
->state
= FWS_INIT
;
1539 w
->node
= fn
->right
;
1540 w
->state
= FWS_INIT
;
1546 if (w
->leaf
&& fn
->fn_flags
& RTN_RTINFO
) {
1568 #ifdef CONFIG_IPV6_SUBTREES
1569 if (FIB6_SUBTREE(pn
) == fn
) {
1570 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1575 if (pn
->left
== fn
) {
1579 if (pn
->right
== fn
) {
1581 w
->leaf
= w
->node
->leaf
;
1591 static int fib6_walk(struct net
*net
, struct fib6_walker
*w
)
1595 w
->state
= FWS_INIT
;
1598 fib6_walker_link(net
, w
);
1599 res
= fib6_walk_continue(w
);
1601 fib6_walker_unlink(net
, w
);
1605 static int fib6_clean_node(struct fib6_walker
*w
)
1608 struct rt6_info
*rt
;
1609 struct fib6_cleaner
*c
= container_of(w
, struct fib6_cleaner
, w
);
1610 struct nl_info info
= {
1614 if (c
->sernum
!= FIB6_NO_SERNUM_CHANGE
&&
1615 w
->node
->fn_sernum
!= c
->sernum
)
1616 w
->node
->fn_sernum
= c
->sernum
;
1619 WARN_ON_ONCE(c
->sernum
== FIB6_NO_SERNUM_CHANGE
);
1624 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1625 res
= c
->func(rt
, c
->arg
);
1628 res
= fib6_del(rt
, &info
);
1631 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
1632 __func__
, rt
, rt
->rt6i_node
, res
);
1645 * Convenient frontend to tree walker.
1647 * func is called on each route.
1648 * It may return -1 -> delete this route.
1649 * 0 -> continue walking
1651 * prune==1 -> only immediate children of node (certainly,
1652 * ignoring pure split nodes) will be scanned.
1655 static void fib6_clean_tree(struct net
*net
, struct fib6_node
*root
,
1656 int (*func
)(struct rt6_info
*, void *arg
),
1657 bool prune
, int sernum
, void *arg
)
1659 struct fib6_cleaner c
;
1662 c
.w
.func
= fib6_clean_node
;
1671 fib6_walk(net
, &c
.w
);
1674 static void __fib6_clean_all(struct net
*net
,
1675 int (*func
)(struct rt6_info
*, void *),
1676 int sernum
, void *arg
)
1678 struct fib6_table
*table
;
1679 struct hlist_head
*head
;
1683 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1684 head
= &net
->ipv6
.fib_table_hash
[h
];
1685 hlist_for_each_entry_rcu(table
, head
, tb6_hlist
) {
1686 write_lock_bh(&table
->tb6_lock
);
1687 fib6_clean_tree(net
, &table
->tb6_root
,
1688 func
, false, sernum
, arg
);
1689 write_unlock_bh(&table
->tb6_lock
);
1695 void fib6_clean_all(struct net
*net
, int (*func
)(struct rt6_info
*, void *),
1698 __fib6_clean_all(net
, func
, FIB6_NO_SERNUM_CHANGE
, arg
);
1701 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1703 if (rt
->rt6i_flags
& RTF_CACHE
) {
1704 RT6_TRACE("pruning clone %p\n", rt
);
1711 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
)
1713 fib6_clean_tree(net
, fn
, fib6_prune_clone
, true,
1714 FIB6_NO_SERNUM_CHANGE
, NULL
);
1717 static void fib6_flush_trees(struct net
*net
)
1719 int new_sernum
= fib6_new_sernum(net
);
1721 __fib6_clean_all(net
, NULL
, new_sernum
, NULL
);
1725 * Garbage collection
1734 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1736 struct fib6_gc_args
*gc_args
= arg
;
1737 unsigned long now
= jiffies
;
1740 * check addrconf expiration here.
1741 * Routes are expired even if they are in use.
1743 * Also age clones. Note, that clones are aged out
1744 * only if they are not in use now.
1747 if (rt
->rt6i_flags
& RTF_EXPIRES
&& rt
->dst
.expires
) {
1748 if (time_after(now
, rt
->dst
.expires
)) {
1749 RT6_TRACE("expiring %p\n", rt
);
1753 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1754 if (atomic_read(&rt
->dst
.__refcnt
) == 0 &&
1755 time_after_eq(now
, rt
->dst
.lastuse
+ gc_args
->timeout
)) {
1756 RT6_TRACE("aging clone %p\n", rt
);
1758 } else if (rt
->rt6i_flags
& RTF_GATEWAY
) {
1759 struct neighbour
*neigh
;
1760 __u8 neigh_flags
= 0;
1762 neigh
= dst_neigh_lookup(&rt
->dst
, &rt
->rt6i_gateway
);
1764 neigh_flags
= neigh
->flags
;
1765 neigh_release(neigh
);
1767 if (!(neigh_flags
& NTF_ROUTER
)) {
1768 RT6_TRACE("purging route %p via non-router but gateway\n",
1779 void fib6_run_gc(unsigned long expires
, struct net
*net
, bool force
)
1781 struct fib6_gc_args gc_args
;
1785 spin_lock_bh(&net
->ipv6
.fib6_gc_lock
);
1786 } else if (!spin_trylock_bh(&net
->ipv6
.fib6_gc_lock
)) {
1787 mod_timer(&net
->ipv6
.ip6_fib_timer
, jiffies
+ HZ
);
1790 gc_args
.timeout
= expires
? (int)expires
:
1791 net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1793 gc_args
.more
= icmp6_dst_gc();
1795 fib6_clean_all(net
, fib6_age
, &gc_args
);
1797 net
->ipv6
.ip6_rt_last_gc
= now
;
1800 mod_timer(&net
->ipv6
.ip6_fib_timer
,
1802 + net
->ipv6
.sysctl
.ip6_rt_gc_interval
));
1804 del_timer(&net
->ipv6
.ip6_fib_timer
);
1805 spin_unlock_bh(&net
->ipv6
.fib6_gc_lock
);
1808 static void fib6_gc_timer_cb(unsigned long arg
)
1810 fib6_run_gc(0, (struct net
*)arg
, true);
1813 static int __net_init
fib6_net_init(struct net
*net
)
1815 size_t size
= sizeof(struct hlist_head
) * FIB6_TABLE_HASHSZ
;
1817 spin_lock_init(&net
->ipv6
.fib6_gc_lock
);
1818 rwlock_init(&net
->ipv6
.fib6_walker_lock
);
1819 INIT_LIST_HEAD(&net
->ipv6
.fib6_walkers
);
1820 setup_timer(&net
->ipv6
.ip6_fib_timer
, fib6_gc_timer_cb
, (unsigned long)net
);
1822 net
->ipv6
.rt6_stats
= kzalloc(sizeof(*net
->ipv6
.rt6_stats
), GFP_KERNEL
);
1823 if (!net
->ipv6
.rt6_stats
)
1826 /* Avoid false sharing : Use at least a full cache line */
1827 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1829 net
->ipv6
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1830 if (!net
->ipv6
.fib_table_hash
)
1833 net
->ipv6
.fib6_main_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_main_tbl
),
1835 if (!net
->ipv6
.fib6_main_tbl
)
1836 goto out_fib_table_hash
;
1838 net
->ipv6
.fib6_main_tbl
->tb6_id
= RT6_TABLE_MAIN
;
1839 net
->ipv6
.fib6_main_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1840 net
->ipv6
.fib6_main_tbl
->tb6_root
.fn_flags
=
1841 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1842 inet_peer_base_init(&net
->ipv6
.fib6_main_tbl
->tb6_peers
);
1844 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1845 net
->ipv6
.fib6_local_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_local_tbl
),
1847 if (!net
->ipv6
.fib6_local_tbl
)
1848 goto out_fib6_main_tbl
;
1849 net
->ipv6
.fib6_local_tbl
->tb6_id
= RT6_TABLE_LOCAL
;
1850 net
->ipv6
.fib6_local_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1851 net
->ipv6
.fib6_local_tbl
->tb6_root
.fn_flags
=
1852 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1853 inet_peer_base_init(&net
->ipv6
.fib6_local_tbl
->tb6_peers
);
1855 fib6_tables_init(net
);
1859 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1861 kfree(net
->ipv6
.fib6_main_tbl
);
1864 kfree(net
->ipv6
.fib_table_hash
);
1866 kfree(net
->ipv6
.rt6_stats
);
1871 static void fib6_net_exit(struct net
*net
)
1873 rt6_ifdown(net
, NULL
);
1874 del_timer_sync(&net
->ipv6
.ip6_fib_timer
);
1876 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1877 inetpeer_invalidate_tree(&net
->ipv6
.fib6_local_tbl
->tb6_peers
);
1878 kfree(net
->ipv6
.fib6_local_tbl
);
1880 inetpeer_invalidate_tree(&net
->ipv6
.fib6_main_tbl
->tb6_peers
);
1881 kfree(net
->ipv6
.fib6_main_tbl
);
1882 kfree(net
->ipv6
.fib_table_hash
);
1883 kfree(net
->ipv6
.rt6_stats
);
1886 static struct pernet_operations fib6_net_ops
= {
1887 .init
= fib6_net_init
,
1888 .exit
= fib6_net_exit
,
1891 int __init
fib6_init(void)
1895 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1896 sizeof(struct fib6_node
),
1897 0, SLAB_HWCACHE_ALIGN
,
1899 if (!fib6_node_kmem
)
1902 ret
= register_pernet_subsys(&fib6_net_ops
);
1904 goto out_kmem_cache_create
;
1906 ret
= __rtnl_register(PF_INET6
, RTM_GETROUTE
, NULL
, inet6_dump_fib
,
1909 goto out_unregister_subsys
;
1911 __fib6_flush_trees
= fib6_flush_trees
;
1915 out_unregister_subsys
:
1916 unregister_pernet_subsys(&fib6_net_ops
);
1917 out_kmem_cache_create
:
1918 kmem_cache_destroy(fib6_node_kmem
);
1922 void fib6_gc_cleanup(void)
1924 unregister_pernet_subsys(&fib6_net_ops
);
1925 kmem_cache_destroy(fib6_node_kmem
);
1928 #ifdef CONFIG_PROC_FS
1930 struct ipv6_route_iter
{
1931 struct seq_net_private p
;
1932 struct fib6_walker w
;
1934 struct fib6_table
*tbl
;
1938 static int ipv6_route_seq_show(struct seq_file
*seq
, void *v
)
1940 struct rt6_info
*rt
= v
;
1941 struct ipv6_route_iter
*iter
= seq
->private;
1943 seq_printf(seq
, "%pi6 %02x ", &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
);
1945 #ifdef CONFIG_IPV6_SUBTREES
1946 seq_printf(seq
, "%pi6 %02x ", &rt
->rt6i_src
.addr
, rt
->rt6i_src
.plen
);
1948 seq_puts(seq
, "00000000000000000000000000000000 00 ");
1950 if (rt
->rt6i_flags
& RTF_GATEWAY
)
1951 seq_printf(seq
, "%pi6", &rt
->rt6i_gateway
);
1953 seq_puts(seq
, "00000000000000000000000000000000");
1955 seq_printf(seq
, " %08x %08x %08x %08x %8s\n",
1956 rt
->rt6i_metric
, atomic_read(&rt
->dst
.__refcnt
),
1957 rt
->dst
.__use
, rt
->rt6i_flags
,
1958 rt
->dst
.dev
? rt
->dst
.dev
->name
: "");
1959 iter
->w
.leaf
= NULL
;
1963 static int ipv6_route_yield(struct fib6_walker
*w
)
1965 struct ipv6_route_iter
*iter
= w
->args
;
1971 iter
->w
.leaf
= iter
->w
.leaf
->dst
.rt6_next
;
1973 if (!iter
->skip
&& iter
->w
.leaf
)
1975 } while (iter
->w
.leaf
);
1980 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter
*iter
,
1983 memset(&iter
->w
, 0, sizeof(iter
->w
));
1984 iter
->w
.func
= ipv6_route_yield
;
1985 iter
->w
.root
= &iter
->tbl
->tb6_root
;
1986 iter
->w
.state
= FWS_INIT
;
1987 iter
->w
.node
= iter
->w
.root
;
1988 iter
->w
.args
= iter
;
1989 iter
->sernum
= iter
->w
.root
->fn_sernum
;
1990 INIT_LIST_HEAD(&iter
->w
.lh
);
1991 fib6_walker_link(net
, &iter
->w
);
1994 static struct fib6_table
*ipv6_route_seq_next_table(struct fib6_table
*tbl
,
1998 struct hlist_node
*node
;
2001 h
= (tbl
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1)) + 1;
2002 node
= rcu_dereference_bh(hlist_next_rcu(&tbl
->tb6_hlist
));
2008 while (!node
&& h
< FIB6_TABLE_HASHSZ
) {
2009 node
= rcu_dereference_bh(
2010 hlist_first_rcu(&net
->ipv6
.fib_table_hash
[h
++]));
2012 return hlist_entry_safe(node
, struct fib6_table
, tb6_hlist
);
2015 static void ipv6_route_check_sernum(struct ipv6_route_iter
*iter
)
2017 if (iter
->sernum
!= iter
->w
.root
->fn_sernum
) {
2018 iter
->sernum
= iter
->w
.root
->fn_sernum
;
2019 iter
->w
.state
= FWS_INIT
;
2020 iter
->w
.node
= iter
->w
.root
;
2021 WARN_ON(iter
->w
.skip
);
2022 iter
->w
.skip
= iter
->w
.count
;
2026 static void *ipv6_route_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2030 struct net
*net
= seq_file_net(seq
);
2031 struct ipv6_route_iter
*iter
= seq
->private;
2036 n
= ((struct rt6_info
*)v
)->dst
.rt6_next
;
2043 ipv6_route_check_sernum(iter
);
2044 read_lock(&iter
->tbl
->tb6_lock
);
2045 r
= fib6_walk_continue(&iter
->w
);
2046 read_unlock(&iter
->tbl
->tb6_lock
);
2050 return iter
->w
.leaf
;
2052 fib6_walker_unlink(net
, &iter
->w
);
2055 fib6_walker_unlink(net
, &iter
->w
);
2057 iter
->tbl
= ipv6_route_seq_next_table(iter
->tbl
, net
);
2061 ipv6_route_seq_setup_walk(iter
, net
);
2065 static void *ipv6_route_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2068 struct net
*net
= seq_file_net(seq
);
2069 struct ipv6_route_iter
*iter
= seq
->private;
2072 iter
->tbl
= ipv6_route_seq_next_table(NULL
, net
);
2076 ipv6_route_seq_setup_walk(iter
, net
);
2077 return ipv6_route_seq_next(seq
, NULL
, pos
);
2083 static bool ipv6_route_iter_active(struct ipv6_route_iter
*iter
)
2085 struct fib6_walker
*w
= &iter
->w
;
2086 return w
->node
&& !(w
->state
== FWS_U
&& w
->node
== w
->root
);
2089 static void ipv6_route_seq_stop(struct seq_file
*seq
, void *v
)
2092 struct net
*net
= seq_file_net(seq
);
2093 struct ipv6_route_iter
*iter
= seq
->private;
2095 if (ipv6_route_iter_active(iter
))
2096 fib6_walker_unlink(net
, &iter
->w
);
2098 rcu_read_unlock_bh();
2101 static const struct seq_operations ipv6_route_seq_ops
= {
2102 .start
= ipv6_route_seq_start
,
2103 .next
= ipv6_route_seq_next
,
2104 .stop
= ipv6_route_seq_stop
,
2105 .show
= ipv6_route_seq_show
2108 int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2110 return seq_open_net(inode
, file
, &ipv6_route_seq_ops
,
2111 sizeof(struct ipv6_route_iter
));
2114 #endif /* CONFIG_PROC_FS */