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 free_percpu(non_pcpu_rt
->rt6i_pcpu
);
181 non_pcpu_rt
->rt6i_pcpu
= NULL
;
184 static void rt6_release(struct rt6_info
*rt
)
186 if (atomic_dec_and_test(&rt
->rt6i_ref
)) {
192 static void fib6_link_table(struct net
*net
, struct fib6_table
*tb
)
197 * Initialize table lock at a single place to give lockdep a key,
198 * tables aren't visible prior to being linked to the list.
200 rwlock_init(&tb
->tb6_lock
);
202 h
= tb
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1);
205 * No protection necessary, this is the only list mutatation
206 * operation, tables never disappear once they exist.
208 hlist_add_head_rcu(&tb
->tb6_hlist
, &net
->ipv6
.fib_table_hash
[h
]);
211 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
213 static struct fib6_table
*fib6_alloc_table(struct net
*net
, u32 id
)
215 struct fib6_table
*table
;
217 table
= kzalloc(sizeof(*table
), GFP_ATOMIC
);
220 table
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
221 table
->tb6_root
.fn_flags
= RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
222 inet_peer_base_init(&table
->tb6_peers
);
228 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
230 struct fib6_table
*tb
;
234 tb
= fib6_get_table(net
, id
);
238 tb
= fib6_alloc_table(net
, id
);
240 fib6_link_table(net
, tb
);
244 EXPORT_SYMBOL_GPL(fib6_new_table
);
246 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
248 struct fib6_table
*tb
;
249 struct hlist_head
*head
;
254 h
= id
& (FIB6_TABLE_HASHSZ
- 1);
256 head
= &net
->ipv6
.fib_table_hash
[h
];
257 hlist_for_each_entry_rcu(tb
, head
, tb6_hlist
) {
258 if (tb
->tb6_id
== id
) {
267 EXPORT_SYMBOL_GPL(fib6_get_table
);
269 static void __net_init
fib6_tables_init(struct net
*net
)
271 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
272 fib6_link_table(net
, net
->ipv6
.fib6_local_tbl
);
276 struct fib6_table
*fib6_new_table(struct net
*net
, u32 id
)
278 return fib6_get_table(net
, id
);
281 struct fib6_table
*fib6_get_table(struct net
*net
, u32 id
)
283 return net
->ipv6
.fib6_main_tbl
;
286 struct dst_entry
*fib6_rule_lookup(struct net
*net
, struct flowi6
*fl6
,
287 int flags
, pol_lookup_t lookup
)
291 rt
= lookup(net
, net
->ipv6
.fib6_main_tbl
, fl6
, flags
);
292 if (rt
->rt6i_flags
& RTF_REJECT
&&
293 rt
->dst
.error
== -EAGAIN
) {
295 rt
= net
->ipv6
.ip6_null_entry
;
302 static void __net_init
fib6_tables_init(struct net
*net
)
304 fib6_link_table(net
, net
->ipv6
.fib6_main_tbl
);
309 static int fib6_dump_node(struct fib6_walker
*w
)
314 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
315 res
= rt6_dump_route(rt
, w
->args
);
317 /* Frame is full, suspend walking */
326 static void fib6_dump_end(struct netlink_callback
*cb
)
328 struct net
*net
= sock_net(cb
->skb
->sk
);
329 struct fib6_walker
*w
= (void *)cb
->args
[2];
334 fib6_walker_unlink(net
, w
);
339 cb
->done
= (void *)cb
->args
[3];
343 static int fib6_dump_done(struct netlink_callback
*cb
)
346 return cb
->done
? cb
->done(cb
) : 0;
349 static int fib6_dump_table(struct fib6_table
*table
, struct sk_buff
*skb
,
350 struct netlink_callback
*cb
)
352 struct net
*net
= sock_net(skb
->sk
);
353 struct fib6_walker
*w
;
356 w
= (void *)cb
->args
[2];
357 w
->root
= &table
->tb6_root
;
359 if (cb
->args
[4] == 0) {
363 read_lock_bh(&table
->tb6_lock
);
364 res
= fib6_walk(net
, w
);
365 read_unlock_bh(&table
->tb6_lock
);
368 cb
->args
[5] = w
->root
->fn_sernum
;
371 if (cb
->args
[5] != w
->root
->fn_sernum
) {
372 /* Begin at the root if the tree changed */
373 cb
->args
[5] = w
->root
->fn_sernum
;
380 read_lock_bh(&table
->tb6_lock
);
381 res
= fib6_walk_continue(w
);
382 read_unlock_bh(&table
->tb6_lock
);
384 fib6_walker_unlink(net
, w
);
392 static int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
394 struct net
*net
= sock_net(skb
->sk
);
396 unsigned int e
= 0, s_e
;
397 struct rt6_rtnl_dump_arg arg
;
398 struct fib6_walker
*w
;
399 struct fib6_table
*tb
;
400 struct hlist_head
*head
;
406 w
= (void *)cb
->args
[2];
410 * 1. hook callback destructor.
412 cb
->args
[3] = (long)cb
->done
;
413 cb
->done
= fib6_dump_done
;
416 * 2. allocate and initialize walker.
418 w
= kzalloc(sizeof(*w
), GFP_ATOMIC
);
421 w
->func
= fib6_dump_node
;
422 cb
->args
[2] = (long)w
;
431 for (h
= s_h
; h
< FIB6_TABLE_HASHSZ
; h
++, s_e
= 0) {
433 head
= &net
->ipv6
.fib_table_hash
[h
];
434 hlist_for_each_entry_rcu(tb
, head
, tb6_hlist
) {
437 res
= fib6_dump_table(tb
, skb
, cb
);
449 res
= res
< 0 ? res
: skb
->len
;
458 * return the appropriate node for a routing tree "add" operation
459 * by either creating and inserting or by returning an existing
463 static struct fib6_node
*fib6_add_1(struct fib6_node
*root
,
464 struct in6_addr
*addr
, int plen
,
465 int offset
, int allow_create
,
466 int replace_required
, int sernum
)
468 struct fib6_node
*fn
, *in
, *ln
;
469 struct fib6_node
*pn
= NULL
;
474 RT6_TRACE("fib6_add_1\n");
476 /* insert node in tree */
481 key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
486 if (plen
< fn
->fn_bit
||
487 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
)) {
489 if (replace_required
) {
490 pr_warn("Can't replace route, no match found\n");
491 return ERR_PTR(-ENOENT
);
493 pr_warn("NLM_F_CREATE should be set when creating new route\n");
502 if (plen
== fn
->fn_bit
) {
503 /* clean up an intermediate node */
504 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
505 rt6_release(fn
->leaf
);
509 fn
->fn_sernum
= sernum
;
515 * We have more bits to go
518 /* Try to walk down on tree. */
519 fn
->fn_sernum
= sernum
;
520 dir
= addr_bit_set(addr
, fn
->fn_bit
);
522 fn
= dir
? fn
->right
: fn
->left
;
526 /* We should not create new node because
527 * NLM_F_REPLACE was specified without NLM_F_CREATE
528 * I assume it is safe to require NLM_F_CREATE when
529 * REPLACE flag is used! Later we may want to remove the
530 * check for replace_required, because according
531 * to netlink specification, NLM_F_CREATE
532 * MUST be specified if new route is created.
533 * That would keep IPv6 consistent with IPv4
535 if (replace_required
) {
536 pr_warn("Can't replace route, no match found\n");
537 return ERR_PTR(-ENOENT
);
539 pr_warn("NLM_F_CREATE should be set when creating new route\n");
542 * We walked to the bottom of tree.
543 * Create new leaf node without children.
549 return ERR_PTR(-ENOMEM
);
553 ln
->fn_sernum
= sernum
;
565 * split since we don't have a common prefix anymore or
566 * we have a less significant route.
567 * we've to insert an intermediate node on the list
568 * this new node will point to the one we need to create
574 /* find 1st bit in difference between the 2 addrs.
576 See comment in __ipv6_addr_diff: bit may be an invalid value,
577 but if it is >= plen, the value is ignored in any case.
580 bit
= __ipv6_addr_diff(addr
, &key
->addr
, sizeof(*addr
));
585 * (new leaf node)[ln] (old node)[fn]
596 return ERR_PTR(-ENOMEM
);
600 * new intermediate node.
602 * be off since that an address that chooses one of
603 * the branches would not match less specific routes
604 * in the other branch
611 atomic_inc(&in
->leaf
->rt6i_ref
);
613 in
->fn_sernum
= sernum
;
615 /* update parent pointer */
626 ln
->fn_sernum
= sernum
;
628 if (addr_bit_set(addr
, bit
)) {
635 } else { /* plen <= bit */
638 * (new leaf node)[ln]
640 * (old node)[fn] NULL
646 return ERR_PTR(-ENOMEM
);
652 ln
->fn_sernum
= sernum
;
659 if (addr_bit_set(&key
->addr
, plen
))
669 static bool rt6_qualify_for_ecmp(struct rt6_info
*rt
)
671 return (rt
->rt6i_flags
& (RTF_GATEWAY
|RTF_ADDRCONF
|RTF_DYNAMIC
)) ==
675 static void fib6_copy_metrics(u32
*mp
, const struct mx6_config
*mxc
)
679 for (i
= 0; i
< RTAX_MAX
; i
++) {
680 if (test_bit(i
, mxc
->mx_valid
))
685 static int fib6_commit_metrics(struct dst_entry
*dst
, struct mx6_config
*mxc
)
690 if (dst
->flags
& DST_HOST
) {
691 u32
*mp
= dst_metrics_write_ptr(dst
);
696 fib6_copy_metrics(mp
, mxc
);
698 dst_init_metrics(dst
, mxc
->mx
, false);
700 /* We've stolen mx now. */
707 static void fib6_purge_rt(struct rt6_info
*rt
, struct fib6_node
*fn
,
710 if (atomic_read(&rt
->rt6i_ref
) != 1) {
711 /* This route is used as dummy address holder in some split
712 * nodes. It is not leaked, but it still holds other resources,
713 * which must be released in time. So, scan ascendant nodes
714 * and replace dummy references to this route with references
715 * to still alive ones.
718 if (!(fn
->fn_flags
& RTN_RTINFO
) && fn
->leaf
== rt
) {
719 fn
->leaf
= fib6_find_prefix(net
, fn
);
720 atomic_inc(&fn
->leaf
->rt6i_ref
);
725 /* No more references are possible at this point. */
726 BUG_ON(atomic_read(&rt
->rt6i_ref
) != 1);
731 * Insert routing information in a node.
734 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
735 struct nl_info
*info
, struct mx6_config
*mxc
)
737 struct rt6_info
*iter
= NULL
;
738 struct rt6_info
**ins
;
739 struct rt6_info
**fallback_ins
= NULL
;
740 int replace
= (info
->nlh
&&
741 (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
));
742 int add
= (!info
->nlh
||
743 (info
->nlh
->nlmsg_flags
& NLM_F_CREATE
));
745 bool rt_can_ecmp
= rt6_qualify_for_ecmp(rt
);
750 for (iter
= fn
->leaf
; iter
; iter
= iter
->dst
.rt6_next
) {
752 * Search for duplicates
755 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
757 * Same priority level
760 (info
->nlh
->nlmsg_flags
& NLM_F_EXCL
))
763 if (rt_can_ecmp
== rt6_qualify_for_ecmp(iter
)) {
768 fallback_ins
= fallback_ins
?: ins
;
772 if (iter
->dst
.dev
== rt
->dst
.dev
&&
773 iter
->rt6i_idev
== rt
->rt6i_idev
&&
774 ipv6_addr_equal(&iter
->rt6i_gateway
,
775 &rt
->rt6i_gateway
)) {
776 if (rt
->rt6i_nsiblings
)
777 rt
->rt6i_nsiblings
= 0;
778 if (!(iter
->rt6i_flags
& RTF_EXPIRES
))
780 if (!(rt
->rt6i_flags
& RTF_EXPIRES
))
781 rt6_clean_expires(iter
);
783 rt6_set_expires(iter
, rt
->dst
.expires
);
784 iter
->rt6i_pmtu
= rt
->rt6i_pmtu
;
787 /* If we have the same destination and the same metric,
788 * but not the same gateway, then the route we try to
789 * add is sibling to this route, increment our counter
790 * of siblings, and later we will add our route to the
792 * Only static routes (which don't have flag
793 * RTF_EXPIRES) are used for ECMPv6.
795 * To avoid long list, we only had siblings if the
796 * route have a gateway.
799 rt6_qualify_for_ecmp(iter
))
800 rt
->rt6i_nsiblings
++;
803 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
807 ins
= &iter
->dst
.rt6_next
;
810 if (fallback_ins
&& !found
) {
811 /* No ECMP-able route found, replace first non-ECMP one */
817 /* Reset round-robin state, if necessary */
818 if (ins
== &fn
->leaf
)
821 /* Link this route to others same route. */
822 if (rt
->rt6i_nsiblings
) {
823 unsigned int rt6i_nsiblings
;
824 struct rt6_info
*sibling
, *temp_sibling
;
826 /* Find the first route that have the same metric */
829 if (sibling
->rt6i_metric
== rt
->rt6i_metric
&&
830 rt6_qualify_for_ecmp(sibling
)) {
831 list_add_tail(&rt
->rt6i_siblings
,
832 &sibling
->rt6i_siblings
);
835 sibling
= sibling
->dst
.rt6_next
;
837 /* For each sibling in the list, increment the counter of
838 * siblings. BUG() if counters does not match, list of siblings
842 list_for_each_entry_safe(sibling
, temp_sibling
,
843 &rt
->rt6i_siblings
, rt6i_siblings
) {
844 sibling
->rt6i_nsiblings
++;
845 BUG_ON(sibling
->rt6i_nsiblings
!= rt
->rt6i_nsiblings
);
848 BUG_ON(rt6i_nsiblings
!= rt
->rt6i_nsiblings
);
856 pr_warn("NLM_F_CREATE should be set when creating new route\n");
859 err
= fib6_commit_metrics(&rt
->dst
, mxc
);
863 rt
->dst
.rt6_next
= iter
;
866 atomic_inc(&rt
->rt6i_ref
);
867 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
, 0);
868 info
->nl_net
->ipv6
.rt6_stats
->fib_rt_entries
++;
870 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
871 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
872 fn
->fn_flags
|= RTN_RTINFO
;
881 pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
885 err
= fib6_commit_metrics(&rt
->dst
, mxc
);
891 rt
->dst
.rt6_next
= iter
->dst
.rt6_next
;
892 atomic_inc(&rt
->rt6i_ref
);
893 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
, NLM_F_REPLACE
);
894 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
895 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
896 fn
->fn_flags
|= RTN_RTINFO
;
898 nsiblings
= iter
->rt6i_nsiblings
;
899 fib6_purge_rt(iter
, fn
, info
->nl_net
);
903 /* Replacing an ECMP route, remove all siblings */
904 ins
= &rt
->dst
.rt6_next
;
907 if (rt6_qualify_for_ecmp(iter
)) {
908 *ins
= iter
->dst
.rt6_next
;
909 fib6_purge_rt(iter
, fn
, info
->nl_net
);
913 ins
= &iter
->dst
.rt6_next
;
917 WARN_ON(nsiblings
!= 0);
924 static void fib6_start_gc(struct net
*net
, struct rt6_info
*rt
)
926 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
) &&
927 (rt
->rt6i_flags
& (RTF_EXPIRES
| RTF_CACHE
)))
928 mod_timer(&net
->ipv6
.ip6_fib_timer
,
929 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
932 void fib6_force_start_gc(struct net
*net
)
934 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
))
935 mod_timer(&net
->ipv6
.ip6_fib_timer
,
936 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
940 * Add routing information to the routing tree.
941 * <destination addr>/<source addr>
942 * with source addr info in sub-trees
945 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
,
946 struct nl_info
*info
, struct mx6_config
*mxc
)
948 struct fib6_node
*fn
, *pn
= NULL
;
950 int allow_create
= 1;
951 int replace_required
= 0;
952 int sernum
= fib6_new_sernum(info
->nl_net
);
954 if (WARN_ON_ONCE((rt
->dst
.flags
& DST_NOCACHE
) &&
955 !atomic_read(&rt
->dst
.__refcnt
)))
959 if (!(info
->nlh
->nlmsg_flags
& NLM_F_CREATE
))
961 if (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
)
962 replace_required
= 1;
964 if (!allow_create
&& !replace_required
)
965 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
967 fn
= fib6_add_1(root
, &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
,
968 offsetof(struct rt6_info
, rt6i_dst
), allow_create
,
969 replace_required
, sernum
);
978 #ifdef CONFIG_IPV6_SUBTREES
979 if (rt
->rt6i_src
.plen
) {
980 struct fib6_node
*sn
;
983 struct fib6_node
*sfn
;
995 /* Create subtree root node */
1000 sfn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
1001 atomic_inc(&info
->nl_net
->ipv6
.ip6_null_entry
->rt6i_ref
);
1002 sfn
->fn_flags
= RTN_ROOT
;
1003 sfn
->fn_sernum
= sernum
;
1005 /* Now add the first leaf node to new subtree */
1007 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
1009 offsetof(struct rt6_info
, rt6i_src
),
1010 allow_create
, replace_required
, sernum
);
1013 /* If it is failed, discard just allocated
1014 root, and then (in st_failure) stale node
1022 /* Now link new subtree to main tree */
1026 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
1028 offsetof(struct rt6_info
, rt6i_src
),
1029 allow_create
, replace_required
, sernum
);
1039 atomic_inc(&rt
->rt6i_ref
);
1045 err
= fib6_add_rt2node(fn
, rt
, info
, mxc
);
1047 fib6_start_gc(info
->nl_net
, rt
);
1048 if (!(rt
->rt6i_flags
& RTF_CACHE
))
1049 fib6_prune_clones(info
->nl_net
, pn
);
1050 rt
->dst
.flags
&= ~DST_NOCACHE
;
1055 #ifdef CONFIG_IPV6_SUBTREES
1057 * If fib6_add_1 has cleared the old leaf pointer in the
1058 * super-tree leaf node we have to find a new one for it.
1060 if (pn
!= fn
&& pn
->leaf
== rt
) {
1062 atomic_dec(&rt
->rt6i_ref
);
1064 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
1065 pn
->leaf
= fib6_find_prefix(info
->nl_net
, pn
);
1068 WARN_ON(pn
->leaf
== NULL
);
1069 pn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
1072 atomic_inc(&pn
->leaf
->rt6i_ref
);
1075 if (!(rt
->dst
.flags
& DST_NOCACHE
))
1080 #ifdef CONFIG_IPV6_SUBTREES
1081 /* Subtree creation failed, probably main tree node
1082 is orphan. If it is, shoot it.
1085 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
1086 fib6_repair_tree(info
->nl_net
, fn
);
1087 if (!(rt
->dst
.flags
& DST_NOCACHE
))
1094 * Routing tree lookup
1098 struct lookup_args
{
1099 int offset
; /* key offset on rt6_info */
1100 const struct in6_addr
*addr
; /* search key */
1103 static struct fib6_node
*fib6_lookup_1(struct fib6_node
*root
,
1104 struct lookup_args
*args
)
1106 struct fib6_node
*fn
;
1109 if (unlikely(args
->offset
== 0))
1119 struct fib6_node
*next
;
1121 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
1123 next
= dir
? fn
->right
: fn
->left
;
1133 if (FIB6_SUBTREE(fn
) || fn
->fn_flags
& RTN_RTINFO
) {
1136 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
1139 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
)) {
1140 #ifdef CONFIG_IPV6_SUBTREES
1142 struct fib6_node
*sfn
;
1143 sfn
= fib6_lookup_1(fn
->subtree
,
1150 if (fn
->fn_flags
& RTN_RTINFO
)
1154 #ifdef CONFIG_IPV6_SUBTREES
1157 if (fn
->fn_flags
& RTN_ROOT
)
1166 struct fib6_node
*fib6_lookup(struct fib6_node
*root
, const struct in6_addr
*daddr
,
1167 const struct in6_addr
*saddr
)
1169 struct fib6_node
*fn
;
1170 struct lookup_args args
[] = {
1172 .offset
= offsetof(struct rt6_info
, rt6i_dst
),
1175 #ifdef CONFIG_IPV6_SUBTREES
1177 .offset
= offsetof(struct rt6_info
, rt6i_src
),
1182 .offset
= 0, /* sentinel */
1186 fn
= fib6_lookup_1(root
, daddr
? args
: args
+ 1);
1187 if (!fn
|| fn
->fn_flags
& RTN_TL_ROOT
)
1194 * Get node with specified destination prefix (and source prefix,
1195 * if subtrees are used)
1199 static struct fib6_node
*fib6_locate_1(struct fib6_node
*root
,
1200 const struct in6_addr
*addr
,
1201 int plen
, int offset
)
1203 struct fib6_node
*fn
;
1205 for (fn
= root
; fn
; ) {
1206 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
1211 if (plen
< fn
->fn_bit
||
1212 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
1215 if (plen
== fn
->fn_bit
)
1219 * We have more bits to go
1221 if (addr_bit_set(addr
, fn
->fn_bit
))
1229 struct fib6_node
*fib6_locate(struct fib6_node
*root
,
1230 const struct in6_addr
*daddr
, int dst_len
,
1231 const struct in6_addr
*saddr
, int src_len
)
1233 struct fib6_node
*fn
;
1235 fn
= fib6_locate_1(root
, daddr
, dst_len
,
1236 offsetof(struct rt6_info
, rt6i_dst
));
1238 #ifdef CONFIG_IPV6_SUBTREES
1240 WARN_ON(saddr
== NULL
);
1241 if (fn
&& fn
->subtree
)
1242 fn
= fib6_locate_1(fn
->subtree
, saddr
, src_len
,
1243 offsetof(struct rt6_info
, rt6i_src
));
1247 if (fn
&& fn
->fn_flags
& RTN_RTINFO
)
1259 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
)
1261 if (fn
->fn_flags
& RTN_ROOT
)
1262 return net
->ipv6
.ip6_null_entry
;
1266 return fn
->left
->leaf
;
1268 return fn
->right
->leaf
;
1270 fn
= FIB6_SUBTREE(fn
);
1276 * Called to trim the tree of intermediate nodes when possible. "fn"
1277 * is the node we want to try and remove.
1280 static struct fib6_node
*fib6_repair_tree(struct net
*net
,
1281 struct fib6_node
*fn
)
1285 struct fib6_node
*child
, *pn
;
1286 struct fib6_walker
*w
;
1290 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
1293 WARN_ON(fn
->fn_flags
& RTN_RTINFO
);
1294 WARN_ON(fn
->fn_flags
& RTN_TL_ROOT
);
1300 child
= fn
->right
, children
|= 1;
1302 child
= fn
->left
, children
|= 2;
1304 if (children
== 3 || FIB6_SUBTREE(fn
)
1305 #ifdef CONFIG_IPV6_SUBTREES
1306 /* Subtree root (i.e. fn) may have one child */
1307 || (children
&& fn
->fn_flags
& RTN_ROOT
)
1310 fn
->leaf
= fib6_find_prefix(net
, fn
);
1314 fn
->leaf
= net
->ipv6
.ip6_null_entry
;
1317 atomic_inc(&fn
->leaf
->rt6i_ref
);
1322 #ifdef CONFIG_IPV6_SUBTREES
1323 if (FIB6_SUBTREE(pn
) == fn
) {
1324 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1325 FIB6_SUBTREE(pn
) = NULL
;
1328 WARN_ON(fn
->fn_flags
& RTN_ROOT
);
1330 if (pn
->right
== fn
)
1332 else if (pn
->left
== fn
)
1341 #ifdef CONFIG_IPV6_SUBTREES
1345 read_lock(&net
->ipv6
.fib6_walker_lock
);
1346 FOR_WALKERS(net
, w
) {
1348 if (w
->root
== fn
) {
1349 w
->root
= w
->node
= NULL
;
1350 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1351 } else if (w
->node
== fn
) {
1352 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1357 if (w
->root
== fn
) {
1359 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1361 if (w
->node
== fn
) {
1364 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1365 w
->state
= w
->state
>= FWS_R
? FWS_U
: FWS_INIT
;
1367 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1368 w
->state
= w
->state
>= FWS_C
? FWS_U
: FWS_INIT
;
1373 read_unlock(&net
->ipv6
.fib6_walker_lock
);
1376 if (pn
->fn_flags
& RTN_RTINFO
|| FIB6_SUBTREE(pn
))
1379 rt6_release(pn
->leaf
);
1385 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1386 struct nl_info
*info
)
1388 struct fib6_walker
*w
;
1389 struct rt6_info
*rt
= *rtp
;
1390 struct net
*net
= info
->nl_net
;
1392 RT6_TRACE("fib6_del_route\n");
1395 *rtp
= rt
->dst
.rt6_next
;
1396 rt
->rt6i_node
= NULL
;
1397 net
->ipv6
.rt6_stats
->fib_rt_entries
--;
1398 net
->ipv6
.rt6_stats
->fib_discarded_routes
++;
1400 /* Reset round-robin state, if necessary */
1401 if (fn
->rr_ptr
== rt
)
1404 /* Remove this entry from other siblings */
1405 if (rt
->rt6i_nsiblings
) {
1406 struct rt6_info
*sibling
, *next_sibling
;
1408 list_for_each_entry_safe(sibling
, next_sibling
,
1409 &rt
->rt6i_siblings
, rt6i_siblings
)
1410 sibling
->rt6i_nsiblings
--;
1411 rt
->rt6i_nsiblings
= 0;
1412 list_del_init(&rt
->rt6i_siblings
);
1415 /* Adjust walkers */
1416 read_lock(&net
->ipv6
.fib6_walker_lock
);
1417 FOR_WALKERS(net
, w
) {
1418 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1419 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1420 w
->leaf
= rt
->dst
.rt6_next
;
1425 read_unlock(&net
->ipv6
.fib6_walker_lock
);
1427 rt
->dst
.rt6_next
= NULL
;
1429 /* If it was last route, expunge its radix tree node */
1431 fn
->fn_flags
&= ~RTN_RTINFO
;
1432 net
->ipv6
.rt6_stats
->fib_route_nodes
--;
1433 fn
= fib6_repair_tree(net
, fn
);
1436 fib6_purge_rt(rt
, fn
, net
);
1438 inet6_rt_notify(RTM_DELROUTE
, rt
, info
, 0);
1442 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1444 struct net
*net
= info
->nl_net
;
1445 struct fib6_node
*fn
= rt
->rt6i_node
;
1446 struct rt6_info
**rtp
;
1449 if (rt
->dst
.obsolete
> 0) {
1454 if (!fn
|| rt
== net
->ipv6
.ip6_null_entry
)
1457 WARN_ON(!(fn
->fn_flags
& RTN_RTINFO
));
1459 if (!(rt
->rt6i_flags
& RTF_CACHE
)) {
1460 struct fib6_node
*pn
= fn
;
1461 #ifdef CONFIG_IPV6_SUBTREES
1462 /* clones of this route might be in another subtree */
1463 if (rt
->rt6i_src
.plen
) {
1464 while (!(pn
->fn_flags
& RTN_ROOT
))
1469 fib6_prune_clones(info
->nl_net
, pn
);
1473 * Walk the leaf entries looking for ourself
1476 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->dst
.rt6_next
) {
1478 fib6_del_route(fn
, rtp
, info
);
1486 * Tree traversal function.
1488 * Certainly, it is not interrupt safe.
1489 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1490 * It means, that we can modify tree during walking
1491 * and use this function for garbage collection, clone pruning,
1492 * cleaning tree when a device goes down etc. etc.
1494 * It guarantees that every node will be traversed,
1495 * and that it will be traversed only once.
1497 * Callback function w->func may return:
1498 * 0 -> continue walking.
1499 * positive value -> walking is suspended (used by tree dumps,
1500 * and probably by gc, if it will be split to several slices)
1501 * negative value -> terminate walking.
1503 * The function itself returns:
1504 * 0 -> walk is complete.
1505 * >0 -> walk is incomplete (i.e. suspended)
1506 * <0 -> walk is terminated by an error.
1509 static int fib6_walk_continue(struct fib6_walker
*w
)
1511 struct fib6_node
*fn
, *pn
;
1518 if (w
->prune
&& fn
!= w
->root
&&
1519 fn
->fn_flags
& RTN_RTINFO
&& w
->state
< FWS_C
) {
1524 #ifdef CONFIG_IPV6_SUBTREES
1526 if (FIB6_SUBTREE(fn
)) {
1527 w
->node
= FIB6_SUBTREE(fn
);
1535 w
->state
= FWS_INIT
;
1541 w
->node
= fn
->right
;
1542 w
->state
= FWS_INIT
;
1548 if (w
->leaf
&& fn
->fn_flags
& RTN_RTINFO
) {
1570 #ifdef CONFIG_IPV6_SUBTREES
1571 if (FIB6_SUBTREE(pn
) == fn
) {
1572 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1577 if (pn
->left
== fn
) {
1581 if (pn
->right
== fn
) {
1583 w
->leaf
= w
->node
->leaf
;
1593 static int fib6_walk(struct net
*net
, struct fib6_walker
*w
)
1597 w
->state
= FWS_INIT
;
1600 fib6_walker_link(net
, w
);
1601 res
= fib6_walk_continue(w
);
1603 fib6_walker_unlink(net
, w
);
1607 static int fib6_clean_node(struct fib6_walker
*w
)
1610 struct rt6_info
*rt
;
1611 struct fib6_cleaner
*c
= container_of(w
, struct fib6_cleaner
, w
);
1612 struct nl_info info
= {
1616 if (c
->sernum
!= FIB6_NO_SERNUM_CHANGE
&&
1617 w
->node
->fn_sernum
!= c
->sernum
)
1618 w
->node
->fn_sernum
= c
->sernum
;
1621 WARN_ON_ONCE(c
->sernum
== FIB6_NO_SERNUM_CHANGE
);
1626 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1627 res
= c
->func(rt
, c
->arg
);
1630 res
= fib6_del(rt
, &info
);
1633 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
1634 __func__
, rt
, rt
->rt6i_node
, res
);
1647 * Convenient frontend to tree walker.
1649 * func is called on each route.
1650 * It may return -1 -> delete this route.
1651 * 0 -> continue walking
1653 * prune==1 -> only immediate children of node (certainly,
1654 * ignoring pure split nodes) will be scanned.
1657 static void fib6_clean_tree(struct net
*net
, struct fib6_node
*root
,
1658 int (*func
)(struct rt6_info
*, void *arg
),
1659 bool prune
, int sernum
, void *arg
)
1661 struct fib6_cleaner c
;
1664 c
.w
.func
= fib6_clean_node
;
1673 fib6_walk(net
, &c
.w
);
1676 static void __fib6_clean_all(struct net
*net
,
1677 int (*func
)(struct rt6_info
*, void *),
1678 int sernum
, void *arg
)
1680 struct fib6_table
*table
;
1681 struct hlist_head
*head
;
1685 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1686 head
= &net
->ipv6
.fib_table_hash
[h
];
1687 hlist_for_each_entry_rcu(table
, head
, tb6_hlist
) {
1688 write_lock_bh(&table
->tb6_lock
);
1689 fib6_clean_tree(net
, &table
->tb6_root
,
1690 func
, false, sernum
, arg
);
1691 write_unlock_bh(&table
->tb6_lock
);
1697 void fib6_clean_all(struct net
*net
, int (*func
)(struct rt6_info
*, void *),
1700 __fib6_clean_all(net
, func
, FIB6_NO_SERNUM_CHANGE
, arg
);
1703 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1705 if (rt
->rt6i_flags
& RTF_CACHE
) {
1706 RT6_TRACE("pruning clone %p\n", rt
);
1713 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
)
1715 fib6_clean_tree(net
, fn
, fib6_prune_clone
, true,
1716 FIB6_NO_SERNUM_CHANGE
, NULL
);
1719 static void fib6_flush_trees(struct net
*net
)
1721 int new_sernum
= fib6_new_sernum(net
);
1723 __fib6_clean_all(net
, NULL
, new_sernum
, NULL
);
1727 * Garbage collection
1736 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1738 struct fib6_gc_args
*gc_args
= arg
;
1739 unsigned long now
= jiffies
;
1742 * check addrconf expiration here.
1743 * Routes are expired even if they are in use.
1745 * Also age clones. Note, that clones are aged out
1746 * only if they are not in use now.
1749 if (rt
->rt6i_flags
& RTF_EXPIRES
&& rt
->dst
.expires
) {
1750 if (time_after(now
, rt
->dst
.expires
)) {
1751 RT6_TRACE("expiring %p\n", rt
);
1755 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1756 if (atomic_read(&rt
->dst
.__refcnt
) == 0 &&
1757 time_after_eq(now
, rt
->dst
.lastuse
+ gc_args
->timeout
)) {
1758 RT6_TRACE("aging clone %p\n", rt
);
1760 } else if (rt
->rt6i_flags
& RTF_GATEWAY
) {
1761 struct neighbour
*neigh
;
1762 __u8 neigh_flags
= 0;
1764 neigh
= dst_neigh_lookup(&rt
->dst
, &rt
->rt6i_gateway
);
1766 neigh_flags
= neigh
->flags
;
1767 neigh_release(neigh
);
1769 if (!(neigh_flags
& NTF_ROUTER
)) {
1770 RT6_TRACE("purging route %p via non-router but gateway\n",
1781 void fib6_run_gc(unsigned long expires
, struct net
*net
, bool force
)
1783 struct fib6_gc_args gc_args
;
1787 spin_lock_bh(&net
->ipv6
.fib6_gc_lock
);
1788 } else if (!spin_trylock_bh(&net
->ipv6
.fib6_gc_lock
)) {
1789 mod_timer(&net
->ipv6
.ip6_fib_timer
, jiffies
+ HZ
);
1792 gc_args
.timeout
= expires
? (int)expires
:
1793 net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1795 gc_args
.more
= icmp6_dst_gc();
1797 fib6_clean_all(net
, fib6_age
, &gc_args
);
1799 net
->ipv6
.ip6_rt_last_gc
= now
;
1802 mod_timer(&net
->ipv6
.ip6_fib_timer
,
1804 + net
->ipv6
.sysctl
.ip6_rt_gc_interval
));
1806 del_timer(&net
->ipv6
.ip6_fib_timer
);
1807 spin_unlock_bh(&net
->ipv6
.fib6_gc_lock
);
1810 static void fib6_gc_timer_cb(unsigned long arg
)
1812 fib6_run_gc(0, (struct net
*)arg
, true);
1815 static int __net_init
fib6_net_init(struct net
*net
)
1817 size_t size
= sizeof(struct hlist_head
) * FIB6_TABLE_HASHSZ
;
1819 spin_lock_init(&net
->ipv6
.fib6_gc_lock
);
1820 rwlock_init(&net
->ipv6
.fib6_walker_lock
);
1821 INIT_LIST_HEAD(&net
->ipv6
.fib6_walkers
);
1822 setup_timer(&net
->ipv6
.ip6_fib_timer
, fib6_gc_timer_cb
, (unsigned long)net
);
1824 net
->ipv6
.rt6_stats
= kzalloc(sizeof(*net
->ipv6
.rt6_stats
), GFP_KERNEL
);
1825 if (!net
->ipv6
.rt6_stats
)
1828 /* Avoid false sharing : Use at least a full cache line */
1829 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1831 net
->ipv6
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1832 if (!net
->ipv6
.fib_table_hash
)
1835 net
->ipv6
.fib6_main_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_main_tbl
),
1837 if (!net
->ipv6
.fib6_main_tbl
)
1838 goto out_fib_table_hash
;
1840 net
->ipv6
.fib6_main_tbl
->tb6_id
= RT6_TABLE_MAIN
;
1841 net
->ipv6
.fib6_main_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1842 net
->ipv6
.fib6_main_tbl
->tb6_root
.fn_flags
=
1843 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1844 inet_peer_base_init(&net
->ipv6
.fib6_main_tbl
->tb6_peers
);
1846 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1847 net
->ipv6
.fib6_local_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_local_tbl
),
1849 if (!net
->ipv6
.fib6_local_tbl
)
1850 goto out_fib6_main_tbl
;
1851 net
->ipv6
.fib6_local_tbl
->tb6_id
= RT6_TABLE_LOCAL
;
1852 net
->ipv6
.fib6_local_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1853 net
->ipv6
.fib6_local_tbl
->tb6_root
.fn_flags
=
1854 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1855 inet_peer_base_init(&net
->ipv6
.fib6_local_tbl
->tb6_peers
);
1857 fib6_tables_init(net
);
1861 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1863 kfree(net
->ipv6
.fib6_main_tbl
);
1866 kfree(net
->ipv6
.fib_table_hash
);
1868 kfree(net
->ipv6
.rt6_stats
);
1873 static void fib6_net_exit(struct net
*net
)
1875 rt6_ifdown(net
, NULL
);
1876 del_timer_sync(&net
->ipv6
.ip6_fib_timer
);
1878 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1879 inetpeer_invalidate_tree(&net
->ipv6
.fib6_local_tbl
->tb6_peers
);
1880 kfree(net
->ipv6
.fib6_local_tbl
);
1882 inetpeer_invalidate_tree(&net
->ipv6
.fib6_main_tbl
->tb6_peers
);
1883 kfree(net
->ipv6
.fib6_main_tbl
);
1884 kfree(net
->ipv6
.fib_table_hash
);
1885 kfree(net
->ipv6
.rt6_stats
);
1888 static struct pernet_operations fib6_net_ops
= {
1889 .init
= fib6_net_init
,
1890 .exit
= fib6_net_exit
,
1893 int __init
fib6_init(void)
1897 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1898 sizeof(struct fib6_node
),
1899 0, SLAB_HWCACHE_ALIGN
,
1901 if (!fib6_node_kmem
)
1904 ret
= register_pernet_subsys(&fib6_net_ops
);
1906 goto out_kmem_cache_create
;
1908 ret
= __rtnl_register(PF_INET6
, RTM_GETROUTE
, NULL
, inet6_dump_fib
,
1911 goto out_unregister_subsys
;
1913 __fib6_flush_trees
= fib6_flush_trees
;
1917 out_unregister_subsys
:
1918 unregister_pernet_subsys(&fib6_net_ops
);
1919 out_kmem_cache_create
:
1920 kmem_cache_destroy(fib6_node_kmem
);
1924 void fib6_gc_cleanup(void)
1926 unregister_pernet_subsys(&fib6_net_ops
);
1927 kmem_cache_destroy(fib6_node_kmem
);
1930 #ifdef CONFIG_PROC_FS
1932 struct ipv6_route_iter
{
1933 struct seq_net_private p
;
1934 struct fib6_walker w
;
1936 struct fib6_table
*tbl
;
1940 static int ipv6_route_seq_show(struct seq_file
*seq
, void *v
)
1942 struct rt6_info
*rt
= v
;
1943 struct ipv6_route_iter
*iter
= seq
->private;
1945 seq_printf(seq
, "%pi6 %02x ", &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
);
1947 #ifdef CONFIG_IPV6_SUBTREES
1948 seq_printf(seq
, "%pi6 %02x ", &rt
->rt6i_src
.addr
, rt
->rt6i_src
.plen
);
1950 seq_puts(seq
, "00000000000000000000000000000000 00 ");
1952 if (rt
->rt6i_flags
& RTF_GATEWAY
)
1953 seq_printf(seq
, "%pi6", &rt
->rt6i_gateway
);
1955 seq_puts(seq
, "00000000000000000000000000000000");
1957 seq_printf(seq
, " %08x %08x %08x %08x %8s\n",
1958 rt
->rt6i_metric
, atomic_read(&rt
->dst
.__refcnt
),
1959 rt
->dst
.__use
, rt
->rt6i_flags
,
1960 rt
->dst
.dev
? rt
->dst
.dev
->name
: "");
1961 iter
->w
.leaf
= NULL
;
1965 static int ipv6_route_yield(struct fib6_walker
*w
)
1967 struct ipv6_route_iter
*iter
= w
->args
;
1973 iter
->w
.leaf
= iter
->w
.leaf
->dst
.rt6_next
;
1975 if (!iter
->skip
&& iter
->w
.leaf
)
1977 } while (iter
->w
.leaf
);
1982 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter
*iter
,
1985 memset(&iter
->w
, 0, sizeof(iter
->w
));
1986 iter
->w
.func
= ipv6_route_yield
;
1987 iter
->w
.root
= &iter
->tbl
->tb6_root
;
1988 iter
->w
.state
= FWS_INIT
;
1989 iter
->w
.node
= iter
->w
.root
;
1990 iter
->w
.args
= iter
;
1991 iter
->sernum
= iter
->w
.root
->fn_sernum
;
1992 INIT_LIST_HEAD(&iter
->w
.lh
);
1993 fib6_walker_link(net
, &iter
->w
);
1996 static struct fib6_table
*ipv6_route_seq_next_table(struct fib6_table
*tbl
,
2000 struct hlist_node
*node
;
2003 h
= (tbl
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1)) + 1;
2004 node
= rcu_dereference_bh(hlist_next_rcu(&tbl
->tb6_hlist
));
2010 while (!node
&& h
< FIB6_TABLE_HASHSZ
) {
2011 node
= rcu_dereference_bh(
2012 hlist_first_rcu(&net
->ipv6
.fib_table_hash
[h
++]));
2014 return hlist_entry_safe(node
, struct fib6_table
, tb6_hlist
);
2017 static void ipv6_route_check_sernum(struct ipv6_route_iter
*iter
)
2019 if (iter
->sernum
!= iter
->w
.root
->fn_sernum
) {
2020 iter
->sernum
= iter
->w
.root
->fn_sernum
;
2021 iter
->w
.state
= FWS_INIT
;
2022 iter
->w
.node
= iter
->w
.root
;
2023 WARN_ON(iter
->w
.skip
);
2024 iter
->w
.skip
= iter
->w
.count
;
2028 static void *ipv6_route_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2032 struct net
*net
= seq_file_net(seq
);
2033 struct ipv6_route_iter
*iter
= seq
->private;
2038 n
= ((struct rt6_info
*)v
)->dst
.rt6_next
;
2045 ipv6_route_check_sernum(iter
);
2046 read_lock(&iter
->tbl
->tb6_lock
);
2047 r
= fib6_walk_continue(&iter
->w
);
2048 read_unlock(&iter
->tbl
->tb6_lock
);
2052 return iter
->w
.leaf
;
2054 fib6_walker_unlink(net
, &iter
->w
);
2057 fib6_walker_unlink(net
, &iter
->w
);
2059 iter
->tbl
= ipv6_route_seq_next_table(iter
->tbl
, net
);
2063 ipv6_route_seq_setup_walk(iter
, net
);
2067 static void *ipv6_route_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2070 struct net
*net
= seq_file_net(seq
);
2071 struct ipv6_route_iter
*iter
= seq
->private;
2074 iter
->tbl
= ipv6_route_seq_next_table(NULL
, net
);
2078 ipv6_route_seq_setup_walk(iter
, net
);
2079 return ipv6_route_seq_next(seq
, NULL
, pos
);
2085 static bool ipv6_route_iter_active(struct ipv6_route_iter
*iter
)
2087 struct fib6_walker
*w
= &iter
->w
;
2088 return w
->node
&& !(w
->state
== FWS_U
&& w
->node
== w
->root
);
2091 static void ipv6_route_seq_stop(struct seq_file
*seq
, void *v
)
2094 struct net
*net
= seq_file_net(seq
);
2095 struct ipv6_route_iter
*iter
= seq
->private;
2097 if (ipv6_route_iter_active(iter
))
2098 fib6_walker_unlink(net
, &iter
->w
);
2100 rcu_read_unlock_bh();
2103 static const struct seq_operations ipv6_route_seq_ops
= {
2104 .start
= ipv6_route_seq_start
,
2105 .next
= ipv6_route_seq_next
,
2106 .stop
= ipv6_route_seq_stop
,
2107 .show
= ipv6_route_seq_show
2110 int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2112 return seq_open_net(inode
, file
, &ipv6_route_seq_ops
,
2113 sizeof(struct ipv6_route_iter
));
2116 #endif /* CONFIG_PROC_FS */