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 */
322 /* Multipath routes are dumped in one route with the
323 * RTA_MULTIPATH attribute. Jump 'rt' to point to the
324 * last sibling of this route (no need to dump the
325 * sibling routes again)
327 if (rt
->rt6i_nsiblings
)
328 rt
= list_last_entry(&rt
->rt6i_siblings
,
336 static void fib6_dump_end(struct netlink_callback
*cb
)
338 struct net
*net
= sock_net(cb
->skb
->sk
);
339 struct fib6_walker
*w
= (void *)cb
->args
[2];
344 fib6_walker_unlink(net
, w
);
349 cb
->done
= (void *)cb
->args
[3];
353 static int fib6_dump_done(struct netlink_callback
*cb
)
356 return cb
->done
? cb
->done(cb
) : 0;
359 static int fib6_dump_table(struct fib6_table
*table
, struct sk_buff
*skb
,
360 struct netlink_callback
*cb
)
362 struct net
*net
= sock_net(skb
->sk
);
363 struct fib6_walker
*w
;
366 w
= (void *)cb
->args
[2];
367 w
->root
= &table
->tb6_root
;
369 if (cb
->args
[4] == 0) {
373 read_lock_bh(&table
->tb6_lock
);
374 res
= fib6_walk(net
, w
);
375 read_unlock_bh(&table
->tb6_lock
);
378 cb
->args
[5] = w
->root
->fn_sernum
;
381 if (cb
->args
[5] != w
->root
->fn_sernum
) {
382 /* Begin at the root if the tree changed */
383 cb
->args
[5] = w
->root
->fn_sernum
;
390 read_lock_bh(&table
->tb6_lock
);
391 res
= fib6_walk_continue(w
);
392 read_unlock_bh(&table
->tb6_lock
);
394 fib6_walker_unlink(net
, w
);
402 static int inet6_dump_fib(struct sk_buff
*skb
, struct netlink_callback
*cb
)
404 struct net
*net
= sock_net(skb
->sk
);
406 unsigned int e
= 0, s_e
;
407 struct rt6_rtnl_dump_arg arg
;
408 struct fib6_walker
*w
;
409 struct fib6_table
*tb
;
410 struct hlist_head
*head
;
416 w
= (void *)cb
->args
[2];
420 * 1. hook callback destructor.
422 cb
->args
[3] = (long)cb
->done
;
423 cb
->done
= fib6_dump_done
;
426 * 2. allocate and initialize walker.
428 w
= kzalloc(sizeof(*w
), GFP_ATOMIC
);
431 w
->func
= fib6_dump_node
;
432 cb
->args
[2] = (long)w
;
441 for (h
= s_h
; h
< FIB6_TABLE_HASHSZ
; h
++, s_e
= 0) {
443 head
= &net
->ipv6
.fib_table_hash
[h
];
444 hlist_for_each_entry_rcu(tb
, head
, tb6_hlist
) {
447 res
= fib6_dump_table(tb
, skb
, cb
);
459 res
= res
< 0 ? res
: skb
->len
;
468 * return the appropriate node for a routing tree "add" operation
469 * by either creating and inserting or by returning an existing
473 static struct fib6_node
*fib6_add_1(struct fib6_node
*root
,
474 struct in6_addr
*addr
, int plen
,
475 int offset
, int allow_create
,
476 int replace_required
, int sernum
)
478 struct fib6_node
*fn
, *in
, *ln
;
479 struct fib6_node
*pn
= NULL
;
484 RT6_TRACE("fib6_add_1\n");
486 /* insert node in tree */
491 key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
496 if (plen
< fn
->fn_bit
||
497 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
)) {
499 if (replace_required
) {
500 pr_warn("Can't replace route, no match found\n");
501 return ERR_PTR(-ENOENT
);
503 pr_warn("NLM_F_CREATE should be set when creating new route\n");
512 if (plen
== fn
->fn_bit
) {
513 /* clean up an intermediate node */
514 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
515 rt6_release(fn
->leaf
);
519 fn
->fn_sernum
= sernum
;
525 * We have more bits to go
528 /* Try to walk down on tree. */
529 fn
->fn_sernum
= sernum
;
530 dir
= addr_bit_set(addr
, fn
->fn_bit
);
532 fn
= dir
? fn
->right
: fn
->left
;
536 /* We should not create new node because
537 * NLM_F_REPLACE was specified without NLM_F_CREATE
538 * I assume it is safe to require NLM_F_CREATE when
539 * REPLACE flag is used! Later we may want to remove the
540 * check for replace_required, because according
541 * to netlink specification, NLM_F_CREATE
542 * MUST be specified if new route is created.
543 * That would keep IPv6 consistent with IPv4
545 if (replace_required
) {
546 pr_warn("Can't replace route, no match found\n");
547 return ERR_PTR(-ENOENT
);
549 pr_warn("NLM_F_CREATE should be set when creating new route\n");
552 * We walked to the bottom of tree.
553 * Create new leaf node without children.
559 return ERR_PTR(-ENOMEM
);
563 ln
->fn_sernum
= sernum
;
575 * split since we don't have a common prefix anymore or
576 * we have a less significant route.
577 * we've to insert an intermediate node on the list
578 * this new node will point to the one we need to create
584 /* find 1st bit in difference between the 2 addrs.
586 See comment in __ipv6_addr_diff: bit may be an invalid value,
587 but if it is >= plen, the value is ignored in any case.
590 bit
= __ipv6_addr_diff(addr
, &key
->addr
, sizeof(*addr
));
595 * (new leaf node)[ln] (old node)[fn]
606 return ERR_PTR(-ENOMEM
);
610 * new intermediate node.
612 * be off since that an address that chooses one of
613 * the branches would not match less specific routes
614 * in the other branch
621 atomic_inc(&in
->leaf
->rt6i_ref
);
623 in
->fn_sernum
= sernum
;
625 /* update parent pointer */
636 ln
->fn_sernum
= sernum
;
638 if (addr_bit_set(addr
, bit
)) {
645 } else { /* plen <= bit */
648 * (new leaf node)[ln]
650 * (old node)[fn] NULL
656 return ERR_PTR(-ENOMEM
);
662 ln
->fn_sernum
= sernum
;
669 if (addr_bit_set(&key
->addr
, plen
))
679 static bool rt6_qualify_for_ecmp(struct rt6_info
*rt
)
681 return (rt
->rt6i_flags
& (RTF_GATEWAY
|RTF_ADDRCONF
|RTF_DYNAMIC
)) ==
685 static void fib6_copy_metrics(u32
*mp
, const struct mx6_config
*mxc
)
689 for (i
= 0; i
< RTAX_MAX
; i
++) {
690 if (test_bit(i
, mxc
->mx_valid
))
695 static int fib6_commit_metrics(struct dst_entry
*dst
, struct mx6_config
*mxc
)
700 if (dst
->flags
& DST_HOST
) {
701 u32
*mp
= dst_metrics_write_ptr(dst
);
706 fib6_copy_metrics(mp
, mxc
);
708 dst_init_metrics(dst
, mxc
->mx
, false);
710 /* We've stolen mx now. */
717 static void fib6_purge_rt(struct rt6_info
*rt
, struct fib6_node
*fn
,
720 if (atomic_read(&rt
->rt6i_ref
) != 1) {
721 /* This route is used as dummy address holder in some split
722 * nodes. It is not leaked, but it still holds other resources,
723 * which must be released in time. So, scan ascendant nodes
724 * and replace dummy references to this route with references
725 * to still alive ones.
728 if (!(fn
->fn_flags
& RTN_RTINFO
) && fn
->leaf
== rt
) {
729 fn
->leaf
= fib6_find_prefix(net
, fn
);
730 atomic_inc(&fn
->leaf
->rt6i_ref
);
735 /* No more references are possible at this point. */
736 BUG_ON(atomic_read(&rt
->rt6i_ref
) != 1);
741 * Insert routing information in a node.
744 static int fib6_add_rt2node(struct fib6_node
*fn
, struct rt6_info
*rt
,
745 struct nl_info
*info
, struct mx6_config
*mxc
)
747 struct rt6_info
*iter
= NULL
;
748 struct rt6_info
**ins
;
749 struct rt6_info
**fallback_ins
= NULL
;
750 int replace
= (info
->nlh
&&
751 (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
));
752 int add
= (!info
->nlh
||
753 (info
->nlh
->nlmsg_flags
& NLM_F_CREATE
));
755 bool rt_can_ecmp
= rt6_qualify_for_ecmp(rt
);
756 u16 nlflags
= NLM_F_EXCL
;
759 if (info
->nlh
&& (info
->nlh
->nlmsg_flags
& NLM_F_APPEND
))
760 nlflags
|= NLM_F_APPEND
;
764 for (iter
= fn
->leaf
; iter
; iter
= iter
->dst
.rt6_next
) {
766 * Search for duplicates
769 if (iter
->rt6i_metric
== rt
->rt6i_metric
) {
771 * Same priority level
774 (info
->nlh
->nlmsg_flags
& NLM_F_EXCL
))
777 nlflags
&= ~NLM_F_EXCL
;
779 if (rt_can_ecmp
== rt6_qualify_for_ecmp(iter
)) {
784 fallback_ins
= fallback_ins
?: ins
;
788 if (iter
->dst
.dev
== rt
->dst
.dev
&&
789 iter
->rt6i_idev
== rt
->rt6i_idev
&&
790 ipv6_addr_equal(&iter
->rt6i_gateway
,
791 &rt
->rt6i_gateway
)) {
792 if (rt
->rt6i_nsiblings
)
793 rt
->rt6i_nsiblings
= 0;
794 if (!(iter
->rt6i_flags
& RTF_EXPIRES
))
796 if (!(rt
->rt6i_flags
& RTF_EXPIRES
))
797 rt6_clean_expires(iter
);
799 rt6_set_expires(iter
, rt
->dst
.expires
);
800 iter
->rt6i_pmtu
= rt
->rt6i_pmtu
;
803 /* If we have the same destination and the same metric,
804 * but not the same gateway, then the route we try to
805 * add is sibling to this route, increment our counter
806 * of siblings, and later we will add our route to the
808 * Only static routes (which don't have flag
809 * RTF_EXPIRES) are used for ECMPv6.
811 * To avoid long list, we only had siblings if the
812 * route have a gateway.
815 rt6_qualify_for_ecmp(iter
))
816 rt
->rt6i_nsiblings
++;
819 if (iter
->rt6i_metric
> rt
->rt6i_metric
)
823 ins
= &iter
->dst
.rt6_next
;
826 if (fallback_ins
&& !found
) {
827 /* No ECMP-able route found, replace first non-ECMP one */
833 /* Reset round-robin state, if necessary */
834 if (ins
== &fn
->leaf
)
837 /* Link this route to others same route. */
838 if (rt
->rt6i_nsiblings
) {
839 unsigned int rt6i_nsiblings
;
840 struct rt6_info
*sibling
, *temp_sibling
;
842 /* Find the first route that have the same metric */
845 if (sibling
->rt6i_metric
== rt
->rt6i_metric
&&
846 rt6_qualify_for_ecmp(sibling
)) {
847 list_add_tail(&rt
->rt6i_siblings
,
848 &sibling
->rt6i_siblings
);
851 sibling
= sibling
->dst
.rt6_next
;
853 /* For each sibling in the list, increment the counter of
854 * siblings. BUG() if counters does not match, list of siblings
858 list_for_each_entry_safe(sibling
, temp_sibling
,
859 &rt
->rt6i_siblings
, rt6i_siblings
) {
860 sibling
->rt6i_nsiblings
++;
861 BUG_ON(sibling
->rt6i_nsiblings
!= rt
->rt6i_nsiblings
);
864 BUG_ON(rt6i_nsiblings
!= rt
->rt6i_nsiblings
);
872 pr_warn("NLM_F_CREATE should be set when creating new route\n");
875 nlflags
|= NLM_F_CREATE
;
876 err
= fib6_commit_metrics(&rt
->dst
, mxc
);
880 rt
->dst
.rt6_next
= iter
;
883 atomic_inc(&rt
->rt6i_ref
);
884 if (!info
->skip_notify
)
885 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
, nlflags
);
886 info
->nl_net
->ipv6
.rt6_stats
->fib_rt_entries
++;
888 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
889 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
890 fn
->fn_flags
|= RTN_RTINFO
;
899 pr_warn("NLM_F_REPLACE set, but no existing node found!\n");
903 err
= fib6_commit_metrics(&rt
->dst
, mxc
);
909 rt
->dst
.rt6_next
= iter
->dst
.rt6_next
;
910 atomic_inc(&rt
->rt6i_ref
);
911 if (!info
->skip_notify
)
912 inet6_rt_notify(RTM_NEWROUTE
, rt
, info
, NLM_F_REPLACE
);
913 if (!(fn
->fn_flags
& RTN_RTINFO
)) {
914 info
->nl_net
->ipv6
.rt6_stats
->fib_route_nodes
++;
915 fn
->fn_flags
|= RTN_RTINFO
;
917 nsiblings
= iter
->rt6i_nsiblings
;
918 fib6_purge_rt(iter
, fn
, info
->nl_net
);
922 /* Replacing an ECMP route, remove all siblings */
923 ins
= &rt
->dst
.rt6_next
;
926 if (rt6_qualify_for_ecmp(iter
)) {
927 *ins
= iter
->dst
.rt6_next
;
928 fib6_purge_rt(iter
, fn
, info
->nl_net
);
932 ins
= &iter
->dst
.rt6_next
;
936 WARN_ON(nsiblings
!= 0);
943 static void fib6_start_gc(struct net
*net
, struct rt6_info
*rt
)
945 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
) &&
946 (rt
->rt6i_flags
& (RTF_EXPIRES
| RTF_CACHE
)))
947 mod_timer(&net
->ipv6
.ip6_fib_timer
,
948 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
951 void fib6_force_start_gc(struct net
*net
)
953 if (!timer_pending(&net
->ipv6
.ip6_fib_timer
))
954 mod_timer(&net
->ipv6
.ip6_fib_timer
,
955 jiffies
+ net
->ipv6
.sysctl
.ip6_rt_gc_interval
);
959 * Add routing information to the routing tree.
960 * <destination addr>/<source addr>
961 * with source addr info in sub-trees
964 int fib6_add(struct fib6_node
*root
, struct rt6_info
*rt
,
965 struct nl_info
*info
, struct mx6_config
*mxc
)
967 struct fib6_node
*fn
, *pn
= NULL
;
969 int allow_create
= 1;
970 int replace_required
= 0;
971 int sernum
= fib6_new_sernum(info
->nl_net
);
973 if (WARN_ON_ONCE((rt
->dst
.flags
& DST_NOCACHE
) &&
974 !atomic_read(&rt
->dst
.__refcnt
)))
978 if (!(info
->nlh
->nlmsg_flags
& NLM_F_CREATE
))
980 if (info
->nlh
->nlmsg_flags
& NLM_F_REPLACE
)
981 replace_required
= 1;
983 if (!allow_create
&& !replace_required
)
984 pr_warn("RTM_NEWROUTE with no NLM_F_CREATE or NLM_F_REPLACE\n");
986 fn
= fib6_add_1(root
, &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
,
987 offsetof(struct rt6_info
, rt6i_dst
), allow_create
,
988 replace_required
, sernum
);
997 #ifdef CONFIG_IPV6_SUBTREES
998 if (rt
->rt6i_src
.plen
) {
999 struct fib6_node
*sn
;
1002 struct fib6_node
*sfn
;
1014 /* Create subtree root node */
1019 sfn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
1020 atomic_inc(&info
->nl_net
->ipv6
.ip6_null_entry
->rt6i_ref
);
1021 sfn
->fn_flags
= RTN_ROOT
;
1022 sfn
->fn_sernum
= sernum
;
1024 /* Now add the first leaf node to new subtree */
1026 sn
= fib6_add_1(sfn
, &rt
->rt6i_src
.addr
,
1028 offsetof(struct rt6_info
, rt6i_src
),
1029 allow_create
, replace_required
, sernum
);
1032 /* If it is failed, discard just allocated
1033 root, and then (in st_failure) stale node
1041 /* Now link new subtree to main tree */
1045 sn
= fib6_add_1(fn
->subtree
, &rt
->rt6i_src
.addr
,
1047 offsetof(struct rt6_info
, rt6i_src
),
1048 allow_create
, replace_required
, sernum
);
1058 atomic_inc(&rt
->rt6i_ref
);
1064 err
= fib6_add_rt2node(fn
, rt
, info
, mxc
);
1066 fib6_start_gc(info
->nl_net
, rt
);
1067 if (!(rt
->rt6i_flags
& RTF_CACHE
))
1068 fib6_prune_clones(info
->nl_net
, pn
);
1069 rt
->dst
.flags
&= ~DST_NOCACHE
;
1074 #ifdef CONFIG_IPV6_SUBTREES
1076 * If fib6_add_1 has cleared the old leaf pointer in the
1077 * super-tree leaf node we have to find a new one for it.
1079 if (pn
!= fn
&& pn
->leaf
== rt
) {
1081 atomic_dec(&rt
->rt6i_ref
);
1083 if (pn
!= fn
&& !pn
->leaf
&& !(pn
->fn_flags
& RTN_RTINFO
)) {
1084 pn
->leaf
= fib6_find_prefix(info
->nl_net
, pn
);
1087 WARN_ON(pn
->leaf
== NULL
);
1088 pn
->leaf
= info
->nl_net
->ipv6
.ip6_null_entry
;
1091 atomic_inc(&pn
->leaf
->rt6i_ref
);
1094 if (!(rt
->dst
.flags
& DST_NOCACHE
))
1099 #ifdef CONFIG_IPV6_SUBTREES
1100 /* Subtree creation failed, probably main tree node
1101 is orphan. If it is, shoot it.
1104 if (fn
&& !(fn
->fn_flags
& (RTN_RTINFO
|RTN_ROOT
)))
1105 fib6_repair_tree(info
->nl_net
, fn
);
1106 if (!(rt
->dst
.flags
& DST_NOCACHE
))
1113 * Routing tree lookup
1117 struct lookup_args
{
1118 int offset
; /* key offset on rt6_info */
1119 const struct in6_addr
*addr
; /* search key */
1122 static struct fib6_node
*fib6_lookup_1(struct fib6_node
*root
,
1123 struct lookup_args
*args
)
1125 struct fib6_node
*fn
;
1128 if (unlikely(args
->offset
== 0))
1138 struct fib6_node
*next
;
1140 dir
= addr_bit_set(args
->addr
, fn
->fn_bit
);
1142 next
= dir
? fn
->right
: fn
->left
;
1152 if (FIB6_SUBTREE(fn
) || fn
->fn_flags
& RTN_RTINFO
) {
1155 key
= (struct rt6key
*) ((u8
*) fn
->leaf
+
1158 if (ipv6_prefix_equal(&key
->addr
, args
->addr
, key
->plen
)) {
1159 #ifdef CONFIG_IPV6_SUBTREES
1161 struct fib6_node
*sfn
;
1162 sfn
= fib6_lookup_1(fn
->subtree
,
1169 if (fn
->fn_flags
& RTN_RTINFO
)
1173 #ifdef CONFIG_IPV6_SUBTREES
1176 if (fn
->fn_flags
& RTN_ROOT
)
1185 struct fib6_node
*fib6_lookup(struct fib6_node
*root
, const struct in6_addr
*daddr
,
1186 const struct in6_addr
*saddr
)
1188 struct fib6_node
*fn
;
1189 struct lookup_args args
[] = {
1191 .offset
= offsetof(struct rt6_info
, rt6i_dst
),
1194 #ifdef CONFIG_IPV6_SUBTREES
1196 .offset
= offsetof(struct rt6_info
, rt6i_src
),
1201 .offset
= 0, /* sentinel */
1205 fn
= fib6_lookup_1(root
, daddr
? args
: args
+ 1);
1206 if (!fn
|| fn
->fn_flags
& RTN_TL_ROOT
)
1213 * Get node with specified destination prefix (and source prefix,
1214 * if subtrees are used)
1218 static struct fib6_node
*fib6_locate_1(struct fib6_node
*root
,
1219 const struct in6_addr
*addr
,
1220 int plen
, int offset
)
1222 struct fib6_node
*fn
;
1224 for (fn
= root
; fn
; ) {
1225 struct rt6key
*key
= (struct rt6key
*)((u8
*)fn
->leaf
+ offset
);
1230 if (plen
< fn
->fn_bit
||
1231 !ipv6_prefix_equal(&key
->addr
, addr
, fn
->fn_bit
))
1234 if (plen
== fn
->fn_bit
)
1238 * We have more bits to go
1240 if (addr_bit_set(addr
, fn
->fn_bit
))
1248 struct fib6_node
*fib6_locate(struct fib6_node
*root
,
1249 const struct in6_addr
*daddr
, int dst_len
,
1250 const struct in6_addr
*saddr
, int src_len
)
1252 struct fib6_node
*fn
;
1254 fn
= fib6_locate_1(root
, daddr
, dst_len
,
1255 offsetof(struct rt6_info
, rt6i_dst
));
1257 #ifdef CONFIG_IPV6_SUBTREES
1259 WARN_ON(saddr
== NULL
);
1260 if (fn
&& fn
->subtree
)
1261 fn
= fib6_locate_1(fn
->subtree
, saddr
, src_len
,
1262 offsetof(struct rt6_info
, rt6i_src
));
1266 if (fn
&& fn
->fn_flags
& RTN_RTINFO
)
1278 static struct rt6_info
*fib6_find_prefix(struct net
*net
, struct fib6_node
*fn
)
1280 if (fn
->fn_flags
& RTN_ROOT
)
1281 return net
->ipv6
.ip6_null_entry
;
1285 return fn
->left
->leaf
;
1287 return fn
->right
->leaf
;
1289 fn
= FIB6_SUBTREE(fn
);
1295 * Called to trim the tree of intermediate nodes when possible. "fn"
1296 * is the node we want to try and remove.
1299 static struct fib6_node
*fib6_repair_tree(struct net
*net
,
1300 struct fib6_node
*fn
)
1304 struct fib6_node
*child
, *pn
;
1305 struct fib6_walker
*w
;
1309 RT6_TRACE("fixing tree: plen=%d iter=%d\n", fn
->fn_bit
, iter
);
1312 WARN_ON(fn
->fn_flags
& RTN_RTINFO
);
1313 WARN_ON(fn
->fn_flags
& RTN_TL_ROOT
);
1319 child
= fn
->right
, children
|= 1;
1321 child
= fn
->left
, children
|= 2;
1323 if (children
== 3 || FIB6_SUBTREE(fn
)
1324 #ifdef CONFIG_IPV6_SUBTREES
1325 /* Subtree root (i.e. fn) may have one child */
1326 || (children
&& fn
->fn_flags
& RTN_ROOT
)
1329 fn
->leaf
= fib6_find_prefix(net
, fn
);
1333 fn
->leaf
= net
->ipv6
.ip6_null_entry
;
1336 atomic_inc(&fn
->leaf
->rt6i_ref
);
1341 #ifdef CONFIG_IPV6_SUBTREES
1342 if (FIB6_SUBTREE(pn
) == fn
) {
1343 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1344 FIB6_SUBTREE(pn
) = NULL
;
1347 WARN_ON(fn
->fn_flags
& RTN_ROOT
);
1349 if (pn
->right
== fn
)
1351 else if (pn
->left
== fn
)
1360 #ifdef CONFIG_IPV6_SUBTREES
1364 read_lock(&net
->ipv6
.fib6_walker_lock
);
1365 FOR_WALKERS(net
, w
) {
1367 if (w
->root
== fn
) {
1368 w
->root
= w
->node
= NULL
;
1369 RT6_TRACE("W %p adjusted by delroot 1\n", w
);
1370 } else if (w
->node
== fn
) {
1371 RT6_TRACE("W %p adjusted by delnode 1, s=%d/%d\n", w
, w
->state
, nstate
);
1376 if (w
->root
== fn
) {
1378 RT6_TRACE("W %p adjusted by delroot 2\n", w
);
1380 if (w
->node
== fn
) {
1383 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1384 w
->state
= w
->state
>= FWS_R
? FWS_U
: FWS_INIT
;
1386 RT6_TRACE("W %p adjusted by delnode 2, s=%d\n", w
, w
->state
);
1387 w
->state
= w
->state
>= FWS_C
? FWS_U
: FWS_INIT
;
1392 read_unlock(&net
->ipv6
.fib6_walker_lock
);
1395 if (pn
->fn_flags
& RTN_RTINFO
|| FIB6_SUBTREE(pn
))
1398 rt6_release(pn
->leaf
);
1404 static void fib6_del_route(struct fib6_node
*fn
, struct rt6_info
**rtp
,
1405 struct nl_info
*info
)
1407 struct fib6_walker
*w
;
1408 struct rt6_info
*rt
= *rtp
;
1409 struct net
*net
= info
->nl_net
;
1411 RT6_TRACE("fib6_del_route\n");
1414 *rtp
= rt
->dst
.rt6_next
;
1415 rt
->rt6i_node
= NULL
;
1416 net
->ipv6
.rt6_stats
->fib_rt_entries
--;
1417 net
->ipv6
.rt6_stats
->fib_discarded_routes
++;
1419 /* Reset round-robin state, if necessary */
1420 if (fn
->rr_ptr
== rt
)
1423 /* Remove this entry from other siblings */
1424 if (rt
->rt6i_nsiblings
) {
1425 struct rt6_info
*sibling
, *next_sibling
;
1427 list_for_each_entry_safe(sibling
, next_sibling
,
1428 &rt
->rt6i_siblings
, rt6i_siblings
)
1429 sibling
->rt6i_nsiblings
--;
1430 rt
->rt6i_nsiblings
= 0;
1431 list_del_init(&rt
->rt6i_siblings
);
1434 /* Adjust walkers */
1435 read_lock(&net
->ipv6
.fib6_walker_lock
);
1436 FOR_WALKERS(net
, w
) {
1437 if (w
->state
== FWS_C
&& w
->leaf
== rt
) {
1438 RT6_TRACE("walker %p adjusted by delroute\n", w
);
1439 w
->leaf
= rt
->dst
.rt6_next
;
1444 read_unlock(&net
->ipv6
.fib6_walker_lock
);
1446 rt
->dst
.rt6_next
= NULL
;
1448 /* If it was last route, expunge its radix tree node */
1450 fn
->fn_flags
&= ~RTN_RTINFO
;
1451 net
->ipv6
.rt6_stats
->fib_route_nodes
--;
1452 fn
= fib6_repair_tree(net
, fn
);
1455 fib6_purge_rt(rt
, fn
, net
);
1457 if (!info
->skip_notify
)
1458 inet6_rt_notify(RTM_DELROUTE
, rt
, info
, 0);
1462 int fib6_del(struct rt6_info
*rt
, struct nl_info
*info
)
1464 struct net
*net
= info
->nl_net
;
1465 struct fib6_node
*fn
= rt
->rt6i_node
;
1466 struct rt6_info
**rtp
;
1469 if (rt
->dst
.obsolete
> 0) {
1474 if (!fn
|| rt
== net
->ipv6
.ip6_null_entry
)
1477 WARN_ON(!(fn
->fn_flags
& RTN_RTINFO
));
1479 if (!(rt
->rt6i_flags
& RTF_CACHE
)) {
1480 struct fib6_node
*pn
= fn
;
1481 #ifdef CONFIG_IPV6_SUBTREES
1482 /* clones of this route might be in another subtree */
1483 if (rt
->rt6i_src
.plen
) {
1484 while (!(pn
->fn_flags
& RTN_ROOT
))
1489 fib6_prune_clones(info
->nl_net
, pn
);
1493 * Walk the leaf entries looking for ourself
1496 for (rtp
= &fn
->leaf
; *rtp
; rtp
= &(*rtp
)->dst
.rt6_next
) {
1498 fib6_del_route(fn
, rtp
, info
);
1506 * Tree traversal function.
1508 * Certainly, it is not interrupt safe.
1509 * However, it is internally reenterable wrt itself and fib6_add/fib6_del.
1510 * It means, that we can modify tree during walking
1511 * and use this function for garbage collection, clone pruning,
1512 * cleaning tree when a device goes down etc. etc.
1514 * It guarantees that every node will be traversed,
1515 * and that it will be traversed only once.
1517 * Callback function w->func may return:
1518 * 0 -> continue walking.
1519 * positive value -> walking is suspended (used by tree dumps,
1520 * and probably by gc, if it will be split to several slices)
1521 * negative value -> terminate walking.
1523 * The function itself returns:
1524 * 0 -> walk is complete.
1525 * >0 -> walk is incomplete (i.e. suspended)
1526 * <0 -> walk is terminated by an error.
1529 static int fib6_walk_continue(struct fib6_walker
*w
)
1531 struct fib6_node
*fn
, *pn
;
1538 if (w
->prune
&& fn
!= w
->root
&&
1539 fn
->fn_flags
& RTN_RTINFO
&& w
->state
< FWS_C
) {
1544 #ifdef CONFIG_IPV6_SUBTREES
1546 if (FIB6_SUBTREE(fn
)) {
1547 w
->node
= FIB6_SUBTREE(fn
);
1555 w
->state
= FWS_INIT
;
1561 w
->node
= fn
->right
;
1562 w
->state
= FWS_INIT
;
1568 if (w
->leaf
&& fn
->fn_flags
& RTN_RTINFO
) {
1590 #ifdef CONFIG_IPV6_SUBTREES
1591 if (FIB6_SUBTREE(pn
) == fn
) {
1592 WARN_ON(!(fn
->fn_flags
& RTN_ROOT
));
1597 if (pn
->left
== fn
) {
1601 if (pn
->right
== fn
) {
1603 w
->leaf
= w
->node
->leaf
;
1613 static int fib6_walk(struct net
*net
, struct fib6_walker
*w
)
1617 w
->state
= FWS_INIT
;
1620 fib6_walker_link(net
, w
);
1621 res
= fib6_walk_continue(w
);
1623 fib6_walker_unlink(net
, w
);
1627 static int fib6_clean_node(struct fib6_walker
*w
)
1630 struct rt6_info
*rt
;
1631 struct fib6_cleaner
*c
= container_of(w
, struct fib6_cleaner
, w
);
1632 struct nl_info info
= {
1636 if (c
->sernum
!= FIB6_NO_SERNUM_CHANGE
&&
1637 w
->node
->fn_sernum
!= c
->sernum
)
1638 w
->node
->fn_sernum
= c
->sernum
;
1641 WARN_ON_ONCE(c
->sernum
== FIB6_NO_SERNUM_CHANGE
);
1646 for (rt
= w
->leaf
; rt
; rt
= rt
->dst
.rt6_next
) {
1647 res
= c
->func(rt
, c
->arg
);
1650 res
= fib6_del(rt
, &info
);
1653 pr_debug("%s: del failed: rt=%p@%p err=%d\n",
1654 __func__
, rt
, rt
->rt6i_node
, res
);
1667 * Convenient frontend to tree walker.
1669 * func is called on each route.
1670 * It may return -1 -> delete this route.
1671 * 0 -> continue walking
1673 * prune==1 -> only immediate children of node (certainly,
1674 * ignoring pure split nodes) will be scanned.
1677 static void fib6_clean_tree(struct net
*net
, struct fib6_node
*root
,
1678 int (*func
)(struct rt6_info
*, void *arg
),
1679 bool prune
, int sernum
, void *arg
)
1681 struct fib6_cleaner c
;
1684 c
.w
.func
= fib6_clean_node
;
1693 fib6_walk(net
, &c
.w
);
1696 static void __fib6_clean_all(struct net
*net
,
1697 int (*func
)(struct rt6_info
*, void *),
1698 int sernum
, void *arg
)
1700 struct fib6_table
*table
;
1701 struct hlist_head
*head
;
1705 for (h
= 0; h
< FIB6_TABLE_HASHSZ
; h
++) {
1706 head
= &net
->ipv6
.fib_table_hash
[h
];
1707 hlist_for_each_entry_rcu(table
, head
, tb6_hlist
) {
1708 write_lock_bh(&table
->tb6_lock
);
1709 fib6_clean_tree(net
, &table
->tb6_root
,
1710 func
, false, sernum
, arg
);
1711 write_unlock_bh(&table
->tb6_lock
);
1717 void fib6_clean_all(struct net
*net
, int (*func
)(struct rt6_info
*, void *),
1720 __fib6_clean_all(net
, func
, FIB6_NO_SERNUM_CHANGE
, arg
);
1723 static int fib6_prune_clone(struct rt6_info
*rt
, void *arg
)
1725 if (rt
->rt6i_flags
& RTF_CACHE
) {
1726 RT6_TRACE("pruning clone %p\n", rt
);
1733 static void fib6_prune_clones(struct net
*net
, struct fib6_node
*fn
)
1735 fib6_clean_tree(net
, fn
, fib6_prune_clone
, true,
1736 FIB6_NO_SERNUM_CHANGE
, NULL
);
1739 static void fib6_flush_trees(struct net
*net
)
1741 int new_sernum
= fib6_new_sernum(net
);
1743 __fib6_clean_all(net
, NULL
, new_sernum
, NULL
);
1747 * Garbage collection
1756 static int fib6_age(struct rt6_info
*rt
, void *arg
)
1758 struct fib6_gc_args
*gc_args
= arg
;
1759 unsigned long now
= jiffies
;
1762 * check addrconf expiration here.
1763 * Routes are expired even if they are in use.
1765 * Also age clones. Note, that clones are aged out
1766 * only if they are not in use now.
1769 if (rt
->rt6i_flags
& RTF_EXPIRES
&& rt
->dst
.expires
) {
1770 if (time_after(now
, rt
->dst
.expires
)) {
1771 RT6_TRACE("expiring %p\n", rt
);
1775 } else if (rt
->rt6i_flags
& RTF_CACHE
) {
1776 if (atomic_read(&rt
->dst
.__refcnt
) == 0 &&
1777 time_after_eq(now
, rt
->dst
.lastuse
+ gc_args
->timeout
)) {
1778 RT6_TRACE("aging clone %p\n", rt
);
1780 } else if (rt
->rt6i_flags
& RTF_GATEWAY
) {
1781 struct neighbour
*neigh
;
1782 __u8 neigh_flags
= 0;
1784 neigh
= dst_neigh_lookup(&rt
->dst
, &rt
->rt6i_gateway
);
1786 neigh_flags
= neigh
->flags
;
1787 neigh_release(neigh
);
1789 if (!(neigh_flags
& NTF_ROUTER
)) {
1790 RT6_TRACE("purging route %p via non-router but gateway\n",
1801 void fib6_run_gc(unsigned long expires
, struct net
*net
, bool force
)
1803 struct fib6_gc_args gc_args
;
1807 spin_lock_bh(&net
->ipv6
.fib6_gc_lock
);
1808 } else if (!spin_trylock_bh(&net
->ipv6
.fib6_gc_lock
)) {
1809 mod_timer(&net
->ipv6
.ip6_fib_timer
, jiffies
+ HZ
);
1812 gc_args
.timeout
= expires
? (int)expires
:
1813 net
->ipv6
.sysctl
.ip6_rt_gc_interval
;
1815 gc_args
.more
= icmp6_dst_gc();
1817 fib6_clean_all(net
, fib6_age
, &gc_args
);
1819 net
->ipv6
.ip6_rt_last_gc
= now
;
1822 mod_timer(&net
->ipv6
.ip6_fib_timer
,
1824 + net
->ipv6
.sysctl
.ip6_rt_gc_interval
));
1826 del_timer(&net
->ipv6
.ip6_fib_timer
);
1827 spin_unlock_bh(&net
->ipv6
.fib6_gc_lock
);
1830 static void fib6_gc_timer_cb(unsigned long arg
)
1832 fib6_run_gc(0, (struct net
*)arg
, true);
1835 static int __net_init
fib6_net_init(struct net
*net
)
1837 size_t size
= sizeof(struct hlist_head
) * FIB6_TABLE_HASHSZ
;
1839 spin_lock_init(&net
->ipv6
.fib6_gc_lock
);
1840 rwlock_init(&net
->ipv6
.fib6_walker_lock
);
1841 INIT_LIST_HEAD(&net
->ipv6
.fib6_walkers
);
1842 setup_timer(&net
->ipv6
.ip6_fib_timer
, fib6_gc_timer_cb
, (unsigned long)net
);
1844 net
->ipv6
.rt6_stats
= kzalloc(sizeof(*net
->ipv6
.rt6_stats
), GFP_KERNEL
);
1845 if (!net
->ipv6
.rt6_stats
)
1848 /* Avoid false sharing : Use at least a full cache line */
1849 size
= max_t(size_t, size
, L1_CACHE_BYTES
);
1851 net
->ipv6
.fib_table_hash
= kzalloc(size
, GFP_KERNEL
);
1852 if (!net
->ipv6
.fib_table_hash
)
1855 net
->ipv6
.fib6_main_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_main_tbl
),
1857 if (!net
->ipv6
.fib6_main_tbl
)
1858 goto out_fib_table_hash
;
1860 net
->ipv6
.fib6_main_tbl
->tb6_id
= RT6_TABLE_MAIN
;
1861 net
->ipv6
.fib6_main_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1862 net
->ipv6
.fib6_main_tbl
->tb6_root
.fn_flags
=
1863 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1864 inet_peer_base_init(&net
->ipv6
.fib6_main_tbl
->tb6_peers
);
1866 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1867 net
->ipv6
.fib6_local_tbl
= kzalloc(sizeof(*net
->ipv6
.fib6_local_tbl
),
1869 if (!net
->ipv6
.fib6_local_tbl
)
1870 goto out_fib6_main_tbl
;
1871 net
->ipv6
.fib6_local_tbl
->tb6_id
= RT6_TABLE_LOCAL
;
1872 net
->ipv6
.fib6_local_tbl
->tb6_root
.leaf
= net
->ipv6
.ip6_null_entry
;
1873 net
->ipv6
.fib6_local_tbl
->tb6_root
.fn_flags
=
1874 RTN_ROOT
| RTN_TL_ROOT
| RTN_RTINFO
;
1875 inet_peer_base_init(&net
->ipv6
.fib6_local_tbl
->tb6_peers
);
1877 fib6_tables_init(net
);
1881 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1883 kfree(net
->ipv6
.fib6_main_tbl
);
1886 kfree(net
->ipv6
.fib_table_hash
);
1888 kfree(net
->ipv6
.rt6_stats
);
1893 static void fib6_net_exit(struct net
*net
)
1895 rt6_ifdown(net
, NULL
);
1896 del_timer_sync(&net
->ipv6
.ip6_fib_timer
);
1898 #ifdef CONFIG_IPV6_MULTIPLE_TABLES
1899 inetpeer_invalidate_tree(&net
->ipv6
.fib6_local_tbl
->tb6_peers
);
1900 kfree(net
->ipv6
.fib6_local_tbl
);
1902 inetpeer_invalidate_tree(&net
->ipv6
.fib6_main_tbl
->tb6_peers
);
1903 kfree(net
->ipv6
.fib6_main_tbl
);
1904 kfree(net
->ipv6
.fib_table_hash
);
1905 kfree(net
->ipv6
.rt6_stats
);
1908 static struct pernet_operations fib6_net_ops
= {
1909 .init
= fib6_net_init
,
1910 .exit
= fib6_net_exit
,
1913 int __init
fib6_init(void)
1917 fib6_node_kmem
= kmem_cache_create("fib6_nodes",
1918 sizeof(struct fib6_node
),
1919 0, SLAB_HWCACHE_ALIGN
,
1921 if (!fib6_node_kmem
)
1924 ret
= register_pernet_subsys(&fib6_net_ops
);
1926 goto out_kmem_cache_create
;
1928 ret
= __rtnl_register(PF_INET6
, RTM_GETROUTE
, NULL
, inet6_dump_fib
,
1931 goto out_unregister_subsys
;
1933 __fib6_flush_trees
= fib6_flush_trees
;
1937 out_unregister_subsys
:
1938 unregister_pernet_subsys(&fib6_net_ops
);
1939 out_kmem_cache_create
:
1940 kmem_cache_destroy(fib6_node_kmem
);
1944 void fib6_gc_cleanup(void)
1946 unregister_pernet_subsys(&fib6_net_ops
);
1947 kmem_cache_destroy(fib6_node_kmem
);
1950 #ifdef CONFIG_PROC_FS
1952 struct ipv6_route_iter
{
1953 struct seq_net_private p
;
1954 struct fib6_walker w
;
1956 struct fib6_table
*tbl
;
1960 static int ipv6_route_seq_show(struct seq_file
*seq
, void *v
)
1962 struct rt6_info
*rt
= v
;
1963 struct ipv6_route_iter
*iter
= seq
->private;
1965 seq_printf(seq
, "%pi6 %02x ", &rt
->rt6i_dst
.addr
, rt
->rt6i_dst
.plen
);
1967 #ifdef CONFIG_IPV6_SUBTREES
1968 seq_printf(seq
, "%pi6 %02x ", &rt
->rt6i_src
.addr
, rt
->rt6i_src
.plen
);
1970 seq_puts(seq
, "00000000000000000000000000000000 00 ");
1972 if (rt
->rt6i_flags
& RTF_GATEWAY
)
1973 seq_printf(seq
, "%pi6", &rt
->rt6i_gateway
);
1975 seq_puts(seq
, "00000000000000000000000000000000");
1977 seq_printf(seq
, " %08x %08x %08x %08x %8s\n",
1978 rt
->rt6i_metric
, atomic_read(&rt
->dst
.__refcnt
),
1979 rt
->dst
.__use
, rt
->rt6i_flags
,
1980 rt
->dst
.dev
? rt
->dst
.dev
->name
: "");
1981 iter
->w
.leaf
= NULL
;
1985 static int ipv6_route_yield(struct fib6_walker
*w
)
1987 struct ipv6_route_iter
*iter
= w
->args
;
1993 iter
->w
.leaf
= iter
->w
.leaf
->dst
.rt6_next
;
1995 if (!iter
->skip
&& iter
->w
.leaf
)
1997 } while (iter
->w
.leaf
);
2002 static void ipv6_route_seq_setup_walk(struct ipv6_route_iter
*iter
,
2005 memset(&iter
->w
, 0, sizeof(iter
->w
));
2006 iter
->w
.func
= ipv6_route_yield
;
2007 iter
->w
.root
= &iter
->tbl
->tb6_root
;
2008 iter
->w
.state
= FWS_INIT
;
2009 iter
->w
.node
= iter
->w
.root
;
2010 iter
->w
.args
= iter
;
2011 iter
->sernum
= iter
->w
.root
->fn_sernum
;
2012 INIT_LIST_HEAD(&iter
->w
.lh
);
2013 fib6_walker_link(net
, &iter
->w
);
2016 static struct fib6_table
*ipv6_route_seq_next_table(struct fib6_table
*tbl
,
2020 struct hlist_node
*node
;
2023 h
= (tbl
->tb6_id
& (FIB6_TABLE_HASHSZ
- 1)) + 1;
2024 node
= rcu_dereference_bh(hlist_next_rcu(&tbl
->tb6_hlist
));
2030 while (!node
&& h
< FIB6_TABLE_HASHSZ
) {
2031 node
= rcu_dereference_bh(
2032 hlist_first_rcu(&net
->ipv6
.fib_table_hash
[h
++]));
2034 return hlist_entry_safe(node
, struct fib6_table
, tb6_hlist
);
2037 static void ipv6_route_check_sernum(struct ipv6_route_iter
*iter
)
2039 if (iter
->sernum
!= iter
->w
.root
->fn_sernum
) {
2040 iter
->sernum
= iter
->w
.root
->fn_sernum
;
2041 iter
->w
.state
= FWS_INIT
;
2042 iter
->w
.node
= iter
->w
.root
;
2043 WARN_ON(iter
->w
.skip
);
2044 iter
->w
.skip
= iter
->w
.count
;
2048 static void *ipv6_route_seq_next(struct seq_file
*seq
, void *v
, loff_t
*pos
)
2052 struct net
*net
= seq_file_net(seq
);
2053 struct ipv6_route_iter
*iter
= seq
->private;
2058 n
= ((struct rt6_info
*)v
)->dst
.rt6_next
;
2065 ipv6_route_check_sernum(iter
);
2066 read_lock(&iter
->tbl
->tb6_lock
);
2067 r
= fib6_walk_continue(&iter
->w
);
2068 read_unlock(&iter
->tbl
->tb6_lock
);
2072 return iter
->w
.leaf
;
2074 fib6_walker_unlink(net
, &iter
->w
);
2077 fib6_walker_unlink(net
, &iter
->w
);
2079 iter
->tbl
= ipv6_route_seq_next_table(iter
->tbl
, net
);
2083 ipv6_route_seq_setup_walk(iter
, net
);
2087 static void *ipv6_route_seq_start(struct seq_file
*seq
, loff_t
*pos
)
2090 struct net
*net
= seq_file_net(seq
);
2091 struct ipv6_route_iter
*iter
= seq
->private;
2094 iter
->tbl
= ipv6_route_seq_next_table(NULL
, net
);
2098 ipv6_route_seq_setup_walk(iter
, net
);
2099 return ipv6_route_seq_next(seq
, NULL
, pos
);
2105 static bool ipv6_route_iter_active(struct ipv6_route_iter
*iter
)
2107 struct fib6_walker
*w
= &iter
->w
;
2108 return w
->node
&& !(w
->state
== FWS_U
&& w
->node
== w
->root
);
2111 static void ipv6_route_seq_stop(struct seq_file
*seq
, void *v
)
2114 struct net
*net
= seq_file_net(seq
);
2115 struct ipv6_route_iter
*iter
= seq
->private;
2117 if (ipv6_route_iter_active(iter
))
2118 fib6_walker_unlink(net
, &iter
->w
);
2120 rcu_read_unlock_bh();
2123 static const struct seq_operations ipv6_route_seq_ops
= {
2124 .start
= ipv6_route_seq_start
,
2125 .next
= ipv6_route_seq_next
,
2126 .stop
= ipv6_route_seq_stop
,
2127 .show
= ipv6_route_seq_show
2130 int ipv6_route_open(struct inode
*inode
, struct file
*file
)
2132 return seq_open_net(inode
, file
, &ipv6_route_seq_ops
,
2133 sizeof(struct ipv6_route_iter
));
2136 #endif /* CONFIG_PROC_FS */