[PATCH bpf v2] bpf: fix memory leak in lpm_trie map_free callback function

[PATCH bpf v2] bpf: fix memory leak in lpm_trie map_free callback function

[Yonghong Song]

There is a memory leak happening in lpm_trie map_free callback
function trie_free. The trie structure itself does not get freed.

Also, trie_free function did not do synchronize_rcu before freeing
various data structures. This is incorrect as some rcu_read_lock
region(s) for lookup, update, delete or get_next_key may not complete yet.
The fix is to add synchronize_rcu in the beginning of trie_free.
The useless spin_lock is removed from this function as well.

Fixes: b95a5c4db09b ("bpf: add a longest prefix match trie map implementation")
Reported-by: Mathieu Malaterre <malat@debian.org>
Reported-by: Alexei Starovoitov <ast@kernel.org>
Tested-by: Mathieu Malaterre <malat@debian.org>
Signed-off-by: Yonghong Song <yhs@fb.com>
---
 kernel/bpf/lpm_trie.c | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

v1->v2:
  Make comments more precise and make label name more appropriate,
  as suggested by Daniel

diff --git a/kernel/bpf/lpm_trie.c b/kernel/bpf/lpm_trie.c
index 7b469d1..a75e02c 100644
--- a/kernel/bpf/lpm_trie.c
+++ b/kernel/bpf/lpm_trie.c
@@ -555,7 +555,10 @@ static void trie_free(struct bpf_map *map)
 	struct lpm_trie_node __rcu **slot;
 	struct lpm_trie_node *node;
 
-	raw_spin_lock(&trie->lock);
+	/* Wait for outstanding programs to complete
+	 * update/lookup/delete/get_next_key and free the trie.
+	 */
+	synchronize_rcu();
 
 	/* Always start at the root and walk down to a node that has no
 	 * children. Then free that node, nullify its reference in the parent
@@ -569,7 +572,7 @@ static void trie_free(struct bpf_map *map)
 			node = rcu_dereference_protected(*slot,
 					lockdep_is_held(&trie->lock));
 			if (!node)
-				goto unlock;
+				goto out;
 
 			if (rcu_access_pointer(node->child[0])) {
 				slot = &node->child[0];
@@ -587,8 +590,8 @@ static void trie_free(struct bpf_map *map)
 		}
 	}
 
-unlock:
-	raw_spin_unlock(&trie->lock);
+out:
+	kfree(trie);
 }
 
 static int trie_get_next_key(struct bpf_map *map, void *_key, void *_next_key)
-- 
2.9.5

Re: [PATCH bpf v2] bpf: fix memory leak in lpm_trie map_free callback function

[Alexei Starovoitov]

On Tue, Feb 13, 2018 at 07:00:21PM -0800, Yonghong Song wrote:
> There is a memory leak happening in lpm_trie map_free callback
> function trie_free. The trie structure itself does not get freed.
> 
> Also, trie_free function did not do synchronize_rcu before freeing
> various data structures. This is incorrect as some rcu_read_lock
> region(s) for lookup, update, delete or get_next_key may not complete yet.
> The fix is to add synchronize_rcu in the beginning of trie_free.
> The useless spin_lock is removed from this function as well.
> 
> Fixes: b95a5c4db09b ("bpf: add a longest prefix match trie map implementation")
> Reported-by: Mathieu Malaterre <malat@debian.org>
> Reported-by: Alexei Starovoitov <ast@kernel.org>
> Tested-by: Mathieu Malaterre <malat@debian.org>
> Signed-off-by: Yonghong Song <yhs@fb.com>
> ---
>  kernel/bpf/lpm_trie.c | 11 +++++++----
>  1 file changed, 7 insertions(+), 4 deletions(-)
> 
> v1->v2:
>   Make comments more precise and make label name more appropriate,
>   as suggested by Daniel

Applied to bpf tree, Thanks Yonghong.

Re: [PATCH bpf v2] bpf: fix memory leak in lpm_trie map_free callback function

[Eric Dumazet]

On Tue, 2018-02-13 at 19:17 -0800, Alexei Starovoitov wrote:
> On Tue, Feb 13, 2018 at 07:00:21PM -0800, Yonghong Song wrote:
> > There is a memory leak happening in lpm_trie map_free callback
> > function trie_free. The trie structure itself does not get freed.
> > 
> > Also, trie_free function did not do synchronize_rcu before freeing
> > various data structures. This is incorrect as some rcu_read_lock
> > region(s) for lookup, update, delete or get_next_key may not complete yet.
> > The fix is to add synchronize_rcu in the beginning of trie_free.
> > The useless spin_lock is removed from this function as well.
> > 
> > Fixes: b95a5c4db09b ("bpf: add a longest prefix match trie map implementation")
> > Reported-by: Mathieu Malaterre <malat@debian.org>
> > Reported-by: Alexei Starovoitov <ast@kernel.org>
> > Tested-by: Mathieu Malaterre <malat@debian.org>
> > Signed-off-by: Yonghong Song <yhs@fb.com>
> > ---
> >  kernel/bpf/lpm_trie.c | 11 +++++++----
> >  1 file changed, 7 insertions(+), 4 deletions(-)
> > 
> > v1->v2:
> >   Make comments more precise and make label name more appropriate,
> >   as suggested by Daniel
> 
> Applied to bpf tree, Thanks Yonghong.


This does not look good.

LOCKDEP surely should complain to

node = rcu_dereference_protected(*slot, lockdep_is_held(&trie->lock));

Since we no longer hold trie->lock

Re: [PATCH bpf v2] bpf: fix memory leak in lpm_trie map_free callback function

[Yonghong Song]

On 2/21/18 7:40 PM, Eric Dumazet wrote:
> On Tue, 2018-02-13 at 19:17 -0800, Alexei Starovoitov wrote:
>> On Tue, Feb 13, 2018 at 07:00:21PM -0800, Yonghong Song wrote:
>>> There is a memory leak happening in lpm_trie map_free callback
>>> function trie_free. The trie structure itself does not get freed.
>>>
>>> Also, trie_free function did not do synchronize_rcu before freeing
>>> various data structures. This is incorrect as some rcu_read_lock
>>> region(s) for lookup, update, delete or get_next_key may not complete yet.
>>> The fix is to add synchronize_rcu in the beginning of trie_free.
>>> The useless spin_lock is removed from this function as well.
>>>
>>> Fixes: b95a5c4db09b ("bpf: add a longest prefix match trie map implementation")
>>> Reported-by: Mathieu Malaterre <malat@debian.org>
>>> Reported-by: Alexei Starovoitov <ast@kernel.org>
>>> Tested-by: Mathieu Malaterre <malat@debian.org>
>>> Signed-off-by: Yonghong Song <yhs@fb.com>
>>> ---
>>>   kernel/bpf/lpm_trie.c | 11 +++++++----
>>>   1 file changed, 7 insertions(+), 4 deletions(-)
>>>
>>> v1->v2:
>>>    Make comments more precise and make label name more appropriate,
>>>    as suggested by Daniel
>>
>> Applied to bpf tree, Thanks Yonghong.
> 
> 
> This does not look good.
> 
> LOCKDEP surely should complain to
> 
> node = rcu_dereference_protected(*slot, lockdep_is_held(&trie->lock));
> 
> Since we no longer hold trie->lock

Eric,

Thanks for spotting this issue. Will fix this issue soon.

Yonghong