[PATCH 1/1] cred: fix memory leak

[PATCH 1/1] cred: fix memory leak

[Jiri Slaby]

Stanse found a memory leak in prepare_exec_creds. tgcred is not
freed/assigned on all paths. Fix that.

I.e. unifdef tgcred and add kfree(tgcred); as it is initialized to
NULL already.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: David Howells <dhowells@redhat.com>
Cc: James Morris <jmorris@namei.org>
Cc: Serge Hallyn <serue@us.ibm.com>
---
 kernel/cred.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/kernel/cred.c b/kernel/cred.c
index dd76cfe..0e10f73 100644
--- a/kernel/cred.c
+++ b/kernel/cred.c
@@ -351,9 +351,7 @@ struct cred *prepare_exec_creds(void)
  */
 struct cred *prepare_usermodehelper_creds(void)
 {
-#ifdef CONFIG_KEYS
 	struct thread_group_cred *tgcred = NULL;
-#endif
 	struct cred *new;
 
 #ifdef CONFIG_KEYS
@@ -363,8 +361,10 @@ struct cred *prepare_usermodehelper_creds(void)
 #endif
 
 	new = kmem_cache_alloc(cred_jar, GFP_ATOMIC);
-	if (!new)
+	if (!new) {
+		kfree(tgcred);
 		return NULL;
+	}
 
 	kdebug("prepare_usermodehelper_creds() alloc %p", new);
 
-- 
1.6.5.7

Re: [PATCH 1/1] cred: fix memory leak

[David Howells]

Jiri Slaby <jslaby@suse.cz> wrote:

> Stanse found a memory leak in prepare_exec_creds. tgcred is not
> freed/assigned on all paths. Fix that.
> 
> I.e. unifdef tgcred and add kfree(tgcred); as it is initialized to
> NULL already.
> 
> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
> Cc: David Howells <dhowells@redhat.com>
> Cc: James Morris <jmorris@namei.org>
> Cc: Serge Hallyn <serue@us.ibm.com>

Acked-by: David Howells <dhowells@redhat.com>

Re: [PATCH 1/1] cred: fix memory leak

[Serge E. Hallyn]

Quoting Jiri Slaby (jslaby@suse.cz):
> Stanse found a memory leak in prepare_exec_creds. tgcred is not
> freed/assigned on all paths. Fix that.
> 
> I.e. unifdef tgcred and add kfree(tgcred); as it is initialized to
> NULL already.

Does this compile with CONFIG_KEYS=n, , though?  I don't see a dummy
define for struct thread_group_cred in cred.h.  Should this patch add
one?

> Signed-off-by: Jiri Slaby <jslaby@suse.cz>
> Cc: David Howells <dhowells@redhat.com>
> Cc: James Morris <jmorris@namei.org>
> Cc: Serge Hallyn <serue@us.ibm.com>
> ---
>  kernel/cred.c |    6 +++---
>  1 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/kernel/cred.c b/kernel/cred.c
> index dd76cfe..0e10f73 100644
> --- a/kernel/cred.c
> +++ b/kernel/cred.c
> @@ -351,9 +351,7 @@ struct cred *prepare_exec_creds(void)
>   */
>  struct cred *prepare_usermodehelper_creds(void)
>  {
> -#ifdef CONFIG_KEYS
>  	struct thread_group_cred *tgcred = NULL;
> -#endif
>  	struct cred *new;
> 
>  #ifdef CONFIG_KEYS
> @@ -363,8 +361,10 @@ struct cred *prepare_usermodehelper_creds(void)
>  #endif
> 
>  	new = kmem_cache_alloc(cred_jar, GFP_ATOMIC);
> -	if (!new)
> +	if (!new) {
> +		kfree(tgcred);
>  		return NULL;
> +	}
> 
>  	kdebug("prepare_usermodehelper_creds() alloc %p", new);
> 
> -- 
> 1.6.5.7

Re: [PATCH 1/1] cred: fix memory leak

[Jiri Slaby]

On 01/06/2010 06:25 PM, Serge E. Hallyn wrote:
> Quoting Jiri Slaby (jslaby@suse.cz):
>> Stanse found a memory leak in prepare_exec_creds. tgcred is not
>> freed/assigned on all paths. Fix that.
>>
>> I.e. unifdef tgcred and add kfree(tgcred); as it is initialized to
>> NULL already.
> 
> Does this compile with CONFIG_KEYS=n, , though? 

Yes and I guess it's due to no dereference of the pointer.

> Should this patch add one?

Hmm, I don't think so. The patch is ugly in the light of not having the
struct defined. I should come up with something where the CONFIG_KEYS is
left there.

thanks,
-- 
js