Re: Linux 2.6.37-rc1 (pcrypt fault)

[Randy Dunlap <randy.dunlap@oracle.com>]

On 11/10/10 03:21, Steffen Klassert wrote:
> On Wed, Nov 03, 2010 at 02:15:19PM -0700, Randy Dunlap wrote:
>>
>> modprobe pcrypt; rmmod pcrypt ==>
>>
>>
>> [   76.081639] calling  pcrypt_init+0x0/0x107 [pcrypt] @ 3016
>> Nov  3 13:02:15 control kernel: [   76.089883] initcall pcrypt_init+0x0/0x107 [pcrypt] returned 0 after 2476 usecs
>> [   76.081639] calling  pcrypt_i
>>
>>
>> [   79.940445] general protection fault: 0000 [#1] SMP DEBUG_PAGEALLOC
> 
> Looks like a use after free of the padata instance.
> Does the patch below fix it?

Yes, it does.  Thanks.

Tested-by: Randy Dunlap <randy.dunlap@oracle.com>


> Thanks for reporting,
> 
> Steffen
> 
> 
> Subject: [PATCH] crypto: pcrypt - Fix use after free on padata_free
> 
> kobject_put is called from padata_free for the padata kobject.
> The kobject's release function frees the padata instance,
> so don't call kobject_put for the padata kobject from pcrypt.
> 
> Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
> ---
>  crypto/pcrypt.c |    1 -
>  1 files changed, 0 insertions(+), 1 deletions(-)
> 
> diff --git a/crypto/pcrypt.c b/crypto/pcrypt.c
> index de30782..75586f1 100644
> --- a/crypto/pcrypt.c
> +++ b/crypto/pcrypt.c
> @@ -504,7 +504,6 @@ err:
>  
>  static void pcrypt_fini_padata(struct padata_pcrypt *pcrypt)
>  {
> -	kobject_put(&pcrypt->pinst->kobj);
>  	free_cpumask_var(pcrypt->cb_cpumask->mask);
>  	kfree(pcrypt->cb_cpumask);
>  


-- 
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***