Re: [PATCH] capabilities: fix sys_prctl() returned uninitialized value

[Shi Weihua <shiwh@cn.fujitsu.com>]

Serge E. Hallyn wrote:
> Quoting Shi Weihua (shiwh@cn.fujitsu.com):
>> When we test kernel by the latest LTP(20080430) on ia64,
>> the following failure occured:
>> -------------------------------------
>> prctl01     1  PASS  :  Test Passed
>> prctl01     0  WARN  :  prctl() returned 2048 errno = 0 : Success
>> prctl01     1  PASS  :  Test Passed
>> prctl01     2  FAIL  :  Test Failed
>> -------------------------------------
>>
>> We found commit 3898b1b4ebff8dcfbcf1807e0661585e06c9a91c
>> causes this failure by git-bisect.
>> And, we found *rc_p has not been initialized if switch-default 
>> of the function cap_task_prctl()(security/commoncap.c). When *rc_p
>> uninitialized, sys_prctl() will return a wrong value.
>>
>> Signed-off-by: Shi Weihua <shiwh@cn.fujitsu.com> 
>> ---
>> diff --git a/security/commoncap.c b/security/commoncap.c
>> index 5edabc7..a4b28c8 100644
>> --- a/security/commoncap.c
>> +++ b/security/commoncap.c
>> @@ -649,6 +649,7 @@ int cap_task_prctl(int option, unsigned long arg2, unsigned long arg3,
>>
>>  	default:
>>  		/* No functionality available - continue with default */
>> +		*rc_p = 0;
>>  		return 0;
>>  	}
> 
> No, this case here means that the capability module is not taking
> responsibility for this call.  So it should not be setting rc_p.

Ok, we noticed the comment as following in include/linux/security.h.
+ *      @rc_p contains a pointer to communicate back the forced return code
+ *     Return 0 if permission is granted, and non-zero if the security module
+ *      has taken responsibility (setting *rc_p) for the prctl call.

> 
> So you'll want to find another path in kernel/sys.c:sys_prctl()
> where error doesn't get set.  Do you know what 'i' was in prctl01
> at the time of failure?

'i' was 1 (PR_SET_PDEATHSIG).

I will create a new patch ASAP.

Thanks.

> 
> For instance, I notice that PR_SET_DUMPABLE doesn't set the value
> of error if arg2 is valid.  Also PR_SET_NAME and PR_GET_NAME
> don't set error.
> 
> -serge
> 
> 
>