Re: [PATCH 6/6] mqueue: fix mq_open() file descriptor leak on user-space processes

[Xiaotian Feng]

On Thu, Feb 25, 2010 at 2:59 PM, AmÃrico Wang <xiyou.wangcong@xxxxxxxxx> wrote:
> On Thu, Feb 25, 2010 at 12:25 PM, AmÃrico Wang <xiyou.wangcong@xxxxxxxxx> wrote:
>> On Thu, Feb 25, 2010 at 12:00 PM, Xiaotian Feng <xtfeng@xxxxxxxxx> wrote:
>>> 2010/2/25 AmÃrico Wang <xiyou.wangcong@xxxxxxxxx>:
>>>> On Tue, Feb 23, 2010 at 3:04 PM, Andrà Goddard Rosa
>>>> <andre.goddard@xxxxxxxxx> wrote:
>>>>> It can be triggered by the following test program:
>>>>>
>>>>
>>>> <snip>
>>>>
>>>>>
>>>>> When not running valgrind, user-space program segfaults trying to execute
>>>>> strerror(errno). With valgrind, it executes successfully and prints the
>>>>> 5 open files: stdin, stdout, stderr, pipe[0] and pipe[1].
>>>>>
>>>>> Signed-off-by: Andrà Goddard Rosa <andre.goddard@xxxxxxxxx>
>>>>> ---
>>>>
>>>> The code has more than just this problem, could you please try
>>>> my patch below?
>>>>
>>>> Thanks.
>>>>
>>>> ---------------------------->
>>>>
>>>> Clean up the failure path of sys_mq_open().
>>>>
>>>> Reorder the goto labels;
>>>> Rename 'upsem' to 'upunlock';
>>>> Remove some unused labels;
>>>> Fix some wrong goto path.
>>>>
>>>
>>> I think it's wrong to move dput after mntput
>>
>> Oh, this is to say mntget() should be called before lookup_one_len(),
>> the original code seems wrong again...
>>
>
> How about the one below?

This is definitely wrong,

 	if (IS_ERR(filp)) {
 		error = PTR_ERR(filp);
-		goto out_putfd;
+		goto out;
 	}

filp is assigned by do_open or do_create in mqueue.c, take a look at
the code, if do_open/do_create is failed, kernel is already dput &
mntput...

So I think original patch from Andrà is enough....

> ---------
>
> Signed-off-by: WANG Cong <xiyou.wangcong@xxxxxxxxx>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/