On 2/7/22 7:33 AM, Alviro Iskandar Setiawan wrote:
> On Mon, Feb 7, 2022 at 9:21 PM Dan Carpenter <dan.carpenter@xxxxxxxxxx> wrote:
>> On Mon, Feb 07, 2022 at 06:45:57AM -0700, Jens Axboe wrote:
>>> On 2/7/22 4:43 AM, Ammar Faizi wrote:
>>>> From: Alviro Iskandar Setiawan <alviro.iskandar@xxxxxxxxx>
>>>>
>>>> In io_recv() if import_single_range() fails, the @flags variable is
>>>> uninitialized, then it will goto out_free.
>>>>
>>>> After the goto, the compiler doesn't know that (ret < min_ret) is
>>>> always true, so it thinks the "if ((flags & MSG_WAITALL) ..." path
>>>> could be taken.
>>>>
>>>> The complaint comes from gcc-9 (Debian 9.3.0-22) 9.3.0:
>>>> ```
>>>> fs/io_uring.c:5238 io_recvfrom() error: uninitialized symbol 'flags'
>>>> ```
>>>> Fix this by bypassing the @ret and @flags check when
>>>> import_single_range() fails.
>>>
>>> The compiler should be able to deduce this, and I guess newer compilers
>>> do which is why we haven't seen this warning before.
>
> The compiler can't deduce this because the import_single_range() is
> located in a different translation unit (different C file), so it
> can't prove that (ret < min_ret) is always true as it can't see the
> function definition (in reality, it is always true because it only
> returns either 0 or -EFAULT).
Yes you are right, I forgot this is the generic helper, and not our
internal one.
>> No, we disabled GCC's uninitialized variable checking a couple years
>> back. Linus got sick of the false positives. You can still see it if
>> you enable W=2
>>
>> fs/io_uring.c: In function ‘io_recv’:
>> fs/io_uring.c:5252:20: warning: ‘flags’ may be used uninitialized in this function [-Wmaybe-uninitialized]
>> } else if ((flags & MSG_WAITALL) && (msg.msg_flags & (MSG_TRUNC | MSG_CTRUNC))) {
>> ~~~~~~~^~~~~~~~~~~~~~
>>
>> If you introduce an uninitialized variable bug then likelyhood is the
>> kbuild-bot will send you a Clang warning or a Smatch warning or both.
>> I don't think anyone looks at GCC W=2 warnings.
>>
>
> This warning is valid, and the compiler should really warn that. But
> again, in reality, this is still a false-positive warning, because
> that "else if" will never be taken from the "goto out_free" path.
Right, as mentioned in my email, there is no bug there. But I do like
the patch as it cleans it up too, the error-out path should not include
non-cleanup items.
--
Jens Axboe
