Re: [PATCH] mtd: spi-nor: stm32-quadspi: avoid unintialized return code

[Geert Uytterhoeven <geert@linux-m68k.org>]

Hi Ludovic,

On Thu, Sep 14, 2017 at 5:13 PM, Ludovic BARRE <ludovic.barre@st.com> wrote:
> On 09/14/2017 03:38 PM, Geert Uytterhoeven wrote:
>> On Thu, Sep 14, 2017 at 1:06 PM, Arnd Bergmann <arnd@arndb.de> wrote:
>>> If we send zero-length data to stm32_qspi_tx_poll() on older
>>> compiler versions such as gcc-4.6, we get warned that the
>>> return code is uninitialized:
>>>
>>> drivers/mtd/spi-nor/stm32-quadspi.c:248:2: error: ‘ret’ may be used
>>> uninitialized in this function [-Werror=uninitialized]
>>>
>>> On newer compiler versions, the return code is always zero
>>> in this case, as the local variable gets optimized away and
>>> is assumed to be zero after the loop completes without error.
>>>
>>> This changes the function to instead return -EINVAL if it
>>> ever gets called with a zero length buffer.
>>>
>>> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82203
>>> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
>>> ---
>>>   drivers/mtd/spi-nor/stm32-quadspi.c | 2 +-
>>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/drivers/mtd/spi-nor/stm32-quadspi.c
>>> b/drivers/mtd/spi-nor/stm32-quadspi.c
>>> index 86c0931543c5..711cfe7aa4bf 100644
>>> --- a/drivers/mtd/spi-nor/stm32-quadspi.c
>>> +++ b/drivers/mtd/spi-nor/stm32-quadspi.c
>>> @@ -227,7 +227,7 @@ static int stm32_qspi_tx_poll(struct stm32_qspi
>>> *qspi,
>>>          void (*tx_fifo)(u8 *, void __iomem *);
>>>          u32 len = cmd->len, sr;
>>>          u8 *buf = cmd->buf;
>>> -       int ret;
>>> +       int ret = -EINVAL;
>>>
>>>          if (cmd->qspimode == CCR_FMODE_INDW)
>>>                  tx_fifo = stm32_qspi_write_fifo;
>>
>>
>> See also "[PATCH] mtd: spi-nor: stm32-quadspi: Fix uninitialized error
>> return code"
>> (https://patchwork.kernel.org/patch/9842173/)
>
> hi Arnd, Geert
>
> sorry, I was forgot this thread while my holidays
>
> Geert: what do you mean like "similar bugs in the future" in "If you
> initialized ret at the beginning, you lose the ability to catch newly
> introduced similar bugs in the future."

If you pre-initialize ret at the top, you loose the ability of the compiler
to detect at compile-time if ret is never written to later. It will just return
-EINVAL at runtime.

With my version, if the code is modified later and another "return ret" is
added, the compiler will detect if there's a code path that forgets
to assign a value to ret.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds