Re: [PATCH] Staging: octeon: Avoid several usecases of strcpy

[Rasmus Villemoes <linux@rasmusvillemoes.dk>]

On 11/09/2019 11.16, Dan Carpenter wrote:
> On Wed, Sep 11, 2019 at 11:04:38AM +0200, Sandro Volery wrote:
>>
>>
>>> On 11 Sep 2019, at 10:52, Dan Carpenter <dan.carpenter@oracle.com> wrote:
>>>
>>> On Wed, Sep 11, 2019 at 08:23:59AM +0200, Sandro Volery wrote:
>>>> strcpy was used multiple times in strcpy to write into dev->name.
>>>> I replaced them with strscpy.

Yes, that's obviously what the patch does. The commit log is supposed to
explain _why_.

>>>> Signed-off-by: Sandro Volery <sandro@volery.com>
>>>> ---
>>>> drivers/staging/octeon/ethernet.c | 16 ++++++++--------
>>>> 1 file changed, 8 insertions(+), 8 deletions(-)
>>>>
>>>> diff --git a/drivers/staging/octeon/ethernet.c b/drivers/staging/octeon/ethernet.c
>>>> index 8889494adf1f..cf8e9a23ebf9 100644
>>>> --- a/drivers/staging/octeon/ethernet.c
>>>> +++ b/drivers/staging/octeon/ethernet.c
>>>> @@ -784,7 +784,7 @@ static int cvm_oct_probe(struct platform_device *pdev)
>>>>            priv->imode = CVMX_HELPER_INTERFACE_MODE_DISABLED;
>>>>            priv->port = CVMX_PIP_NUM_INPUT_PORTS;
>>>>            priv->queue = -1;
>>>> -            strcpy(dev->name, "pow%d");
>>>> +            strscpy(dev->name, "pow%d", sizeof(dev->name));
>>>
>>> Is there a program which is generating a warning for this code?  We know
>>> that "pow%d" is 6 characters and static analysis tools can understand
>>> this code fine so we know it's safe.
>>
>> Well I was confused too but checkpatch complained about 
>> it so I figured I'd clean it up quick
> 
> Ah.  It's a new checkpatch warning.  I don't care in that case.  I'm
> fine with replacing all of these in that case.

But why? It actually gives _less_ compile-time checking (gcc and all
static tools know perfectly well what strcpy is and does, but knows
nothing of strscpy). And using sizeof() instead of ARRAY_SIZE() means a
future reader is not even sure dev->name is not just a pointer.

Moreover, it's very likely also a runtime and .text pessimization, again
because gcc knows what strcpy does, so it can just do a few immediate
stores (e.g. 0x25776f70 for the "pow%" part) instead of emitting an
actual function call.

Rasmus
_______________________________________________
devel mailing list
devel@linuxdriverproject.org
http://driverdev.linuxdriverproject.org/mailman/listinfo/driverdev-devel