On Mon, Jul 03, 2023 at 08:23:30PM +0200, Krzysztof Kozlowski wrote:
> On 30/06/2023 00:38, Andi Shyti wrote:
> > Hi Michal,
> >> -static struct at91_twi_pdata *at91_twi_get_driver_data(
> >> +static const struct at91_twi_pdata *at91_twi_get_driver_data(
> >> struct platform_device *pdev)
> >> {
> >> if (pdev->dev.of_node) {
> >> @@ -189,9 +189,9 @@ static struct at91_twi_pdata *at91_twi_get_driver_data(
> >> match = of_match_node(atmel_twi_dt_ids, pdev->dev.of_node);
> >> if (!match)
> >> return NULL;
> >> - return (struct at91_twi_pdata *)match->data;
> >> + return match->data;
> >> }
> >> - return (struct at91_twi_pdata *) platform_get_device_id(pdev)->driver_data;
> >> + return (const void *) platform_get_device_id(pdev)->driver_data;
> >
> > the const's always confuse me... do you get an error here? Is
> > this cast really needed?
>
> I think this change is not necessary and actually should not matter. See
> for example drivers/tty/serial/samsung_tty.c after my refactorings in
> commit 3aec40096550 ("tty: serial: samsung: reduce number of casts").
I added this two fragments to avoid having suggesting that the pdata might
not be in a read-only segment. For the compiler it doesn't matter -- the
pointer is cast to const on return.
Best Regards
Michał Mirosław
