On Sat, Mar 09, 2019 at 11:18:53PM +0800, Axel Lin wrote:
> Current code test wrong value so it does not verify if the written
> data is correctly read back. Fix it.
>
> Signed-off-by: Axel Lin <axel.lin@xxxxxxxxxx>
> ---
> BTW, I'm wondering why it needs to read back the written vlaue for double
> checking if write is success or not.
> Without any comment, it looks like a leftover of testing code.
> Does it make sense to remove the read after write?
The purpose of this was to support read-only and write-only pins in the
controller. The controller would not allow a bit to be set (or cleared)
if the pin was read- or write-only. And yes, that would've probably
deserved a comment.
> drivers/gpio/gpio-adnp.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpio/gpio-adnp.c b/drivers/gpio/gpio-adnp.c
> index 91b90c0cea73..d334dc302dac 100644
> --- a/drivers/gpio/gpio-adnp.c
> +++ b/drivers/gpio/gpio-adnp.c
> @@ -132,8 +132,10 @@ static int adnp_gpio_direction_input(struct gpio_chip *chip, unsigned offset)
> if (err < 0)
> goto out;
>
> - if (err & BIT(pos))
> + if (value & BIT(pos)) {
> err = -EACCES;
> + goto out;
> + }
Perhaps while at it also make this return -EPERM just like it done for
adnp_gpio_direction_output()?
With that:
Reviewed-by: Thierry Reding <thierry.reding@xxxxxxxxx>
Attachment:
signature.asc
Description: PGP signature
