[PATCH] gpio-uclass: fix gpio flags save condition

[Simon Glass <sjg@chromium.org>]

Hi K?ry,

On Mon, 25 Jan 2021 at 07:27, K?ry Maincent <kory.maincent@bootlin.com> wrote:
>
> Hello Simon,
>
> On Sun, 24 Jan 2021 13:24:14 -0700
> Simon Glass <sjg@chromium.org> wrote:
>
> > Hi Kory,
> >
> > On Fri, 22 Jan 2021 at 08:23, Kory Maincent <kory.maincent@bootlin.com> wrote:
> > >
> > > The commit cd2faeba1a moves the location where we save the flags to the gpio
> > > descriptor but it adds a "!" character.
> > > This breaks the condition we expect to save the flags.
> > >
> > > Signed-off-by: Kory Maincent <kory.maincent@bootlin.com>
> > > ---
> > >  drivers/gpio/gpio-uclass.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> >
> > This does indeed seem to be a bug, and the tests do not cover it. But
> > this happens to be fixed by my GPIO series. I'm about to send out v3
> > so will copy you on it.
>
> Thanks.
>
> I have tested your series of patch and it seems to have also an issue with my
> use case.
> I am working with w1-gpio.
> I got this issue:
> => w1 bus
> check_dir_flags: flags 0x16 has GPIOD_IS_OUT and GPIOD_IS_IN
> gpio_sunxi PD: PD error: set_dir_flags for gpio PD2 has invalid dir flags 0x16
> Bus 0:  onewire  (active)
>         w1_eeprom at 0 (0) uclass w1_eeprom : family 0x0

Thank you for that. I think that is because dm_gpio_set_dir_flags()
does not clear the flags first. But we now do actually update
desc->flags, where as before it was just supposed to happen. So I
think this function needs to update to clear the direction flags, then
add in the ones passed.

I have pushed this to u-boot-dm/gpio-working if you can test it again.

Regards,
Simon