[18/48] ARM: pxa: hx4700: use gpio descriptors for audio

From: Arnd Bergmann <arnd@arndb.de>

The audio driver should not use a hardwired gpio number
from the header. Change it to use a lookup table.

Cc: Philipp Zabel <philipp.zabel@gmail.com>
Cc: Paul Parsons <lost.distance@yahoo.com>
Acked-by: Mark Brown <broonie@kernel.org>
Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
Cc: alsa-devel@alsa-project.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 arch/arm/mach-pxa/hx4700-pcmcia.c             |  2 +-
 arch/arm/mach-pxa/hx4700.c                    | 16 ++++++++-
 arch/arm/mach-pxa/{include/mach => }/hx4700.h |  2 +-
 sound/soc/pxa/hx4700.c                        | 34 ++++++++-----------
 4 files changed, 31 insertions(+), 23 deletions(-)
 rename arch/arm/mach-pxa/{include/mach => }/hx4700.h (99%)

On Tue, Apr 19, 2022 at 6:41 PM Arnd Bergmann <arnd@kernel.org> wrote:

> From: Arnd Bergmann <arnd@arndb.de>
>
> The audio driver should not use a hardwired gpio number
> from the header. Change it to use a lookup table.
>
> Cc: Philipp Zabel <philipp.zabel@gmail.com>
> Cc: Paul Parsons <lost.distance@yahoo.com>
> Acked-by: Mark Brown <broonie@kernel.org>
> Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
> Cc: alsa-devel@alsa-project.org
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

(...)
> +static struct gpiod_lookup_table hx4700_audio_gpio_table = {
> +       .dev_id = "hx4700-audio",
> +       .table = {
> +               GPIO_LOOKUP("gpio-pxa", GPIO75_HX4700_EARPHONE_nDET,
> +                           "earphone-ndet", GPIO_ACTIVE_HIGH),

This looks wrong. The n in nDET in the end of the name of the GPIO line
means active low does it not?

What I usually do when I see this is to properly set it to
GPIO_ACTIVE_LOW in the descriptor table, then invert the logic
where it's getting used.

Also rename to earphone-det instead of -ndet

> +               GPIO_LOOKUP("gpio-pxa", GPIO92_HX4700_HP_DRIVER,
> +                           "hp-driver", GPIO_ACTIVE_HIGH),
> +               GPIO_LOOKUP("gpio-pxa", GPIO107_HX4700_SPK_nSD,
> +                           "spk-nsd", GPIO_ACTIVE_HIGH),

Same here. Rename spk-sd

Yours,
Linus Walleij

On Mon, May 2, 2022 at 9:08 AM Arnd Bergmann <arnd@kernel.org> wrote:
> On Sun, May 1, 2022 at 11:41 PM Linus Walleij <linus.walleij@linaro.org> wrote:
> >
> > (...)
> > > +static struct gpiod_lookup_table hx4700_audio_gpio_table = {
> > > +       .dev_id = "hx4700-audio",
> > > +       .table = {
> > > +               GPIO_LOOKUP("gpio-pxa", GPIO75_HX4700_EARPHONE_nDET,
> > > +                           "earphone-ndet", GPIO_ACTIVE_HIGH),
> >
> > This looks wrong. The n in nDET in the end of the name of the GPIO line
> > means active low does it not?
> >
> > What I usually do when I see this is to properly set it to
> > GPIO_ACTIVE_LOW in the descriptor table, then invert the logic
> > where it's getting used.
> >
> > Also rename to earphone-det instead of -ndet
>
> Thanks for taking a look! I changed it now, but I don't know if
> I got the correct number of inversions in the end. How does this look?

Looks wrong, you can just invert the argument to any statement of set_value()
after tagging respective line as active low. Then gpilob will do a second
inversion.

> +               GPIO_LOOKUP("gpio-pxa", GPIO75_HX4700_EARPHONE_nDET,
> +                           "earphone-det", GPIO_ACTIVE_LOW),
> +               GPIO_LOOKUP("gpio-pxa", GPIO107_HX4700_SPK_nSD,
> +                           "spk-sd", GPIO_ACTIVE_LOW),

So those two have switched polarity.

> @@ -81,14 +79,14 @@ static const struct snd_soc_ops hx4700_ops = {
>  static int hx4700_spk_power(struct snd_soc_dapm_widget *w,
>                             struct snd_kcontrol *k, int event)
>  {
> -       gpio_set_value(GPIO107_HX4700_SPK_nSD, !!SND_SOC_DAPM_EVENT_ON(event));
> +       gpiod_set_value(gpiod_spk_sd, !!SND_SOC_DAPM_EVENT_ON(event));

Thus drop one ! in front of the expression, just !SND_SOC_DAPM_EVENT_ON(event)

> -       gpio_set_value(GPIO92_HX4700_HP_DRIVER, !!SND_SOC_DAPM_EVENT_ON(event));
> +       gpiod_set_value(gpiod_hp_driver, !!SND_SOC_DAPM_EVENT_ON(event));

But not this.

> +       gpiod_spk_sd = devm_gpiod_get(&pdev->dev, "spk-sd", GPIOD_OUT_LOW);

These initial values don't seem to be set in the old code you could
just use GPIOD_ASIS as flag to make sure the new code behaves
the same.

Yours,
Linus Walleij

On Thu, May 05, 2022 at 04:33:06PM +0200, Linus Walleij wrote:
> On Thu, May 5, 2022 at 8:04 AM Arnd Bergmann <arnd@kernel.org> wrote:

> > /* Headphones jack detection DAPM pin */
> > static struct snd_soc_jack_pin hs_jack_pin[] = {
> >         {
> >                 .pin    = "Headphone Jack",
> >                 .mask   = SND_JACK_HEADPHONE,
> >         },
> >         {
> >                 .pin    = "Speaker",
> >                 /* disable speaker when hp jack is inserted */
> >                 .mask   = SND_JACK_HEADPHONE,
> >                 .invert = 1,
> >         },

> Hm some ASoC thingie. No idea what that is, but I suppose another
> place where a subsystem for legacy reasons try to do the gpiolib
> inversion on it's own accord. That one isn't flagged as active low in the
> descriptor so it's fine I guess.

It's saying that when the headphone is inserted the headphone output
should be enabled and the speaker output should be disabled, and vice
versa.

> Possible this should be fixed in ASoC to rely on gpiolib but we can't
> fix the entire world.

I don't think so.

On Thu, May 05, 2022 at 04:59:35PM +0200, Arnd Bergmann wrote:
> On Thu, May 5, 2022 at 4:39 PM Mark Brown <broonie@kernel.org> wrote:
> > On Thu, May 05, 2022 at 04:33:06PM +0200, Linus Walleij wrote:
> > > On Thu, May 5, 2022 at 8:04 AM Arnd Bergmann <arnd@kernel.org> wrote:

> > > > static struct snd_soc_jack_pin hs_jack_pin[] = {
> > > >         {
> > > >                 .pin    = "Headphone Jack",
> > > >                 .mask   = SND_JACK_HEADPHONE,
> > > >         },
> > > >         {
> > > >                 .pin    = "Speaker",
> > > >                 /* disable speaker when hp jack is inserted */
> > > >                 .mask   = SND_JACK_HEADPHONE,
> > > >                 .invert = 1,
> > > >         },

> > > Hm some ASoC thingie. No idea what that is, but I suppose another
> > > place where a subsystem for legacy reasons try to do the gpiolib
> > > inversion on it's own accord. That one isn't flagged as active low in the
> > > descriptor so it's fine I guess.

> > It's saying that when the headphone is inserted the headphone output
> > should be enabled and the speaker output should be disabled, and vice
> > versa.

> Ok, that sounds like I should remove the flag here if I declare the
> GPIO line as GPIO_ACTIVE_LOW instead of GPIO_ACTIVE_HIGH, right?

If you change the sense of the GPIO you'll need to flip the invert to
the headphone instead of the speaker - whichever way round the GPIO
sense is each of the pins should be taking the opposite sense from the
GPIO state to the other.