Re: [PATCH] ASoC: rt5677: Reintroduce I2C device IDs

[Andy Shevchenko]

On Fri, 2017-08-25 at 17:05 +0100, John Keeping wrote:
> On Fri, 25 Aug 2017 10:24:26 -0400, Tom Rini wrote:
> > On Fri, Aug 25, 2017 at 04:56:47PM +0300, Andy Shevchenko wrote:

> > > Apparently you are the one who tested the commit
> > > 	89128534f925 ("ASoC: rt5677: Add ACPI support")
> > > year ago.ÂÂ
> > 
> > Yes.
> > 
> > > The commit states that ACPI properties that are used in Chromebook
> > > Pixel
> > > 2015 is non-standard (not the same as for DT).
> > > 
> > > However, DSDT shows the opposite!ÂÂ
> > 
> > Interesting.ÂÂI'm not an ACPI person, I just tested what John came
> > up
> > with.
> 
> And the patch adding this was the first (and still only) time I've
> really looked at ACPI, so it's quite possible that I misunderstood
> something at the time.

Maybe.

> From memory, I think the particular problem I was referring to in the
> commit message was that certain GPIOs were only defined by index and
> not
> by property name (specifically "plug-det-gpios", "mic-present-gpios"
> and
> "headphone-enable-gpios"), and having dumped DSDT just now I do not
> see
> those strings appearing anywhere.

Exactly, and this part of the patch I'm _not_ talking about (it's pretty
much good and working).

What I'm talking about is a specific function called

rt5677_read_acpi_properties()

in the rt5677.c codec driver.

The question is do we have _real publicly available_ hardware with that
kind of properties?

Because now it's a mess (wrt to real DSDT attached to the thread).

The proposed change to fix this is like

diff --git a/sound/soc/codecs/rt5677.c b/sound/soc/codecs/rt5677.c
index 6448b7a00203..28bde5f50ed9 100644
--- a/sound/soc/codecs/rt5677.c
+++ b/sound/soc/codecs/rt5677.c
@@ -5145,20 +5145,18 @@ static int rt5677_i2c_probe(struct i2c_client
*i2c)
Â		match_id = of_match_device(rt5677_of_match, &i2c->dev);
Â		if (match_id)
Â			rt5677->type = (enum rt5677_type)match_id-
>data;
-
-		rt5677_read_device_properties(rt5677, &i2c->dev);
Â	} else if (ACPI_HANDLE(&i2c->dev)) {
Â		const struct acpi_device_id *acpi_id;
Â
Â		acpi_id = acpi_match_device(rt5677_acpi_match, &i2c-
>dev);
Â		if (acpi_id)
Â			rt5677->type = (enum rt5677_type)acpi_id-
>driver_data;
-
-		rt5677_read_acpi_properties(rt5677, &i2c->dev);
Â	} else {
Â		return -EINVAL;
Â	}
Â
+	rt5677_read_device_properties(rt5677, &i2c->dev);
+
Â	/* pow-ldo2 and reset are optional. The codec pins may be
statically
Â	Â* connected on the board without gpios. If the gpio device
property
Â	Â* isn't specified, devm_gpiod_get_optional returns NULL.

+ removing rt5677_read_acpi_properties() completely.

Tom, if you can test it (basic test + might be quality of sound) on your
Chromebook, it would be nice!


-- 
Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
Intel Finland Oy