Re: [PATCH v1 2/2] ASoC: Add initial ZL38060 driver

[Sven Van Asbroeck]

Thank you Mark for the constructive feedback !
A few follow-up questions below.

On Thu, Apr 16, 2020 at 8:42 AM Mark Brown <broonie@xxxxxxxxxx> wrote:
>
> > +++ b/sound/soc/codecs/zl38060.c
> > @@ -0,0 +1,643 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Codec driver for Microsemi ZL38060 Connected Home Audio Processor.
> > + *
>
> Please make the entire comment a C++ one so things look more
> intentional.

The 'weird' combination of // SPDX and /* Description/copyright */ seems to
be a kernel-wide standard (for C files, at least) ?

E.g.:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/sound/soc/codecs/wm9090.c?h=v5.7-rc1#n2
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/sound/soc/codecs/wm8904.c?h=v5.7-rc1#n2

Ok to keep?

>
> > +skip_setup:
> > +     if (priv->amp_en_gpio && tx) {
> > +             /* enable the external amplifier before playback starts */
> > +             gpiod_set_value_cansleep(priv->amp_en_gpio, 1);
> > +             if (priv->amp_startup_delay_ms)
> > +                     msleep(priv->amp_startup_delay_ms);
> > +     }
>
> This external amplifier support shouldn't be here, if there's other
> devices in the system then they will have their own drivers and the
> machine driver will take care of linking things together.

In our application, the amp is a "dumb" class-D amp with a single enable line:
https://www.onsemi.com/pub/Collateral/FAB3103-D.pdf

I am not sure how I could make this more general. Could you point me to an
example somewhere in the tree?

>
> > +     priv->regmap = devm_regmap_init(dev, &zl38_regmap_bus, spi,
> > +                                     &zl38_regmap_conf);
> > +     if (IS_ERR(priv->regmap))
> > +             return PTR_ERR(priv->regmap);
>
> devm_regmap_init_spi()

I wish !! This chip has complex SPI addressing, using an "address" which:
- is variable length, depending on the page of the register being accessed;
- contains a field with the length of the data to follow.

Unfortunately, during firmware programming, multi-writes are mandatory
(usually address header + 32 data bytes).

Implementing my own regmap_bus looked like the only way out.