[PATCH v2 2/2] ASoC: codecs: add support for Everest ES7202 PDM ADC

[Jun Yan]

Hi,

Thanks for your review.

> > Add support for the Everest ES7202, a 2-channel PDM analog-to-digital
> > converter controlled over I2C. The device outputs PDM data to the SoC
> > PDM controller and is configured through registers for supply voltage,
> > analog front-end and per-channel PGA gain.
> 
> This looks pretty good, a few relatively minor issues below but nothing
> major.
> 
> > --- a/sound/soc/codecs/Kconfig
> > +++ b/sound/soc/codecs/Kconfig
> > @@ -1274,6 +1274,16 @@ config SND_SOC_HDMI_CODEC
> >  config SND_SOC_ES7134
> >  	tristate "Everest Semi ES7134 CODEC"
> >  
> > +config SND_SOC_ES7202
> > +	tristate "Everest Semi ES7202 CODEC"
> > +	depends on I2C
> > +	select REGMAP_I2C
> 
> You should also add this to SND_SOC_ALL_CODECS.

Added.

> 
> > --- /dev/null
> > +++ b/sound/soc/codecs/es7202.c
> > @@ -0,0 +1,367 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Copyright (C) 2020 Everest Semiconductor Co Ltd.,
> > + * Copyright (C) 2026 Jun Yan <jerrysteve1101 at gmail.com>
> > + */
> 
> Please make the whole comment block C++ so things look more intentional.

Done.

> 
> > +#include <linux/bitfield.h>
> > +#include <linux/delay.h>
> > +#include <linux/device.h>
> > +#include <linux/gpio/consumer.h>
> > +#include <linux/i2c.h>
> > +#include <linux/module.h>
> > +#include <linux/of.h>
> > +#include <linux/regmap.h>
> > +#include <linux/regulator/consumer.h>
> > +#include <sound/pcm.h>
> > +#include <sound/pcm_params.h>
> 
> linux/of.h and sound/pcm_params.h aren't needed I think.
> 
Right, both removed.

> > +static bool es7202_volatile_reg(struct device *dev, unsigned int reg)
> > +{
> > +	/* status and device-id registers are read-only, never cached */
> > +	return reg == ES7202_CHIP_STA ||
> > +	       reg == ES7202_DEV_ID1 ||
> > +	       reg == ES7202_DEV_ID0;
> > +}
> 
> A switch statement would be neater and would probably generate
> equivalent code.  Are you sure that the ID registers are actually
> volatile, that's quite unusual?  If they've volatile the hardware is
> accessed on each read, if you just omit the default then they will be
> cached after the first read.
> 

Thanks for catching this. The ID registers are static and don't need
volatile marking. I'll drop them from es7202_volatile_reg() and use
.writeable_reg to make them read-only, so they are still cached after
the first read but skipped by regcache_sync().

> Also should the rest register be uncached so we don't try to write it
> back out as part of a cache sync by mistake?
> 

Agreed, I missed this. I'll mark ES7202_RESET volatile so it can't be
written back during a cache sync.

> > +static const struct regmap_config es7202_regmap = {
> > +	.reg_bits	= 8,
> > +	.val_bits	= 8,
> > +	.max_register	= ES7202_DEV_ID0,
> > +	.reg_defaults	= es7202_reg_defaults,
> > +	.num_reg_defaults = ARRAY_SIZE(es7202_reg_defaults),
> > +	.volatile_reg	= es7202_volatile_reg,
> > +	.cache_type	= REGCACHE_MAPLE,
> 
> I notice that other Everest devices set use_single_read and
> use_single_write - especially for write we can easily trigger
> multi-register write as part of a regcache sync so if the device lacks
> support it'll create bugs.
> 

Thanks. The other Everest codecs are the only precedent here - the
ES7202 datasheet does not document multi-register (auto-increment)
writes, so I cannot rely on them. I'll set use_single_read and
use_single_write so a regcache sync can never turn into a burst write.

> > +	es7202->supply = devm_regulator_get_optional(dev, "vdd");
> > +	if (IS_ERR(es7202->supply)) {
> 
> Unless the device can operate without power this should be a normal
> regulator_get(), the core will substitude a dummy regulator if it's not
> configurd by the firmware.
> 

Right, VDD supply is required. I'll use devm_regulator_get() instead.

I'll send a v3 with these changes.

> -- 
> Mark Brown <broonie at kernel.org>