Re: [PATCH v2 1/4] ASoC: cs4271: Fix cs4271 I2C and SPI drivers automatic module loading

[Alexander Sverdlin <alexander.sverdlin@gmail.com>]

Hi Herve,

On Thu, 2025-10-30 at 14:54 +0100, Alexander Sverdlin wrote:
> > > > --- a/sound/soc/codecs/cs4271-spi.c
> > > > +++ b/sound/soc/codecs/cs4271-spi.c
> > > > @@ -23,11 +23,24 @@ static int cs4271_spi_probe(struct spi_device *spi)
> > > >  	return cs4271_probe(&spi->dev, devm_regmap_init_spi(spi, &config));
> > > >  }
> > > >  
> > > > +static const struct spi_device_id cs4271_id_spi[] = {
> > > > +	{ "cs4271", 0 },
> > > > +	{}
> > > > +};
> > > > +MODULE_DEVICE_TABLE(spi, cs4271_id_spi);
> > > > +
> > > > +static const struct of_device_id cs4271_dt_ids[] = {
> > > > +	{ .compatible = "cirrus,cs4271", },
> > > > +	{ }
> > > > +};
> > > > +MODULE_DEVICE_TABLE(of, cs4271_dt_ids);  
> > > 
> > > So currently SPI core doesn't generate "of:" prefixed uevents, therefore this
> > > currently doesn't help? However, imagine, you'd have both backends enabled
> > > as modules, -spi and -i2c. udev/modprobe currently load just one module it
> > > finds first. What is the guarantee that the loaded module for the "of:"
> > > prefixed I2C uevent would be the -i2c module?
> > > 
> > 
> > I hesitate to fully remove cs4271_dt_ids in the SPI part.
> > 
> > I understood having it could lead to issues if both SPI and I2C parts
> > are compiled as modules but this is the pattern used in quite a lot of
> > drivers.
> > 
> > Maybe this could be handle globally out of this series instead of introducing
> > a specific pattern in this series.
> > 
> > But well, if you and Mark are ok to fully remove the cs4271_dt_ids from the
> > SPI part and so unset the of_match_table from the cs4271_spi_driver, I can
> > do the modification.
> > 
> > Let me know if I should send a new iteration with cs4271_dt_ids fully removed
> > from the SPI part.
> > 
> > Also, last point, I don't have any cs4271 connected to a SPI bus.
> > I use only the I2C version and will not be able to check for correct
> > modifications on the SPI part.
> 
> I'd propose to drop SPI modifications in this case, because by doing this
> you actually introduce yet another problem for the I2C case you are interested
> in (namely if you'd enable both modules).

sorry again for the confusion, I meant only to drop "MODULE_DEVICE_TABLE(of, "
from SPI...

But seems that Mark actually has different opinion... Indeed 

$ grep -F "MODULE_DEVICE_TABLE(of, " *spi*.c | wc -l

currently reports "17" out of 23 SPI-capable CODECs. I just don't see how
you are helping the I2C use case by doing this...

-- 
Alexander Sverdlin.