On Mon, Jun 19, 2023 at 09:30:05AM +0100, Lee Jones wrote:
> On Fri, 16 Jun 2023, Charles Keepax wrote:
> > On Thu, Jun 15, 2023 at 06:11:24PM +0100, Lee Jones wrote:
> > > On Mon, 05 Jun 2023, Charles Keepax wrote:
> > > > +static struct i2c_device_id cs42l43_i2c_id[] = {
> > > > + { "cs42l43", 0 },
> > > > + {}
> > > > +};
> > > > +MODULE_DEVICE_TABLE(i2c, cs42l43_i2c_id);
> > >
> > > Is this required anymore?
> > >
> >
> > I was not aware of it not being required, I think it will still
> > be used for the purposes of module naming. Perhaps someone more
> > knowledgable than me can comment?
>
> Since this table isn't providing any information which cannot be derived
> from the other (OF, ACPI) tables, the I2C subsystem should be able to
> obtain it from those sources instead.
>
Sorry I literally just sent a v4 then saw this email. I will test
removing this table and send a v5.
> > > > +#if IS_ENABLED(CONFIG_MFD_CS42L43_I2C)
> > > > +const struct regmap_config cs42l43_i2c_regmap = {
> > > > + .reg_bits = 32,
> > > > + .reg_stride = 4,
> > > > + .val_bits = 32,
> > > > + .reg_format_endian = REGMAP_ENDIAN_BIG,
> > > > + .val_format_endian = REGMAP_ENDIAN_BIG,
> > > > +
> > > > + .max_register = CS42L43_MCU_RAM_MAX,
> > > > + .readable_reg = cs42l43_readable_register,
> > > > + .volatile_reg = cs42l43_volatile_register,
> > > > + .precious_reg = cs42l43_precious_register,
> > > > +
> > > > + .cache_type = REGCACHE_RBTREE,
> > > > + .reg_defaults = cs42l43_reg_default,
> > > > + .num_reg_defaults = ARRAY_SIZE(cs42l43_reg_default),
> > > > +};
> > > > +EXPORT_SYMBOL_NS_GPL(cs42l43_i2c_regmap, MFD_CS42L43);
> > > > +#endif
> > >
> > > We don't tend to like #ifery in C files.
> > >
> > > Why is it required?
> > >
> > > And why not just put them were they're consumed?
> >
> > The trouble is the cs42l43_reg_default array and the array size.
> > There is no good way to statically initialise those two fields
> > from a single array in both the I2C and SDW modules.
>
> Can you have a little think for another way to solve this please?
>
I will have another go at it, if memory serves the vague options
were:
1) this approach
2) some sort of horrible #include to put the defaults array in
both modules, although I would really prefer to avoid this one.
3) dynamically allocate the regmap_configs so those two fields
can be filled in with non-static data.
If I fail to come up with an option 4 would you prefer 1 or 3?
Well or 2 but I really would prefer not to do 2.
> > > Perhaps some simple function headers would help?
> > You mean add some kernel doc for these functions, right? Assuming
> > that is what you mean, will do.
>
> I'd suggest not using kernel-doc formatting, but that type of thing,
> yes.
Ok I will remove the kernel doc bits for v5.
Thanks,
Charles
