RE: [PATCH v11 09/11] clk: realtek: Add RTD1625-CRT clock controller driver

> On Mon, Aug 03, 2026 at 09:56:01AM +0000, Yu-Chun Lin [林祐君] wrote:
> > > On Fri, Jul 31, 2026 at 08:17:44AM +0000, Yu-Chun Lin [林祐君] wrote:
> > > > > On Tue, Jul 28, 2026 at 10:28:04PM +0800, Yu-Chun Lin wrote:
> > > > > > From: Cheng-Yu Lee <[email protected]>
> > > > > >
> > > >
> > > > (...)
> > > >
> > > > > > +     { /* sentinel */ }
> > > > > > +};
> > > > > > +MODULE_DEVICE_TABLE(of, rtd1625_crt_match);
> > > > > > +
> > > > > > +static struct platform_driver rtd1625_crt_driver = {
> > > > > > +     .probe = rtd1625_crt_probe,
> > > > > > +     .driver = {
> > > > > > +             .name = "rtk-rtd1625-crt-clk",
> > > > > > +             .of_match_table = rtd1625_crt_match,
> > > > > > +             .suppress_bind_attrs = true,
> > > > >
> > > > > This driver is listed as a tristate in the Kconfig but then you
> > > > > have this. Is the intention of this driver to be built in only?
> > > > >
> > > >
> > > > I will change the Kconfig options for this driver, as well as the
> > > > related config, from tristate to bool.
> > >
> > > Keep it as a tristate if you are going to use module_platform_driver().
> > > You can also drop suppress_bind_attrs unless you have a specific need why.
> > >
> > > Brian
> >
> > You are right. Keeping it tristate is beneficial, especially for
> > aligning with the Android GKI design.
> >
> > To properly support module unloading and driver unbinding, I will add
> > the .remove callback and drop suppress_bind_attrs.
> 
> If you are able to use all devm_*() APIs, then you shouldn't need a remove
> callback.
> 

The reason I planned to add a .remove callback is not to free memory or
unregister clocks, but to clear the regmap pointers.

I want to set desc->clks[i]->regmap = NULL; in .remove callback.

Best Regards,
Yu-Chun

> Brian
>