Re: [PATCH v3 1/9] clk: bcm: kona: Move CLOCK_COUNT defines into the driver

[Artur Weber]

On 10.03.2025 09:40, Krzysztof Kozlowski wrote:
> On Sat, Mar 08, 2025 at 08:50:39AM +0100, Artur Weber wrote:
> > CLOCK_COUNT defines for each CCU are stored in the DT binding header.
> > This is not correct - they are not used by device trees, only internally
> > by the driver.
> >
> > Move the CLOCK_COUNT defines directly into the driver in preparation
> > for dropping them from the DT binding include.
> >
> > Signed-off-by: Artur Weber <aweber.kernel@xxxxxxxxx>
> > ---
> >   drivers/clk/bcm/clk-bcm21664.c |  8 ++++++++
> >   drivers/clk/bcm/clk-bcm281xx.c | 10 ++++++++++
> >   2 files changed, 18 insertions(+)
> >
> > diff --git a/drivers/clk/bcm/clk-bcm21664.c b/drivers/clk/bcm/clk-bcm21664.c
> > index 520c3aeb4ea9c4a431512c0909f9545c1761d17a..fa6e1649d6f5f459b63026109caea9e2f72e22dd 100644
> > --- a/drivers/clk/bcm/clk-bcm21664.c
> > +++ b/drivers/clk/bcm/clk-bcm21664.c
> > @@ -17,6 +17,8 @@ static struct peri_clk_data frac_1m_data = {
> >   	.clocks		= CLOCKS("ref_crystal"),
> >   };
> >   
> > +#define BCM21664_ROOT_CCU_CLOCK_COUNT	(BCM21664_ROOT_CCU_FRAC_1M + 1)
>
> I hit that wall too, no worries. It might surprise you but 0+1 != 1 :),

Do you mean that I should specify the clock count directly rather than
incrementing the last ID? Some other drivers seem to do this the way I
did here (samsung/clk-exynos*, renesas/r9a06g032-clocks.c).

> so you redefine a define. You need to test this patch bisectability.

I assume renaming the define to not collide with the old name is the way
to go?

Best regards
Artur