The RISC-V platform management interface (RPMI) is only available on
RISC-V platforms.  Hence add a dependency on RISCV, to prevent asking
the user about this driver when configuring a kernel for a different
architecture.

Fixes: 5ba9f520f41a33c9 ("clk: Add clock driver for the RISC-V RPMI clock service group")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
And perhaps the "default RISCV" should be dropped, too?
---
 drivers/clk/Kconfig | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/clk/Kconfig b/drivers/clk/Kconfig
index b03fafb402b19153..5c17a22d59920f5e 100644
--- a/drivers/clk/Kconfig
+++ b/drivers/clk/Kconfig
@@ -510,6 +510,7 @@ config COMMON_CLK_LITEX
 
 config COMMON_CLK_RPMI
 	tristate "Clock driver based on RISC-V RPMI"
+	depends on RISCV || COMPILE_TEST
 	depends on MAILBOX
 	default RISCV
 	help
-- 
2.43.0

On Tue, 30 Sep 2025, Geert Uytterhoeven wrote:

> The RISC-V platform management interface (RPMI) is only available on
> RISC-V platforms.  Hence add a dependency on RISCV, to prevent asking
> the user about this driver when configuring a kernel for a different
> architecture.
> 
> Fixes: 5ba9f520f41a33c9 ("clk: Add clock driver for the RISC-V RPMI clock service group")
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>

Thanks Geert for catching this.

This patch is against unmerged patches in -next.  So I'll plan to add this 
to the PR that I plan to send to Linus tomorrow -- unless any of the 
drivers/clk maintainers would prefer that I not.

> And perhaps the "default RISCV" should be dropped, too?

Probably.  I guess we should just add this to the arch/riscv defconfig 
instead.  Let's wait on this one for a few days to see if anyone has any 
comments, and consider that change for v6.18-rc fixes.


thanks again,

- Paul

On Wed, Oct 01, 2025 at 07:15:56PM -0600, Paul Walmsley wrote:
> On Tue, 30 Sep 2025, Geert Uytterhoeven wrote:
> 
> > The RISC-V platform management interface (RPMI) is only available on
> > RISC-V platforms.  Hence add a dependency on RISCV, to prevent asking
> > the user about this driver when configuring a kernel for a different
> > architecture.
> > 
> > Fixes: 5ba9f520f41a33c9 ("clk: Add clock driver for the RISC-V RPMI clock service group")
> > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> 
> Thanks Geert for catching this.
> 
> This patch is against unmerged patches in -next.  So I'll plan to add this 
> to the PR that I plan to send to Linus tomorrow -- unless any of the 
> drivers/clk maintainers would prefer that I not.
> 
> > And perhaps the "default RISCV" should be dropped, too?
> 
> Probably.  I guess we should just add this to the arch/riscv defconfig 
> instead.  Let's wait on this one for a few days to see if anyone has any 
> comments, and consider that change for v6.18-rc fixes.

There's little point having "default RISCV" if it's only available on
RISCV in the first place, may as well just be "default y" and be
simpler.

My 2c is that putting it in defconfig is barely worth doing, unless there
are actual platforms that use it.
Does QEMU provide a useful test for it that exercises the various code
paths, that would make it worthwhile to have in defconfig Anup?

On Thu, Oct 2, 2025 at 3:06 PM Conor Dooley <conor@kernel.org> wrote:
>
> On Wed, Oct 01, 2025 at 07:15:56PM -0600, Paul Walmsley wrote:
> > On Tue, 30 Sep 2025, Geert Uytterhoeven wrote:
> >
> > > The RISC-V platform management interface (RPMI) is only available on
> > > RISC-V platforms.  Hence add a dependency on RISCV, to prevent asking
> > > the user about this driver when configuring a kernel for a different
> > > architecture.
> > >
> > > Fixes: 5ba9f520f41a33c9 ("clk: Add clock driver for the RISC-V RPMI clock service group")
> > > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> >
> > Thanks Geert for catching this.
> >
> > This patch is against unmerged patches in -next.  So I'll plan to add this
> > to the PR that I plan to send to Linus tomorrow -- unless any of the
> > drivers/clk maintainers would prefer that I not.
> >
> > > And perhaps the "default RISCV" should be dropped, too?
> >
> > Probably.  I guess we should just add this to the arch/riscv defconfig
> > instead.  Let's wait on this one for a few days to see if anyone has any
> > comments, and consider that change for v6.18-rc fixes.
>
> There's little point having "default RISCV" if it's only available on
> RISCV in the first place, may as well just be "default y" and be
> simpler.
>
> My 2c is that putting it in defconfig is barely worth doing, unless there
> are actual platforms that use it.
> Does QEMU provide a useful test for it that exercises the various code
> paths, that would make it worthwhile to have in defconfig Anup?

Yes, QEMU RPMI support is in the pipeline. The OpenSBI
upstreaming was completed few months back and Linux patches
are going in this merge window (Linux-6.18)

We (Ventana) have already made the QEMU RPMI implementation
public which can be tried until QEMU RPMI upstreaming is completed.
(Refer, dev-upstream branch of https://github.com/ventanamicro/qemu.git)

NOTE: these details are mentioned in cover-letter as well.
(Refer, https://www.spinics.net/lists/linux-clk/msg118229.html)

Regards,
Anup

On Wed, 1 Oct 2025, Paul Walmsley wrote:

> On Tue, 30 Sep 2025, Geert Uytterhoeven wrote:
> 
> > And perhaps the "default RISCV" should be dropped, too?
> 
> Probably.  I guess we should just add this to the arch/riscv defconfig 
> instead.  Let's wait on this one for a few days to see if anyone has any 
> comments, and consider that change for v6.18-rc fixes.

Sorry, this one got garbled on the way out.  I meant to say that we should 
think about whether it makes sense to use a more fine-grained way to 
enable this by default.


- Paul