Re: [PATCH] firmware: arm_scmi: Drop the ARM architecture dependency
From: Sudeep Holla
Date: Fri Nov 20 2026 - 09:16:54 EST
On Fri, Nov 20, 2026 at 01:37:13PM +0100, Michal Simek wrote:
>
>
> On 11/20/26 09:51, Geert Uytterhoeven wrote:
> > Hi Han,
> >
> > On Fri, 20 Nov 2026 at 09:43, Han Gao <rabenda.cn@xxxxxxxxx> wrote:
> > > On Fri, 20 Nov 2026 09:26:47 +0100 Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> > > > On Fri, 6 Nov 2026 at 20:40, Han Gao via B4 Relay
> > > > <devnull+gaohan.iscas.ac.cn@xxxxxxxxxx> wrote:
> > > > > From: Han Gao <gaohan@xxxxxxxxxxx>
> > > > >
> > > > > The SCMI core and the mailbox and shared memory transports only use
> > > > > generic kernel interfaces. Transports that need architecture support,
> > > > > such as SMC, already carry their own dependencies.
> > > > >
> > > > > Drop the ARM and ARM64 dependency so that other architectures can use
> > > > > SCMI. Depend on HAS_IOMEM instead, for devm_ioremap().
> > > > >
> > > > > Suggested-by: Michal Simek <michal.simek@xxxxxxx>
> > > > > Signed-off-by: Han Gao <gaohan@xxxxxxxxxxx>
> > > >
> > > > Thanks for your patch, which is now commit fe87ede443a15969 ("firmware:
> > > > arm: scmi: Drop the ARM architecture dependency") in scmi/for-linux-next.
> > > >
> > > > > ---
> > > > > This was patch 1 of the Zhihe A210 SoC series [1]. As Michal asked, it
> > > > > now goes on its own through the SCMI tree.
> > > > >
> > > > > Since [1], it depends on HAS_IOMEM: Sashiko found that UML, which has
> > > > > no HAS_IOMEM, failed to link devm_ioremap().
> > > > >
> > > > > Build tested on x86_64 without OF, i386, loongarch64, ppc64le, riscv64
> > > > > and arm64. On UML, SCMI can no longer be selected.
> > > > >
> > > > > [1] https://patch.msgid.link/20261103-a210-scmi-v3-0-ee6b31580808@xxxxxxxxxxx
> > > >
> > > > > --- a/drivers/firmware/arm_scmi/Kconfig
> > > > > +++ b/drivers/firmware/arm_scmi/Kconfig
> > > > > @@ -3,7 +3,7 @@ menu "ARM System Control and Management Interface Protocol"
> > > > >
> > > > > config ARM_SCMI_PROTOCOL
> > > > > tristate "ARM System Control and Management Interface (SCMI) Message Protocol"
> > > > > - depends on ARM || ARM64 || COMPILE_TEST
> > > > > + depends on HAS_IOMEM || COMPILE_TEST
> > > >
> > > > Please extend the dependencies to RISCV (from the link above, I assume
> > > > you are targeting riscv?) instead. There is no point in showing this
> > > > option on architectures that have no users of SCMI; one already has
> > > > too many relevant symbols to configure.
> > > > People who want to build test on x86_64 without OF, i386, loongarch64,
> > > > ppc64le, ... can just enable COMPILE_TEST.
> > >
> > > Following Michal's suggestion[1], I'm dropping ARM/ARM64.
> > >
> > > [1] https://lore.kernel.org/all/3d38c4a5-1a12-4af1-ac2e-413a457ce4c5@xxxxxxx/
> >
> > Yes, I read that.
> >
> > For reference, other firmware interfaces are also not exposed to the
> > whole world:
> >
> > $ git grep -w ARCH_SUPPORTS_ACPI
> > arch/arm64/Kconfig: select ARCH_SUPPORTS_ACPI
> > arch/loongarch/Kconfig: select ARCH_SUPPORTS_ACPI
> > arch/riscv/Kconfig: select ARCH_SUPPORTS_ACPI if 64BIT
> > arch/x86/Kconfig: select ARCH_SUPPORTS_ACPI
> > drivers/acpi/Kconfig:config ARCH_SUPPORTS_ACPI
> > drivers/acpi/Kconfig: depends on ARCH_SUPPORTS_ACPI
> > $ git grep -w "config EFI"
> > arch/arm/Kconfig:config EFI
> > arch/arm64/Kconfig:config EFI
> > arch/loongarch/Kconfig:config EFI
> > arch/riscv/Kconfig:config EFI
> > arch/x86/Kconfig:config EFI
> I can see it's usage with x86 too.
> Microblaze is another one which can also use it.
>
> SCMI from my perspective should be open for any architecture that's why no
> restriction make sense.
>
IIUC, Geert's concern is it now shows up asking the users to choose on all
other architecture. We should have to default off on all archs that don't
need it now and can enable it if needed. I agree and like that approach.
--
Regards,
Sudeep