[PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Geert Uytterhoeven]

The Intel Management Engine Interface is only present on x86 platforms
and Intel Xe graphics cards.  Hence add a dependency on X86 or DRM_XE,
to prevent asking the user about this driver when configuring a kernel
for a non-x86 architecture and without Xe graphics support.

Fixes: 25f9b0d351552f63 ("misc/mei: Allow building Intel ME interface on non-x86")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
 drivers/misc/mei/Kconfig | 1 +
 1 file changed, 1 insertion(+)

diff --git a/drivers/misc/mei/Kconfig b/drivers/misc/mei/Kconfig
index 5902dd1ee44baf16..094fb1dde0fe80c5 100644
--- a/drivers/misc/mei/Kconfig
+++ b/drivers/misc/mei/Kconfig
@@ -3,6 +3,7 @@
 config INTEL_MEI
 	tristate "Intel Management Engine Interface"
 	depends on PCI
+	depends on X86 || DRM_XE!=n || COMPILE_TEST
 	default X86_64 || MATOM
 	help
 	  The Intel Management Engine (Intel ME) provides Manageability,
-- 
2.43.0

RE: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Usyskin, Alexander]

> Subject: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE
> 
> The Intel Management Engine Interface is only present on x86 platforms
> and Intel Xe graphics cards.  Hence add a dependency on X86 or DRM_XE,
> to prevent asking the user about this driver when configuring a kernel
> for a non-x86 architecture and without Xe graphics support.
> 
> Fixes: 25f9b0d351552f63 ("misc/mei: Allow building Intel ME interface on
> non-x86")
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
>  drivers/misc/mei/Kconfig | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/misc/mei/Kconfig b/drivers/misc/mei/Kconfig
> index 5902dd1ee44baf16..094fb1dde0fe80c5 100644
> --- a/drivers/misc/mei/Kconfig
> +++ b/drivers/misc/mei/Kconfig
> @@ -3,6 +3,7 @@
>  config INTEL_MEI
>  	tristate "Intel Management Engine Interface"
>  	depends on PCI
> +	depends on X86 || DRM_XE!=n || COMPILE_TEST

Add DRM_I915 here too, MEI may depend on it.

- - 
Thanks,
Sasha



>  	default X86_64 || MATOM
>  	help
>  	  The Intel Management Engine (Intel ME) provides Manageability,
> --
> 2.43.0

Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Arnd Bergmann]

On Wed, Jan 28, 2026, at 07:41, Usyskin, Alexander wrote:
>> Subject: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE
>> 
>> The Intel Management Engine Interface is only present on x86 platforms
>> and Intel Xe graphics cards.  Hence add a dependency on X86 or DRM_XE,
>> to prevent asking the user about this driver when configuring a kernel
>> for a non-x86 architecture and without Xe graphics support.
>> 
>> Fixes: 25f9b0d351552f63 ("misc/mei: Allow building Intel ME interface on
>> non-x86")
>> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
>> ---
>>  drivers/misc/mei/Kconfig | 1 +
>>  1 file changed, 1 insertion(+)
>> 
>> diff --git a/drivers/misc/mei/Kconfig b/drivers/misc/mei/Kconfig
>> index 5902dd1ee44baf16..094fb1dde0fe80c5 100644
>> --- a/drivers/misc/mei/Kconfig
>> +++ b/drivers/misc/mei/Kconfig
>> @@ -3,6 +3,7 @@
>>  config INTEL_MEI
>>  	tristate "Intel Management Engine Interface"
>>  	depends on PCI
>> +	depends on X86 || DRM_XE!=n || COMPILE_TEST
>
> Add DRM_I915 here too, MEI may depend on it.

Based on the patch description, shouldn't the dependency allow
compile-testing only if both XE and I915 are turned off, as in

      depends on X86  || (DRM_XE=n && DRM_I915=n && COMPILE_TEST)

?

As far as I can tell, i915 and xe will attempt (and fail) to
use MEI if that is enabled, so there is no point in enabling
INTEL_MEI outside of compile-testing even if XE is built.

       Arnd

RE: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Usyskin, Alexander]

> Subject: Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE
> 
> On Wed, Jan 28, 2026, at 07:41, Usyskin, Alexander wrote:
> >> Subject: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE
> >>
> >> The Intel Management Engine Interface is only present on x86 platforms
> >> and Intel Xe graphics cards.  Hence add a dependency on X86 or DRM_XE,
> >> to prevent asking the user about this driver when configuring a kernel
> >> for a non-x86 architecture and without Xe graphics support.
> >>
> >> Fixes: 25f9b0d351552f63 ("misc/mei: Allow building Intel ME interface on
> >> non-x86")
> >> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> >> ---
> >>  drivers/misc/mei/Kconfig | 1 +
> >>  1 file changed, 1 insertion(+)
> >>
> >> diff --git a/drivers/misc/mei/Kconfig b/drivers/misc/mei/Kconfig
> >> index 5902dd1ee44baf16..094fb1dde0fe80c5 100644
> >> --- a/drivers/misc/mei/Kconfig
> >> +++ b/drivers/misc/mei/Kconfig
> >> @@ -3,6 +3,7 @@
> >>  config INTEL_MEI
> >>  	tristate "Intel Management Engine Interface"
> >>  	depends on PCI
> >> +	depends on X86 || DRM_XE!=n || COMPILE_TEST
> >
> > Add DRM_I915 here too, MEI may depend on it.
> 
> Based on the patch description, shouldn't the dependency allow
> compile-testing only if both XE and I915 are turned off, as in
> 
>       depends on X86  || (DRM_XE=n && DRM_I915=n && COMPILE_TEST)
> 
> ?
> 
> As far as I can tell, i915 and xe will attempt (and fail) to
> use MEI if that is enabled, so there is no point in enabling
> INTEL_MEI outside of compile-testing even if XE is built.
> 

There are different usages of mei:
- for discrete graphics xe/i915 will load mei via auxiliary device also on non x86.
- for integrated xe/i915 communicate with mei for on-chip CSME - this scenario is x86 only.

IMO this should cover all:
depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST

- - 
Thanks,
Sasha

Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Arnd Bergmann]

On Wed, Jan 28, 2026, at 07:57, Usyskin, Alexander wrote:
>> Subject: Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE
>> As far as I can tell, i915 and xe will attempt (and fail) to
>> use MEI if that is enabled, so there is no point in enabling
>> INTEL_MEI outside of compile-testing even if XE is built.
>> 
>
> There are different usages of mei:
> - for discrete graphics xe/i915 will load mei via auxiliary device also 
> on non x86.

Right, I see.

> - for integrated xe/i915 communicate with mei for on-chip CSME - this 
> scenario is x86 only.
>
> IMO this should cover all:
> depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST

Ok

        Arnd

Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Simon Richter]

Hi,

On 1/28/26 3:57 PM, Usyskin, Alexander wrote:

> There are different usages of mei:
> - for discrete graphics xe/i915 will load mei via auxiliary device also on non x86.
> - for integrated xe/i915 communicate with mei for on-chip CSME - this scenario is x86 only.

> IMO this should cover all:
> depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST

Is there discrete i915?

DRM_I915 depends on PCI && X86 (so "X86 || DRM_I915" is redundant, which 
is why I didn't add it in my version of the patch. If DRM_I915 is useful 
on non-x86 we might want to change that as well.

Also, is there any policy on dependency ordering? On non-X86, INTEL_MEI 
becomes available only after selecting DRM_XE, which is in a directory 
included much later, so the question is not shown initially, is that a 
problem?

    Simon

RE: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Usyskin, Alexander]

> Subject: Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE
> 
> Hi,
> 
> On 1/28/26 3:57 PM, Usyskin, Alexander wrote:
> 
> > There are different usages of mei:
> > - for discrete graphics xe/i915 will load mei via auxiliary device also on non
> x86.
> > - for integrated xe/i915 communicate with mei for on-chip CSME - this
> scenario is x86 only.
> 
> > IMO this should cover all:
> > depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST
> 
> Is there discrete i915?
> 

DG1, Alchemist and some server cards are supported by i915

> DRM_I915 depends on PCI && X86 (so "X86 || DRM_I915" is redundant,
> which
> is why I didn't add it in my version of the patch. If DRM_I915 is useful
> on non-x86 we might want to change that as well.
> 

Let ask gfx people.
Rodrigo, is DRM_I915 useful on non-x86?

> Also, is there any policy on dependency ordering? On non-X86, INTEL_MEI
> becomes available only after selecting DRM_XE, which is in a directory
> included much later, so the question is not shown initially, is that a
> problem?
> 

By not enabling INTEL_MEI the biggest thing that you loose are the firmware update.
I'm not sure how to sort this out and if it is worth the effort.

- - 
Thanks,
Sasha

Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Rodrigo Vivi]

On Thu, Jan 29, 2026 at 12:53:12AM -0500, Usyskin, Alexander wrote:
> > Subject: Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE
> > 
> > Hi,
> > 
> > On 1/28/26 3:57 PM, Usyskin, Alexander wrote:
> > 
> > > There are different usages of mei:
> > > - for discrete graphics xe/i915 will load mei via auxiliary device also on non
> > x86.
> > > - for integrated xe/i915 communicate with mei for on-chip CSME - this
> > scenario is x86 only.
> > 
> > > IMO this should cover all:
> > > depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST
> > 
> > Is there discrete i915?
> > 
> 
> DG1, Alchemist and some server cards are supported by i915
> 
> > DRM_I915 depends on PCI && X86 (so "X86 || DRM_I915" is redundant,
> > which
> > is why I didn't add it in my version of the patch. If DRM_I915 is useful
> > on non-x86 we might want to change that as well.
> > 
> 
> Let ask gfx people.

Cc: intel-gfx@lists.freedesktop.org

> Rodrigo, is DRM_I915 useful on non-x86?

No, I really don't believe it would work out.

> 
> > Also, is there any policy on dependency ordering? On non-X86, INTEL_MEI
> > becomes available only after selecting DRM_XE, which is in a directory
> > included much later, so the question is not shown initially, is that a
> > problem?
> > 
> 
> By not enabling INTEL_MEI the biggest thing that you loose are the firmware update.
> I'm not sure how to sort this out and if it is worth the effort.
> 
> - - 
> Thanks,
> Sasha
> 
> 

RE: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Usyskin, Alexander]

> > >
> > > > There are different usages of mei:
> > > > - for discrete graphics xe/i915 will load mei via auxiliary device also on non
> > > x86.
> > > > - for integrated xe/i915 communicate with mei for on-chip CSME - this
> > > scenario is x86 only.
> > >
> > > > IMO this should cover all:
> > > > depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST
> > >
> > > Is there discrete i915?
> > >
> >
> > DG1, Alchemist and some server cards are supported by i915
> >
> > > DRM_I915 depends on PCI && X86 (so "X86 || DRM_I915" is redundant,
> > > which
> > > is why I didn't add it in my version of the patch. If DRM_I915 is useful
> > > on non-x86 we might want to change that as well.
> > >
> >
> > Let ask gfx people.
> 
> Cc: intel-gfx@lists.freedesktop.org
> 
> > Rodrigo, is DRM_I915 useful on non-x86?
> 
> No, I really don't believe it would work out.
> 

In this case the below should be good, I suppose:

depends on X86 || DRM_XE!=n || COMPILE_TEST

- - 
Thanks,
Sasha

Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Geert Uytterhoeven]

Hi Sasha

On Sun, 1 Feb 2026 at 11:15, Usyskin, Alexander
<alexander.usyskin@intel.com> wrote:
> > > > > There are different usages of mei:
> > > > > - for discrete graphics xe/i915 will load mei via auxiliary device also on non
> > > > x86.
> > > > > - for integrated xe/i915 communicate with mei for on-chip CSME - this
> > > > scenario is x86 only.
> > > >
> > > > > IMO this should cover all:
> > > > > depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST
> > > >
> > > > Is there discrete i915?
> > >
> > > DG1, Alchemist and some server cards are supported by i915
> > >
> > > > DRM_I915 depends on PCI && X86 (so "X86 || DRM_I915" is redundant,
> > > > which
> > > > is why I didn't add it in my version of the patch. If DRM_I915 is useful
> > > > on non-x86 we might want to change that as well.
> > >
> > > Let ask gfx people.
> >
> > Cc: intel-gfx@lists.freedesktop.org
> >
> > > Rodrigo, is DRM_I915 useful on non-x86?
> >
> > No, I really don't believe it would work out.
>
> In this case the below should be good, I suppose:
>
> depends on X86 || DRM_XE!=n || COMPILE_TEST

... which is exactly what I added in my patch. So life's good?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

RE: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Usyskin, Alexander]

> Subject: Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE
> 
> Hi Sasha
> 
> On Sun, 1 Feb 2026 at 11:15, Usyskin, Alexander
> <alexander.usyskin@intel.com> wrote:
> > > > > > There are different usages of mei:
> > > > > > - for discrete graphics xe/i915 will load mei via auxiliary device also on
> non
> > > > > x86.
> > > > > > - for integrated xe/i915 communicate with mei for on-chip CSME - this
> > > > > scenario is x86 only.
> > > > >
> > > > > > IMO this should cover all:
> > > > > > depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST
> > > > >
> > > > > Is there discrete i915?
> > > >
> > > > DG1, Alchemist and some server cards are supported by i915
> > > >
> > > > > DRM_I915 depends on PCI && X86 (so "X86 || DRM_I915" is
> redundant,
> > > > > which
> > > > > is why I didn't add it in my version of the patch. If DRM_I915 is useful
> > > > > on non-x86 we might want to change that as well.
> > > >
> > > > Let ask gfx people.
> > >
> > > Cc: intel-gfx@lists.freedesktop.org
> > >
> > > > Rodrigo, is DRM_I915 useful on non-x86?
> > >
> > > No, I really don't believe it would work out.
> >
> > In this case the below should be good, I suppose:
> >
> > depends on X86 || DRM_XE!=n || COMPILE_TEST
> 
> ... which is exactly what I added in my patch. So life's good?
> 

From my point of view now - yes, it is reasonable.

- - 
Thanks,
Sasha

Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Geert Uytterhoeven]

Hi Simon,

On Thu, 29 Jan 2026 at 05:45, Simon Richter <Simon.Richter@hogyros.de> wrote:
> On 1/28/26 3:57 PM, Usyskin, Alexander wrote:
> > There are different usages of mei:
> > - for discrete graphics xe/i915 will load mei via auxiliary device also on non x86.
> > - for integrated xe/i915 communicate with mei for on-chip CSME - this scenario is x86 only.
>
> > IMO this should cover all:
> > depends on X86 || DRM_XE!=n || DRM_I915!=n || COMPILE_TEST
>
> Is there discrete i915?
>
> DRM_I915 depends on PCI && X86 (so "X86 || DRM_I915" is redundant, which
> is why I didn't add it in my version of the patch. If DRM_I915 is useful
> on non-x86 we might want to change that as well.

Then the dependency of DRM_I915 on X86 should be dropped first...

> Also, is there any policy on dependency ordering? On non-X86, INTEL_MEI
> becomes available only after selecting DRM_XE, which is in a directory
> included much later, so the question is not shown initially, is that a
> problem?

Let's give it a try on arm64.
Remove "# CONFIG_DRM_XE is not set" from .config.

    $ make oldconfig
    [...]
      Intel Xe2 Graphics (DRM_XE) [N/y/?] (NEW) y
    [...]
    Intel Management Engine Interface (INTEL_MEI) [N/y/?] (NEW) y
      ME Enabled Intel Chipsets (INTEL_MEI_ME) [Y/n/?] (NEW)
    [...]

So the ordering is not an issue.

I only wonder about some of the other INTEL_MEI_* options.
I think e.g. INTEL_MEI_ME now needs a dependency on X86, as I doubt
any of these Intel Chipsets are used on non-x86?

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Arnd Bergmann]

On Thu, Jan 29, 2026, at 09:17, Geert Uytterhoeven wrote:
> On Thu, 29 Jan 2026 at 05:45, Simon Richter <Simon.Richter@hogyros.de> wrote:
>>
>> DRM_I915 depends on PCI && X86 (so "X86 || DRM_I915" is redundant, which
>> is why I didn't add it in my version of the patch. If DRM_I915 is useful
>> on non-x86 we might want to change that as well.
>
> Then the dependency of DRM_I915 on X86 should be dropped first...

This is not easily done, with the amount of 'select' statements
from the I915 drive, including X86_PLATFORM_DEVICES and ACPI_WMI.

At the minimum, changing this should involve turning a lot of
these into 'depends on' to avoid dependency loops, plus
build testing random configurations.

    Arnd

RE: [PATCH] misc/mei: INTEL_MEI should depend on X86 or DRM_XE

[Usyskin, Alexander]

> 
> > Also, is there any policy on dependency ordering? On non-X86, INTEL_MEI
> > becomes available only after selecting DRM_XE, which is in a directory
> > included much later, so the question is not shown initially, is that a
> > problem?
> 
> Let's give it a try on arm64.
> Remove "# CONFIG_DRM_XE is not set" from .config.
> 
>     $ make oldconfig
>     [...]
>       Intel Xe2 Graphics (DRM_XE) [N/y/?] (NEW) y
>     [...]
>     Intel Management Engine Interface (INTEL_MEI) [N/y/?] (NEW) y
>       ME Enabled Intel Chipsets (INTEL_MEI_ME) [Y/n/?] (NEW)
>     [...]
> 
> So the ordering is not an issue.
> 
> I only wonder about some of the other INTEL_MEI_* options.
> I think e.g. INTEL_MEI_ME now needs a dependency on X86, as I doubt
> any of these Intel Chipsets are used on non-x86?
> 

INTEL_MEI_ME is re-used by INTEL_MEI_GSC (the HW access code, not PCI one),
so it should not be bound to x86.

- - 
Thanks,
Sasha