Skip to contentSkip to conversation

drm/xe/vsec: fix CONFIG_INTEL_VSEC dependency

PATCH In review Arnd Bergmann 6 participants 9 messages May 23, 2025 – May 28, 2025

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 drm/xe/vsec: fix CONFIG_INTEL_VSEC dependency

From: Arnd Bergmann <[email protected]>

The XE driver can be built with or without VSEC support, but fails to link as built-in if vsec is in a loadable module:

x86_64-linux-ld: vmlinux.o: in function `xe_vsec_init':
(.text+0x1e83e16): undefined reference to `intel_vsec_register'

The normal fix for this is to add a 'depends on INTEL_VSEC || !INTEL_VSEC', forcing XE to be a loadable module as well, but that causes a circular dependency:

symbol DRM_XE depends on INTEL_VSEC symbol INTEL_VSEC depends on X86_PLATFORM_DEVICES symbol X86_PLATFORM_DEVICES is selected by DRM_XE

The problem here is selecting a symbol from another subsystem, so change that as well and rephrase the 'select' into the corresponding dependency. Since X86_PLATFORM_DEVICES is 'default y', there is no change to defconfig builds here.

Fixes 0c45e76fcc62 ("drm/xe/vsec: Support BMG devices")
---
 drivers/gpu/drm/xe/Kconfig | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/xe/Kconfig +2 -1
index beddd153c28c..c870b680431c 100644
@@ -2,6 +2,8 @@
 config DRM_XE
 	tristate "Intel Xe Graphics"
 	depends on DRM && PCI && (m || (y && KUNIT=y))
+	depends on INTEL_VSEC || !INTEL_VSEC
+	depends on INTEL_PLATFORM_DEVICES || !(X86 && ACPI)
 	select INTERVAL_TREE
 	# we need shmfs for the swappable backing store, and in particular
 	# the shmem_readpage() which depends upon tmpfs
@@ -29,7 +31,6 @@ config DRM_XE
 	select INPUT if ACPI
 	select ACPI_VIDEO if X86 && ACPI
 	select ACPI_BUTTON if ACPI
-	select X86_PLATFORM_DEVICES if X86 && ACPI
 	select ACPI_WMI if X86 && ACPI
 	select SYNC_FILE
 	select IOSF_MBI
Signature · 1 line
2.39.5
#1 lore raw

✗ CI.Patch_applied: failure for drm/xe/vsec: fix CONFIG_INTEL_VSEC dependency

== Series Details ==

== Series Details ==

Series: drm/xe/vsec: fix CONFIG_INTEL_VSEC dependency
URL : https://patchwork.freedesktop.org/series/149427/
State : failure

== Summary ==

=== Applying kernel patches on branch 'drm-tip' with base: ===
Base commit: 5ed61f4a36b3 drm-tip: 2025y-05m-23d-13h-19m-23s UTC integration manifest === git am output follows ===
error: patch failed: drivers/gpu/drm/xe/Kconfig:2
error: drivers/gpu/drm/xe/Kconfig: patch does not apply
hint: Use 'git am --show-current-patch=diff' to see the failed patch
Applying: drm/xe/vsec: fix CONFIG_INTEL_VSEC dependency
Patch failed at 0001 drm/xe/vsec: fix CONFIG_INTEL_VSEC dependency
When you have resolved this problem, run "git am --continue".
If you prefer to skip this patch, run "git am --skip" instead.
To restore the original branch and stop patching, run "git am --abort".

#2 lore raw
In reply to Arnd Bergmann
Quoted 35 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>

The XE driver can be built with or without VSEC support, but fails to link as built-in if vsec is in a loadable module:

x86_64-linux-ld: vmlinux.o: in function `xe_vsec_init':
(.text+0x1e83e16): undefined reference to `intel_vsec_register'

The normal fix for this is to add a 'depends on INTEL_VSEC || !INTEL_VSEC', forcing XE to be a loadable module as well, but that causes a circular dependency:

symbol DRM_XE depends on INTEL_VSEC symbol INTEL_VSEC depends on X86_PLATFORM_DEVICES symbol X86_PLATFORM_DEVICES is selected by DRM_XE

The problem here is selecting a symbol from another subsystem, so change that as well and rephrase the 'select' into the corresponding dependency. Since X86_PLATFORM_DEVICES is 'default y', there is no change to defconfig builds here.

Fixes 0c45e76fcc62 ("drm/xe/vsec: Support BMG devices")

---
drivers/gpu/drm/xe/Kconfig | 3 ++-

1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/xe/Kconfig +0 -0
index beddd153c28c..c870b680431c 100644
@@ -2,6 +2,8 @@

config DRM_XE

	tristate "Intel Xe Graphics"
	depends on DRM && PCI && (m || (y && KUNIT=y))

drivers/gpu/drm/xe/Kconfig line 6

Quoted 2 lines
drivers/gpu/drm/xe/Kconfig +2 -0
+	depends on INTEL_VSEC || !INTEL_VSEC
+	depends on INTEL_PLATFORM_DEVICES || !(X86 && ACPI)
		   ^

Did you mean X86_PLATFORM_DEVICES here?

With that, Reviewed-by: Lucas De Marchi <[email protected]>

I see several drivers selecting
X86_PLATFORM_DEVICES though. Maybe they should also be translated to dependencies instead?

$ git grep "select X86_PLATFORM_DEVICES"       drivers/gpu/drm/amd/amdgpu/Kconfig:     select X86_PLATFORM_DEVICES if ACPI && X86
drivers/gpu/drm/gma500/Kconfig: select X86_PLATFORM_DEVICES if ACPI               drivers/gpu/drm/i915/Kconfig:   select X86_PLATFORM_DEVICES if ACPI               drivers/gpu/drm/nouveau/Kconfig:        select X86_PLATFORM_DEVICES if ACPI && X86
drivers/gpu/drm/radeon/Kconfig: select X86_PLATFORM_DEVICES if ACPI && X86        

thanks
Lucas De Marchi

#3 lore raw
Quoted 1 line · Lucas De Marchi

On Fri, May 23, 2025 at 02:10:46PM +0200, Arnd Bergmann wrote:

...

Quoted 4 lines

drivers/gpu/drm/xe/Kconfig line 6

Quoted 1 line
drivers/gpu/drm/xe/Kconfig +1 -0
+	depends on INTEL_PLATFORM_DEVICES || !(X86 && ACPI)
		   ^

Did you mean X86_PLATFORM_DEVICES here?

Why do we need to depend on the whole thingy (yes, it will be enabled at the end) if we only talking about Intel?

Quoted 5 lines

With that, Reviewed-by: Lucas De Marchi <[email protected]>

I see several drivers selecting
X86_PLATFORM_DEVICES though. Maybe they should also be translated to dependencies instead?

I think so, selecting that sounds wrong.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#4 lore raw
Quoted 9 lines · Andy Shevchenko
Quoted 1 line · Lucas De Marchi

On Fri, May 23, 2025 at 02:10:46PM +0200, Arnd Bergmann wrote:

...

Quoted 4 lines

drivers/gpu/drm/xe/Kconfig line 6

Quoted 1 line
drivers/gpu/drm/xe/Kconfig +1 -0
+	depends on INTEL_PLATFORM_DEVICES || !(X86 && ACPI)
		   ^

Did you mean X86_PLATFORM_DEVICES here?

Yes, my mistake.

Quoted 2 lines

Why do we need to depend on the whole thingy (yes, it will be enabled at the end) if we only talking about Intel?

I don't understand what you mean with 'the whole thing'. My change changed the existing 'select X86_PLATFORM_DEVICES if X86 && ACPI' into the corresponding dependency, in order to change it the least.

The dependency itself is needed because of

       select ACPI_WMI if X86 && ACPI

and this in turn is needed for

       select ACPI_VIDEO if X86 && ACPI
Quoted 7 lines
Quoted 5 lines

With that, Reviewed-by: Lucas De Marchi <[email protected]>

I see several drivers selecting
X86_PLATFORM_DEVICES though. Maybe they should also be translated to dependencies instead?

I think so, selecting that sounds wrong.

Agreed. Overall, what I'd really like to see is to remove all those 'select' of drivers from other subsystems. I think ACPI_VIDEO is at the center here, and changing all the 'select ACPI_VIDEO if ACPI' instances to
'depends on ACPI_VIDEO || !ACPI_VIDEO' would solve a lot of the recurring dependency loop problems in drivers/gpu/.

Actually doing it without regressions is going to be nontrivial though, because any change in this area is likely to trigger another dependency loop somewhere.

Arnd

#5 lore raw
Quoted 13 lines · Arnd Bergmann
Quoted 9 lines · Andy Shevchenko
Quoted 1 line · Lucas De Marchi

On Fri, May 23, 2025 at 02:10:46PM +0200, Arnd Bergmann wrote:

...

Quoted 4 lines

drivers/gpu/drm/xe/Kconfig line 6

Quoted 1 line
drivers/gpu/drm/xe/Kconfig +1 -0
+	depends on INTEL_PLATFORM_DEVICES || !(X86 && ACPI)
		   ^

Did you mean X86_PLATFORM_DEVICES here?

Yes, my mistake.

Quoted 7 lines
Quoted 2 lines

Why do we need to depend on the whole thingy (yes, it will be enabled at the end) if we only talking about Intel?

I don't understand what you mean with 'the whole thing'. My change changed the existing 'select X86_PLATFORM_DEVICES if X86 && ACPI' into the corresponding dependency, in order to change it the least.

It used to be (for only one or two releases) X86_PLATFORM_DRIVERS_INTEL, but it doesn't look closer to the mistake above, which I was thinking of. So, Lucas is right.

Quoted 10 lines

The dependency itself is needed because of

       select ACPI_WMI if X86 && ACPI

and this in turn is needed for

       select ACPI_VIDEO if X86 && ACPI
Quoted 2 lines
Quoted 2 lines

With that, Reviewed-by: Lucas De Marchi <[email protected]>

Quoted 8 lines
Quoted 5 lines
Quoted 3 lines

I see several drivers selecting
X86_PLATFORM_DEVICES though. Maybe they should also be translated to dependencies instead?

I think so, selecting that sounds wrong.

Agreed. Overall, what I'd really like to see is to remove all those 'select' of drivers from other subsystems.

Let's start from some low-hanging fruits?

Quoted 9 lines

I think
ACPI_VIDEO is at the center here, and changing all the 'select ACPI_VIDEO if ACPI' instances to
'depends on ACPI_VIDEO || !ACPI_VIDEO' would solve a lot of the recurring dependency loop problems in drivers/gpu/.

Actually doing it without regressions is going to be nontrivial though, because any change in this area is likely to trigger another dependency loop somewhere.

True and I agree this requires more comprehensive testing.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#6 lore raw
In reply to Arnd Bergmann
Quoted 35 lines · Arnd Bergmann
Quoted 9 lines · Andy Shevchenko
Quoted 1 line · Lucas De Marchi

On Fri, May 23, 2025 at 02:10:46PM +0200, Arnd Bergmann wrote:

...

Quoted 4 lines

drivers/gpu/drm/xe/Kconfig line 6

Quoted 1 line
drivers/gpu/drm/xe/Kconfig +1 -0
+	depends on INTEL_PLATFORM_DEVICES || !(X86 && ACPI)
		   ^

Did you mean X86_PLATFORM_DEVICES here?

Yes, my mistake.

Quoted 2 lines

Why do we need to depend on the whole thingy (yes, it will be enabled at the end) if we only talking about Intel?

I don't understand what you mean with 'the whole thing'. My change changed the existing 'select X86_PLATFORM_DEVICES if X86 && ACPI' into the corresponding dependency, in order to change it the least.

The dependency itself is needed because of

       select ACPI_WMI if X86 && ACPI

and this in turn is needed for

       select ACPI_VIDEO if X86 && ACPI
Quoted 6 lines
Quoted 5 lines

With that, Reviewed-by: Lucas De Marchi <[email protected]>

I see several drivers selecting
X86_PLATFORM_DEVICES though. Maybe they should also be translated to dependencies instead?

Quoted 7 lines
Quoted 1 line

I think so, selecting that sounds wrong.

Agreed. Overall, what I'd really like to see is to remove all those 'select' of drivers from other subsystems. I think ACPI_VIDEO is at the center here, and changing all the 'select ACPI_VIDEO if ACPI' instances to
'depends on ACPI_VIDEO || !ACPI_VIDEO' would solve a lot of

Maybe you meant 'depends on ACPI_VIDEO || !ACPI' ?

Quoted 7 lines

the recurring dependency loop problems in drivers/gpu/.

Actually doing it without regressions is going to be nontrivial though, because any change in this area is likely to trigger another dependency loop somewhere.

Arnd

#7 lore raw
Quoted 4 lines · Christopher Snowhill
Quoted 3 lines · Arnd Bergmann
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · Lucas De Marchi

On Fri, May 23, 2025 at 02:10:46PM +0200, Arnd Bergmann wrote:

...

Quoted 5 lines
Quoted 3 lines

I think ACPI_VIDEO is at the center here, and changing all the 'select ACPI_VIDEO if ACPI' instances to
'depends on ACPI_VIDEO || !ACPI_VIDEO' would solve a lot of

Maybe you meant 'depends on ACPI_VIDEO || !ACPI' ?

I believe not. The depends on FOO || FOO=n is idiomatic in Kconfig.

Quoted 5 lines
Quoted 5 lines

the recurring dependency loop problems in drivers/gpu/.

Actually doing it without regressions is going to be nontrivial though, because any change in this area is likely to trigger another dependency loop somewhere.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#8 lore raw
Quoted 11 lines · Andy Shevchenko
Quoted 8 lines · Christopher Snowhill
Quoted 5 lines · Arnd Bergmann
Quoted 1 line · Andy Shevchenko

On Tue, May 27, 2025 at 03:55:46PM -0500, Lucas De Marchi wrote:

I think ACPI_VIDEO is at the center here, and changing all the 'select ACPI_VIDEO if ACPI' instances to
'depends on ACPI_VIDEO || !ACPI_VIDEO' would solve a lot of

Maybe you meant 'depends on ACPI_VIDEO || !ACPI' ?

I believe not. The depends on FOO || FOO=n is idiomatic in Kconfig.

It depends on what we want here. 'ACPI_VIDEO || !ACPI' would be the direct equivalent of the existing 'select ACPI_VIDEO if ACPI', while 'ACPI_VIDEO || !ACPI_VIDEO' would allow building with ACPI=y and ACPI_VIDEO=n, which is not possible today. I'd probably start with the version that Christopher suggested to have a lower regression risk.

Arnd