Re: [PATCH] genpt: Make GENERIC_PT invisible

[Geert Uytterhoeven]

Hi Jason,

CC kunit

On Thu, 20 Nov 2025 at 18:07, Jason Gunthorpe <j...@ziepe.ca> wrote:
> On Thu, Nov 20, 2025 at 12:49:33PM -0400, Jason Gunthorpe wrote:
> > On Wed, Nov 12, 2025 at 03:08:05PM +0100, Geert Uytterhoeven wrote:
> > > There is no point in asking the user about the Generic Radix Page
> > > Table API:
> > > - All IOMMU drivers that use this API already select GENERIC_PT when
> > > needed,
> > > - Most users probably do not know what to answer anyway.
> > >
> > > Fixes: 7c5b184db7145fd4 ("genpt: Generic Page Table base API")
> > > Signed-off-by: Geert Uytterhoeven <geert+...@glider.be>
> > > ---
> > > drivers/iommu/generic_pt/Kconfig | 2 +-
> > > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > Reviewed-by: Jason Gunthorpe <j...@nvidia.com>
>
> Actually, it doesn't work :\
>
> $ tools/testing/kunit/kunit.py run --build_dir build_kunit_x86_64 --arch x86_64 --kunitconfig ./drivers/iommu/generic_pt/.kunitconfig
> [13:01:26] Configuring KUnit Kernel ...
> [13:01:26] Building KUnit Kernel ...
> Populating config with:
> $ make ARCH=x86_64 O=build_kunit_x86_64 olddefconfig
> Building with:
> $ make all compile_commands.json scripts_gdb ARCH=x86_64 O=build_kunit_x86_64 --jobs=20
> ERROR:root:Not all Kconfig options selected in kunitconfig were in the generated .config.
> This is probably due to unsatisfied dependencies.
> Missing: CONFIG_IOMMUFD_TEST=y, CONFIG_DEBUG_GENERIC_PT=y, CONFIG_IOMMU_PT_VTDSS=y, CONFIG_IOMMU_PT=y, CONFIG_IOMMU_PT_AMDV1=y, CONFIG_IOMMU_PT_X86_64=y, CONFIG_GENERIC_PT=y, CONFIG_IOMMU_PT_KUNIT_TEST=y
>
> Can you add this hunk and send a v2?
>
> --- a/drivers/iommu/generic_pt/.kunitconfig
> +++ b/drivers/iommu/generic_pt/.kunitconfig
> @@ -1,4 +1,5 @@
> CONFIG_KUNIT=y
> +CONFIG_COMPILE_TEST=y
> CONFIG_GENERIC_PT=y
> CONFIG_DEBUG_GENERIC_PT=y
> CONFIG_IOMMU_PT=y

Do you really want to enable CONFIG_COMPILE_TEST in a .kunitconfig?

Hm, that .kunitconfig already enables IOMMUFD_TEST, which is
documented to be dangerous (why?), and already enabled by allyesconfig
(except on GENERIC_ATOMIC64 architectures).
IOMMUFD_TEST cannot select GENERIC_PT, as that would lead to
a recursive dependency (and I am not a huge fan of test code auto-enabling
extra attack surfaces^W^W functionality).

Or perhaps:

- bool "Generic Radix Page Table"
+ bool "Generic Radix Page Table" if COMPILE_TEST || KUNIT

?

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@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

[Jason Gunthorpe]

IDK, why not?

> Hm, that .kunitconfig already enables IOMMUFD_TEST, which is
> documented to be dangerous (why?)

It builds in a kernel module with a uapi that is kind of unsafe.

Though, hmm, maybe that is some weird a leftover I don't recall that
this kunit needed IOMMUFD_TEST stanza at all..

> and already enabled by allyesconfig (except on GENERIC_ATOMIC64
> architectures).

I guess allyesconfig would do that.

> IOMMUFD_TEST cannot select GENERIC_PT, as that would lead to a
> recursive dependency (and I am not a huge fan of test code
> auto-enabling extra attack surfaces^W^W functionality).

Yes

> Or perhaps:
>
> - bool "Generic Radix Page Table"
> + bool "Generic Radix Page Table" if COMPILE_TEST || KUNIT
>
> ?

It would work, that does seem like a better choice if someone wants to
make the kunit run in a normal disto kernel.

Thanks,
Jason