Re: [PATCH v2 1/8] drivers/virt: Drop VIRT_DRIVERS build dependency

[<dan.j.williams@intel.com>]

Nathan Chancellor wrote:
> Hi Dan,
> 
> On Wed, Nov 12, 2025 at 06:14:39PM -0800, Dan Williams wrote:
> > All of the objects in drivers/virt/ have their own configuration symbols to
> > gate compilation. I.e. nothing gets added to the kernel with
> > CONFIG_VIRT_DRIVERS=y in isolation.
> > 
> > Unconditionally descend into drivers/virt/ so that consumers do not need to
> > add an additional CONFIG_VIRT_DRIVERS dependency.
> > 
> > Fix warnings of the form:
> > 
> >     Kconfig warnings: (for reference only)
> >        WARNING: unmet direct dependencies detected for TSM
> >        Depends on [n]: VIRT_DRIVERS [=n]
> >        Selected by [y]:
> >        - PCI_TSM [=y] && PCI [=y]
> 
> I can still trigger this warning on next-20251202, which contains this
> change as commit 110c155e8a68 ("drivers/virt: Drop VIRT_DRIVERS build
> dependency").
> 
>   $ git merge-base --is-ancestor 110c155e8a684d8b2423a72cfde147903881f765 HEAD
>     echo $status
>   0
> 
>   $ make -skj"$(nproc)" ARCH=arm64 CROSS_COMPILE=aarch64-linux- clean defconfig
> 
>   $ scripts/config -e PCI_TSM
> 
>   $ make -skj"$(nproc)" ARCH=arm64 CROSS_COMPILE=aarch64-linux- olddefconfig
>   WARNING: unmet direct dependencies detected for TSM
>     Depends on [n]: VIRT_DRIVERS [=n]
>     Selected by [y]:
>     - PCI_TSM [=y] && PCI [=y]
> 
> I would think you would need something like this avoid the problem but I
> am not sure if it would be acceptable, hence just the report.

Oh, yes, thanks for that. It turns out I fixed the compilation problem
associated with this report [1].

   ld: drivers/pci/ide.o: in function `pci_ide_stream_release':
   ide.c:(.text+0xffb): undefined reference to `tsm_ide_stream_unregister'
   ld: drivers/pci/tsm.o: in function `connect_store':
   tsm.c:(.text+0x1d42): undefined reference to `find_tsm_dev'

I missed that the report I ended up quoting [2] was only about the
Kconfig problem not the build error.

[1]: http://lore.kernel.org/202511140712.NubhamPy-lkp@intel.com
[2]: http://lore.kernel.org/202511041832.ylcgIiqN-lkp@intel.com

> Cheers,
> Nathan
> 
> diff --git a/drivers/virt/Kconfig b/drivers/virt/Kconfig
> index d8c848cf09a6..52eb7e4ba71f 100644
> --- a/drivers/virt/Kconfig
> +++ b/drivers/virt/Kconfig
> @@ -47,6 +47,6 @@ source "drivers/virt/nitro_enclaves/Kconfig"
>  
>  source "drivers/virt/acrn/Kconfig"
>  
> -source "drivers/virt/coco/Kconfig"
> -
>  endif
> +
> +source "drivers/virt/coco/Kconfig"
> diff --git a/drivers/virt/coco/Kconfig b/drivers/virt/coco/Kconfig
> index bb0c6d6ddcc8..df1cfaf26c65 100644
> --- a/drivers/virt/coco/Kconfig
> +++ b/drivers/virt/coco/Kconfig
> @@ -3,6 +3,7 @@
>  # Confidential computing related collateral
>  #
>  
> +if VIRT_DRIVERS
>  source "drivers/virt/coco/efi_secret/Kconfig"
>  
>  source "drivers/virt/coco/pkvm-guest/Kconfig"
> @@ -14,6 +15,7 @@ source "drivers/virt/coco/tdx-guest/Kconfig"
>  source "drivers/virt/coco/arm-cca-guest/Kconfig"
>  
>  source "drivers/virt/coco/guest/Kconfig"
> +endif
>  
>  config TSM
>  	bool

Yes, I think that looks right to me as core infrastructure pieces are
not user selectable like the leaf drivers.

For now I will note the mixup in the pull request and would be happy to
take an incremental patch fixing this up.