Re: [PATCH 1/2] memory: tegra: Add missing dependencies

[Thierry Reding <thierry.reding@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 3188 bytes --]

On Wed, Jun 09, 2021 at 03:19:12PM +0200, Krzysztof Kozlowski wrote:
> On 09/06/2021 13:58, Dmitry Osipenko wrote:
> > 09.06.2021 14:28, Thierry Reding пишет:
> >> From: Thierry Reding <treding@nvidia.com>
> >>
> >> When enabling the COMPILE_TEST Kconfig option, the Tegra memory
> >> controller can be built without ARCH_TEGRA being selected. However, the
> >> driver implicitly depends on some symbols pulled in via ARCH_TEGRA,
> >> which causes the build to break.
> >>
> >> Add explicit dependencies for OF_EARLY_FLATTREE and OF_RESERVED_MEM to
> >> the Tegra MC Kconfig option to make sure they are selected even if
> >> ARCH_TEGRA is not.
> >>
> >> Reported-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
> >> Signed-off-by: Thierry Reding <treding@nvidia.com>
> >> ---
> >>  drivers/memory/tegra/Kconfig | 2 ++
> >>  1 file changed, 2 insertions(+)
> >>
> >> diff --git a/drivers/memory/tegra/Kconfig b/drivers/memory/tegra/Kconfig
> >> index f9bae36c03a3..ecfb071fc4f4 100644
> >> --- a/drivers/memory/tegra/Kconfig
> >> +++ b/drivers/memory/tegra/Kconfig
> >> @@ -48,6 +48,8 @@ config TEGRA124_EMC
> >>  config TEGRA210_EMC_TABLE
> >>  	bool
> >>  	depends on ARCH_TEGRA_210_SOC || COMPILE_TEST
> >> +	select OF_EARLY_FLATTREE
> >> +	select OF_RESERVED_MEM
> >>  
> >>  config TEGRA210_EMC
> >>  	tristate "NVIDIA Tegra210 External Memory Controller driver"
> >>
> > 
> > Will this work if CONFIG_OF is disabled?
> 
> Yeah, good question. That's why I propose "depends on". No issues with
> unmet or circular dependencies.

I couldn't find a way to make this work with "depends on" because
OF_RESERVED_MEM is not user-visible and the only way to get it enabled
is if something also selects OF_EARLY_FLATTREE, which is only ever done
at the architecture Kconfig level (and for OF unit testing).

So switching this to a "depends on" causes the TEGRA210_EMC never to get
enabled.

However, with OF disabled, the above causes issues because it can lead
to unmet direct dependencies. That, in turn, can be fixed by appending
&& OF to the COMPILE_TEST branch, which seems like a good enough
compromise.

Here's what I have on top of the above patch and that seems to do the
trick.

--- >8 ---
diff --git a/drivers/memory/tegra/Kconfig b/drivers/memory/tegra/Kconfig
index ecfb071fc4f4..1c553895160c 100644
--- a/drivers/memory/tegra/Kconfig
+++ b/drivers/memory/tegra/Kconfig
@@ -47,13 +47,13 @@ config TEGRA124_EMC
 
 config TEGRA210_EMC_TABLE
 	bool
-	depends on ARCH_TEGRA_210_SOC || COMPILE_TEST
+	depends on ARCH_TEGRA_210_SOC || (COMPILE_TEST && OF)
 	select OF_EARLY_FLATTREE
 	select OF_RESERVED_MEM
 
 config TEGRA210_EMC
 	tristate "NVIDIA Tegra210 External Memory Controller driver"
-	depends on ARCH_TEGRA_210_SOC || COMPILE_TEST
+	depends on ARCH_TEGRA_210_SOC || (COMPILE_TEST && OF)
 	select TEGRA210_EMC_TABLE
 	help
 	  This driver is for the External Memory Controller (EMC) found on
--- >8 ---

So in a nutshell this will only get compile-tested if OF is enabled, but
then it will select OF_RESERVED_MEM and OF_EARLY_FLATTREE to get the
required symbols.

Thierry

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]