[RFC][PATCH] iommu: Kconfig: Fix missing config option for CONFIG_ARM_SMMU_QCOM

[RFC][PATCH] iommu: Kconfig: Fix missing config option for CONFIG_ARM_SMMU_QCOM

[John Stultz]

In commit 424953cf3c665 ("qcom_scm: hide Kconfig symbol"), some
logic was added to make the code depend on CONFIG_ARM_SMMU_QCOM,
however no such option exist.

This caused regressions on db845c (and I suspect other qcom
platforms), but can be easily fixed up by adding a new option
to re-enable the newly conditionalized code.

I set it to default to ARCH_QCOM, so folks don't have to
discover this new option to keep things booting, but if folks
would rather, I'm ok to drop that line.

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Bjorn Andersson <bjorn.andersson@linaro.org>
Cc: Amit Pundir <amit.pundir@linaro.org>
Cc: Caleb Connolly <caleb.connolly@linaro.org>
Cc: Kalle Valo <kvalo@codeaurora.org>
Cc: linux-arm-kernel <linux-arm-kernel@lists.infradead.org>
Cc: linux-arm-msm <linux-arm-msm@vger.kernel.org>
Fixes: 424953cf3c665 ("qcom_scm: hide Kconfig symbol")
Signed-off-by: John Stultz <john.stultz@linaro.org>
---
 drivers/iommu/Kconfig | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/drivers/iommu/Kconfig b/drivers/iommu/Kconfig
index c5c71b7ab7e83..e674796b1174d 100644
--- a/drivers/iommu/Kconfig
+++ b/drivers/iommu/Kconfig
@@ -380,6 +380,16 @@ config ARM_SMMU_V3_SVA
 	  Say Y here if your system supports SVA extensions such as PCIe PASID
 	  and PRI.
 
+config ARM_SMMU_QCOM
+	bool "QCOM Specific ARM SMMU implementation support"
+	depends on ARM_SMMU
+	default ARCH_QCOM
+	help
+          Support for the QCOM implementation of the ARM System MMU architecture.
+
+          Say Y here if your SoC includes an QCOM IOMMU device implementing
+          the ARM SMMU architecture.
+
 config S390_IOMMU
 	def_bool y if S390 && PCI
 	depends on S390 && PCI
-- 
2.25.1

Re: [RFC][PATCH] iommu: Kconfig: Fix missing config option for CONFIG_ARM_SMMU_QCOM

[Bjorn Andersson]

On Mon 11 Oct 22:26 CDT 2021, John Stultz wrote:

> In commit 424953cf3c665 ("qcom_scm: hide Kconfig symbol"), some
> logic was added to make the code depend on CONFIG_ARM_SMMU_QCOM,
> however no such option exist.
> 
> This caused regressions on db845c (and I suspect other qcom
> platforms), but can be easily fixed up by adding a new option
> to re-enable the newly conditionalized code.
> 
> I set it to default to ARCH_QCOM, so folks don't have to
> discover this new option to keep things booting, but if folks
> would rather, I'm ok to drop that line.
> 

I've not looked where the discussion went, but Dmitry posted this end of
last week:

https://lore.kernel.org/linux-arm-msm/20211010023350.978638-1-dmitry.baryshkov@linaro.org/

Regards,
Bjorn

> Cc: Arnd Bergmann <arnd@arndb.de>
> Cc: Bjorn Andersson <bjorn.andersson@linaro.org>
> Cc: Amit Pundir <amit.pundir@linaro.org>
> Cc: Caleb Connolly <caleb.connolly@linaro.org>
> Cc: Kalle Valo <kvalo@codeaurora.org>
> Cc: linux-arm-kernel <linux-arm-kernel@lists.infradead.org>
> Cc: linux-arm-msm <linux-arm-msm@vger.kernel.org>
> Fixes: 424953cf3c665 ("qcom_scm: hide Kconfig symbol")
> Signed-off-by: John Stultz <john.stultz@linaro.org>
> ---
>  drivers/iommu/Kconfig | 10 ++++++++++
>  1 file changed, 10 insertions(+)
> 
> diff --git a/drivers/iommu/Kconfig b/drivers/iommu/Kconfig
> index c5c71b7ab7e83..e674796b1174d 100644
> --- a/drivers/iommu/Kconfig
> +++ b/drivers/iommu/Kconfig
> @@ -380,6 +380,16 @@ config ARM_SMMU_V3_SVA
>  	  Say Y here if your system supports SVA extensions such as PCIe PASID
>  	  and PRI.
>  
> +config ARM_SMMU_QCOM
> +	bool "QCOM Specific ARM SMMU implementation support"
> +	depends on ARM_SMMU
> +	default ARCH_QCOM
> +	help
> +          Support for the QCOM implementation of the ARM System MMU architecture.
> +
> +          Say Y here if your SoC includes an QCOM IOMMU device implementing
> +          the ARM SMMU architecture.
> +
>  config S390_IOMMU
>  	def_bool y if S390 && PCI
>  	depends on S390 && PCI
> -- 
> 2.25.1
> 

Re: [RFC][PATCH] iommu: Kconfig: Fix missing config option for CONFIG_ARM_SMMU_QCOM

[John Stultz]

On Mon, Oct 11, 2021 at 8:37 PM Bjorn Andersson
<bjorn.andersson@linaro.org> wrote:
>
> On Mon 11 Oct 22:26 CDT 2021, John Stultz wrote:
>
> > In commit 424953cf3c665 ("qcom_scm: hide Kconfig symbol"), some
> > logic was added to make the code depend on CONFIG_ARM_SMMU_QCOM,
> > however no such option exist.
> >
> > This caused regressions on db845c (and I suspect other qcom
> > platforms), but can be easily fixed up by adding a new option
> > to re-enable the newly conditionalized code.
> >
> > I set it to default to ARCH_QCOM, so folks don't have to
> > discover this new option to keep things booting, but if folks
> > would rather, I'm ok to drop that line.
> >
>
> I've not looked where the discussion went, but Dmitry posted this end of
> last week:
>
> https://lore.kernel.org/linux-arm-msm/20211010023350.978638-1-dmitry.baryshkov@linaro.org/

Oh, my apologies for not seeing this! I did search my mail log
earlier, but I must have missed it.

So yes, I validated earlier that reverting the problematic chunks
works as well, and I see Arnd did already have a similar patch to what
I sent out.

I'd probably lean towards Arnd's patch, as he has been uncovering the
problematic Kconfig edge cases.

thanks
-john