[PATCH] perf: ARM_SMMU_V3_PMU should depend on ARM_SMMU_V3

[PATCH] perf: ARM_SMMU_V3_PMU should depend on ARM_SMMU_V3

[Geert Uytterhoeven]

There is no point in monitoring transactions passing through the SMMU
when ARM Ltd. System MMU Version 3 (SMMUv3) Support is disabled.
Hence replace the dependency on ARM64 by a dependency on ARM_SMMU_V3
(which implies the former).

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
This caught my eye after commit 7c3f204e544dfa37 ("perf/smmuv3: Remove
build dependency on ACPI") in arm64/for-next/core.

Perhaps my understanding is wrong? Is there anything to monitor when
ARM_SMMU_V3=n?

 drivers/perf/Kconfig | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/perf/Kconfig b/drivers/perf/Kconfig
index 273d67ecf6d2530f..c33324692fbf0141 100644
--- a/drivers/perf/Kconfig
+++ b/drivers/perf/Kconfig
@@ -92,7 +92,7 @@ config ARM_PMU_ACPI
 
 config ARM_SMMU_V3_PMU
 	 tristate "ARM SMMUv3 Performance Monitors Extension"
-	 depends on ARM64 || (COMPILE_TEST && 64BIT)
+	 depends on ARM_SMMU_V3 || (COMPILE_TEST && 64BIT)
 	 depends on GENERIC_MSI_IRQ
 	   help
 	   Provides support for the ARM SMMUv3 Performance Monitor Counter
-- 
2.34.1

Re: [PATCH] perf: ARM_SMMU_V3_PMU should depend on ARM_SMMU_V3

[Robin Murphy]

On 2023-08-15 16:36, Geert Uytterhoeven wrote:
> There is no point in monitoring transactions passing through the SMMU
> when ARM Ltd. System MMU Version 3 (SMMUv3) Support is disabled.
> Hence replace the dependency on ARM64 by a dependency on ARM_SMMU_V3
> (which implies the former).
> 
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
> This caught my eye after commit 7c3f204e544dfa37 ("perf/smmuv3: Remove
> build dependency on ACPI") in arm64/for-next/core.
> 
> Perhaps my understanding is wrong? Is there anything to monitor when
> ARM_SMMU_V3=n?

Yes, at least TBU event 1 still counts bypass transactions even when the 
SMMU is disabled, so PMCGs can be useful as basic traffic monitors in 
their own right. Plus the original design intent was that PMCGs may also 
be implemented by other things that interact with the SMMU, like 
ATS-capable PCIe root complexes, or devices with their own internal 
TLBs, thus they could potentially count any manner of 
implementation-defined events that aren't necessarily related to SMMU 
translation.

Thanks,
Robin.

>   drivers/perf/Kconfig | 2 +-
>   1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/perf/Kconfig b/drivers/perf/Kconfig
> index 273d67ecf6d2530f..c33324692fbf0141 100644
> --- a/drivers/perf/Kconfig
> +++ b/drivers/perf/Kconfig
> @@ -92,7 +92,7 @@ config ARM_PMU_ACPI
>   
>   config ARM_SMMU_V3_PMU
>   	 tristate "ARM SMMUv3 Performance Monitors Extension"
> -	 depends on ARM64 || (COMPILE_TEST && 64BIT)
> +	 depends on ARM_SMMU_V3 || (COMPILE_TEST && 64BIT)
>   	 depends on GENERIC_MSI_IRQ
>   	   help
>   	   Provides support for the ARM SMMUv3 Performance Monitor Counter

Re: [PATCH] perf: ARM_SMMU_V3_PMU should depend on ARM_SMMU_V3

[Geert Uytterhoeven]

Hi Robin,

On Wed, Aug 16, 2023 at 1:23 PM Robin Murphy <robin.murphy@arm.com> wrote:
> On 2023-08-15 16:36, Geert Uytterhoeven wrote:
> > There is no point in monitoring transactions passing through the SMMU
> > when ARM Ltd. System MMU Version 3 (SMMUv3) Support is disabled.
> > Hence replace the dependency on ARM64 by a dependency on ARM_SMMU_V3
> > (which implies the former).
> >
> > Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> > ---
> > This caught my eye after commit 7c3f204e544dfa37 ("perf/smmuv3: Remove
> > build dependency on ACPI") in arm64/for-next/core.
> >
> > Perhaps my understanding is wrong? Is there anything to monitor when
> > ARM_SMMU_V3=n?
>
> Yes, at least TBU event 1 still counts bypass transactions even when the
> SMMU is disabled, so PMCGs can be useful as basic traffic monitors in
> their own right. Plus the original design intent was that PMCGs may also
> be implemented by other things that interact with the SMMU, like
> ATS-capable PCIe root complexes, or devices with their own internal
> TLBs, thus they could potentially count any manner of
> implementation-defined events that aren't necessarily related to SMMU
> translation.

Thanks for the explanation!
Hereby I withdraw my patch.

Gr{oetje,eeting}s,

                        Geert

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