Re: [PATCH v2 14/17] arm64: dts: realtek: Add EL2 virtual timer interrupt

[Marc Zyngier <maz@kernel.org>]

On Fri, 15 May 2026 00:18:03 +0100,
sashiko-bot@kernel.org wrote:
> 
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [High] Incomplete implementation: missing update for Realtek Kent SoC family which specifies an incorrect EL2 virtual timer interrupt.
> --
> 
> commit 3b73638950a28ecc324e1c28fa4fb9fcb13d5567
> Author: Marc Zyngier <maz@kernel.org>
> 
> arm64: dts: realtek: Add EL2 virtual timer interrupt
> 
> This commit adds the missing EL2 virtual timer interrupt to ARMv8.2-based
> Realtek SoCs. It updates rtd16xx.dtsi by adding the fifth timer interrupt
> with GIC_PPI 12.
> 
> > diff --git a/arch/arm64/boot/dts/realtek/rtd16xx.dtsi b/arch/arm64/boot/dts/realtek/rtd16xx.dtsi
> > index 3a7f6e35b7f74..43b13d133c324 100644
> > --- a/arch/arm64/boot/dts/realtek/rtd16xx.dtsi
> > +++ b/arch/arm64/boot/dts/realtek/rtd16xx.dtsi
> > @@ -105,7 +105,8 @@ timer {
> >  		interrupts = <GIC_PPI 13 IRQ_TYPE_LEVEL_LOW>,
> >  			     <GIC_PPI 14 IRQ_TYPE_LEVEL_LOW>,
> >  			     <GIC_PPI 11 IRQ_TYPE_LEVEL_LOW>,
> > -			     <GIC_PPI 10 IRQ_TYPE_LEVEL_LOW>;
> > +			     <GIC_PPI 10 IRQ_TYPE_LEVEL_LOW>,
> > +			     <GIC_PPI 12 IRQ_TYPE_LEVEL_LOW>;
> >  	};
> 
> Does this update miss another Realtek ARMv8.2+ SoC file?
> 
> Looking at arch/arm64/boot/dts/realtek/kent.dtsi (used by RTD1501, RTD1861,
> and RTD1920), the fifth timer interrupt is currently defined as
> <GIC_PPI 9 IRQ_TYPE_LEVEL_HIGH>.
> 
> PPI 9 is the GIC vCPU maintenance interrupt, rather than the EL2 virtual
> timer interrupt which should be PPI 12:
> 
> arch/arm64/boot/dts/realtek/kent.dtsi:
> 	timer {
> 		compatible = "arm,armv8-timer";
> 		interrupts = <GIC_PPI 13 IRQ_TYPE_LEVEL_LOW>,
> 			     <GIC_PPI 14 IRQ_TYPE_LEVEL_LOW>,
> 			     <GIC_PPI 11 IRQ_TYPE_LEVEL_LOW>,
> 			     <GIC_PPI 10 IRQ_TYPE_LEVEL_LOW>,
> 			     <GIC_PPI  9 IRQ_TYPE_LEVEL_HIGH>;
> 	};
> 
> Another patch in this series (clocksource/drivers/arm_arch_timer: Default
> to EL2 virtual timer when running VHE) changes the driver to use the EL2
> virtual timer by default on VHE systems.
> 
> If the kernel uses the EL2 virtual timer on these Realtek SoCs, could it
> request the wrong interrupt (PPI 9 instead of PPI 12), causing timer
> interrupts to not be delivered and resulting in a boot hang regression?

That's definitely odd, as PPI9 is present in two interrupt specifiers
(timer and GIC). Obviously, that's not possible.

I'll add a tentative fix for that in the next round. Maybe Andreas or
Yu-Chun can check in the meantime?

	M.

-- 
Jazz isn't dead. It just smells funny.