[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ZtGCJincjO7z-WFm@linaro.org>
Date: Fri, 30 Aug 2024 10:26:14 +0200
From: Stephan Gerhold <stephan.gerhold@...aro.org>
To: Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
Cc: Danila Tikhonov <danila@...xyga.com>, andersson@...nel.org,
konradybcio@...nel.org, robh@...nel.org, krzk+dt@...nel.org,
conor+dt@...nel.org, cros-qcom-dts-watchers@...omium.org,
linux-arm-msm@...r.kernel.org, devicetree@...r.kernel.org,
linux-kernel@...r.kernel.org, linux@...nlining.org
Subject: Re: [PATCH] arm64: dts: qcom: sc7280.dtsi: Fix PMU nodes for Cortex
A55 and A78
On Fri, Aug 30, 2024 at 11:13:28AM +0300, Dmitry Baryshkov wrote:
> On Sun, Aug 18, 2024 at 10:29:05PM GMT, Danila Tikhonov wrote:
> > The SC7280, SM7325, and QCM6490 platforms feature an 8-core setup
> > consisting of:
> > - 1x Kryo 670 Prime (Cortex-A78) / Kryo 670 Gold Plus (Cortex-A78)
> > - 3x Kryo 670 Gold (Cortex-A78)
> > - 4x Kryo 670 Silver (Cortex-A55)
> > (The CPU cores in the SC7280 are simply called Kryo, but are
> > nevertheless based on the same Cortex A78 and A55).
> >
> > Use the correct compatibility.
> >
> > Signed-off-by: Danila Tikhonov <danila@...xyga.com>
> > ---
> > arch/arm64/boot/dts/qcom/sc7280.dtsi | 9 +++++++--
> > 1 file changed, 7 insertions(+), 2 deletions(-)
> >
> > diff --git a/arch/arm64/boot/dts/qcom/sc7280.dtsi b/arch/arm64/boot/dts/qcom/sc7280.dtsi
> > index 91cc5e74d8f5..ab024a3c3653 100644
> > --- a/arch/arm64/boot/dts/qcom/sc7280.dtsi
> > +++ b/arch/arm64/boot/dts/qcom/sc7280.dtsi
> > @@ -845,8 +845,13 @@ wlan_smp2p_in: wlan-wpss-to-ap {
> > };
> > };
> >
> > - pmu {
> > - compatible = "arm,armv8-pmuv3";
> > + pmu-a55 {
> > + compatible = "arm,cortex-a55-pmu";
> > + interrupts = <GIC_PPI 7 IRQ_TYPE_LEVEL_LOW>;
> > + };
> > +
> > + pmu-a78 {
> > + compatible = "arm,cortex-a78-pmu";
> > interrupts = <GIC_PPI 7 IRQ_TYPE_LEVEL_LOW>;
>
> Shouldn't these two entries have GIC_CPU_MASK_RAW(), limiting interrupts
> to the corresponding cores? I see that in [1] Rob used masks for older
> SoCs, but skipped them for newer ones.
>
> [1] https://lore.kernel.org/all/20240417204247.3216703-1-robh@kernel.org/
>
According to the bindings, the cpu mask bits apply only to gic-v2, but
not gic-v3. It looks correct in Rob's changes. And here too, since
SC7280 uses gic-v3.
Thanks,
Stephan
Powered by blists - more mailing lists