[PATCH 1/2] ARM: dts: rockchip: Limit GPU frequency on veyron mickey to 300 MHz when the CPU gets very hot

[PATCH 1/2] ARM: dts: rockchip: Limit GPU frequency on veyron mickey to 300 MHz when the CPU gets very hot

[Matthias Kaehlcke]

On rk3288 the CPU and GPU temperatures are correlated. Limit the GPU
frequency on veyron mickey to 300 MHz for CPU temperatures >= 85°C.

This matches the configuration of the downstream Chrome OS 3.14 kernel,
the 'official' kernel for mickey.

Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
---
Note: this patch depends on "ARM: dts: rockchip: Add #cooling-cells
entry for rk3288 GPU" (https://lore.kernel.org/patchwork/patch/1075005/)
---
 arch/arm/boot/dts/rk3288-veyron-mickey.dts | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/arch/arm/boot/dts/rk3288-veyron-mickey.dts b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
index d889ab3c8235..f118d92a49d0 100644
--- a/arch/arm/boot/dts/rk3288-veyron-mickey.dts
+++ b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
@@ -125,6 +125,12 @@
 					 <&cpu2 8 THERMAL_NO_LIMIT>,
 					 <&cpu3 8 THERMAL_NO_LIMIT>;
 		};
+
+		/* At very hot, don't let GPU go over 300 MHz */
+		cpu_very_hot_limit_gpu {
+			trip = <&cpu_alert_very_hot>;
+			cooling-device = <&gpu 2 2>;
+		};
 	};
 };
 
-- 
2.21.0.1020.gf2820cf01a-goog

[PATCH 2/2] ARM: dts: rockchip: Configure the GPU thermal zone for mickey

[Matthias Kaehlcke]

mickey crams a lot of hardware into a tiny package, which requires
more aggressive thermal throttling than for devices with a larger
footprint. Configure the GPU thermal zone to throttle the GPU
progressively at temperatures >= 60°C. Heat dissipated by the
CPUs also affects the GPU temperature, hence we cap the CPU
frequency to 1.4 GHz for temperatures above 65°C. Further throttling
of the CPUs may be performed by the CPU thermal zone.

The configuration matches that of the downstram Chrome OS 3.14
kernel, the 'official' kernel for mickey.

Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
---
Note: this patch depends on "ARM: dts: rockchip: Add #cooling-cells
entry for rk3288 GPU" (https://lore.kernel.org/patchwork/patch/1075005/)
---
 arch/arm/boot/dts/rk3288-veyron-mickey.dts | 64 ++++++++++++++++++++++
 1 file changed, 64 insertions(+)

diff --git a/arch/arm/boot/dts/rk3288-veyron-mickey.dts b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
index f118d92a49d0..f0b83afa2a60 100644
--- a/arch/arm/boot/dts/rk3288-veyron-mickey.dts
+++ b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
@@ -138,6 +138,70 @@
 	/delete-property/mmc-hs200-1_8v;
 };
 
+&gpu_thermal {
+	/delete-node/ trips;
+	/delete-node/ cooling-maps;
+
+	trips {
+		gpu_alert_warmish: gpu_alert_warmish {
+			temperature = <60000>; /* millicelsius */
+			hysteresis = <2000>; /* millicelsius */
+			type = "passive";
+		};
+		gpu_alert_warm: gpu_alert_warm {
+			temperature = <65000>; /* millicelsius */
+			hysteresis = <2000>; /* millicelsius */
+			type = "passive";
+		};
+		gpu_alert_hotter: gpu_alert_hotter {
+			temperature = <84000>; /* millicelsius */
+			hysteresis = <2000>; /* millicelsius */
+			type = "passive";
+		};
+		gpu_alert_very_very_hot: gpu_alert_very_very_hot {
+			temperature = <86000>; /* millicelsius */
+			hysteresis = <2000>; /* millicelsius */
+			type = "passive";
+		};
+		gpu_crit: gpu_crit {
+			temperature = <90000>; /* millicelsius */
+			hysteresis = <2000>; /* millicelsius */
+			type = "critical";
+		};
+	};
+
+	cooling-maps {
+		/* After 1st level throttle the GPU down to as low as 400 MHz */
+		gpu_warmish_limit_gpu {
+			trip = <&gpu_alert_warmish>;
+			cooling-device = <&gpu THERMAL_NO_LIMIT 1>;
+		};
+
+		/*
+		 * Slightly after we throttle the GPU, we'll also make sure that
+		 * the CPU can't go faster than 1.4 GHz.  Note that we won't
+		 * throttle the CPU lower than 1.4 GHz due to GPU heat--we'll
+		 * let the CPU do the rest itself.
+		 */
+		gpu_warm_limit_cpu {
+			trip = <&gpu_alert_warm>;
+			cooling-device = <&cpu0 4 4>;
+		};
+
+		/* When hot, GPU goes down to 300 MHz */
+		gpu_hotter_limit_gpu {
+			trip = <&gpu_alert_hotter>;
+			cooling-device = <&gpu 2 2>;
+		};
+
+		/* When really hot, don't let GPU go _above_ 300 MHz */
+		gpu_very_very_hot_limit_gpu {
+			trip = <&gpu_alert_very_very_hot>;
+			cooling-device = <&gpu 2 THERMAL_NO_LIMIT>;
+		};
+	};
+};
+
 &i2c2 {
 	status = "disabled";
 };
-- 
2.21.0.1020.gf2820cf01a-goog

Re: [PATCH 2/2] ARM: dts: rockchip: Configure the GPU thermal zone for mickey

[Doug Anderson]

Hi,

On Mon, May 20, 2019 at 10:01 AM Matthias Kaehlcke <mka@chromium.org> wrote:
>
> mickey crams a lot of hardware into a tiny package, which requires
> more aggressive thermal throttling than for devices with a larger
> footprint. Configure the GPU thermal zone to throttle the GPU
> progressively at temperatures >= 60°C. Heat dissipated by the
> CPUs also affects the GPU temperature, hence we cap the CPU
> frequency to 1.4 GHz for temperatures above 65°C. Further throttling
> of the CPUs may be performed by the CPU thermal zone.
>
> The configuration matches that of the downstram Chrome OS 3.14

s/downstram/downstream


> +       cooling-maps {
> +               /* After 1st level throttle the GPU down to as low as 400 MHz */
> +               gpu_warmish_limit_gpu {
> +                       trip = <&gpu_alert_warmish>;
> +                       cooling-device = <&gpu THERMAL_NO_LIMIT 1>;

As per my comment in patch #1, you are probably ending up throttling
to 500 MHz, not 400 MHz.  Below will all have similar problems unless
we actually delete the 500 MHz operating point.


> +               };
> +
> +               /*
> +                * Slightly after we throttle the GPU, we'll also make sure that
> +                * the CPU can't go faster than 1.4 GHz.  Note that we won't
> +                * throttle the CPU lower than 1.4 GHz due to GPU heat--we'll
> +                * let the CPU do the rest itself.
> +                */
> +               gpu_warm_limit_cpu {
> +                       trip = <&gpu_alert_warm>;
> +                       cooling-device = <&cpu0 4 4>;

Shouldn't you list cpu1, cpu2, and cpu3 too?  That'd match what
upstream did elsewhere in this file?

Re: [PATCH 2/2] ARM: dts: rockchip: Configure the GPU thermal zone for mickey

[Matthias Kaehlcke]

On Mon, May 20, 2019 at 01:21:33PM -0700, Doug Anderson wrote:
> Hi,
> 
> On Mon, May 20, 2019 at 10:01 AM Matthias Kaehlcke <mka@chromium.org> wrote:
> >
> > mickey crams a lot of hardware into a tiny package, which requires
> > more aggressive thermal throttling than for devices with a larger
> > footprint. Configure the GPU thermal zone to throttle the GPU
> > progressively at temperatures >= 60°C. Heat dissipated by the
> > CPUs also affects the GPU temperature, hence we cap the CPU
> > frequency to 1.4 GHz for temperatures above 65°C. Further throttling
> > of the CPUs may be performed by the CPU thermal zone.
> >
> > The configuration matches that of the downstram Chrome OS 3.14
> 
> s/downstram/downstream

ack

> 
> > +       cooling-maps {
> > +               /* After 1st level throttle the GPU down to as low as 400 MHz */
> > +               gpu_warmish_limit_gpu {
> > +                       trip = <&gpu_alert_warmish>;
> > +                       cooling-device = <&gpu THERMAL_NO_LIMIT 1>;
> 
> As per my comment in patch #1, you are probably ending up throttling
> to 500 MHz, not 400 MHz.  Below will all have similar problems unless
> we actually delete the 500 MHz operating point.

Thanks for pointing that out. As per disussion on patch #1 we'll
disable the 500 MHz OPP to stay in sync with downstream and avoid
problems in case someone decides to re-purpose NPLL.

> > +               };
> > +
> > +               /*
> > +                * Slightly after we throttle the GPU, we'll also make sure that
> > +                * the CPU can't go faster than 1.4 GHz.  Note that we won't
> > +                * throttle the CPU lower than 1.4 GHz due to GPU heat--we'll
> > +                * let the CPU do the rest itself.
> > +                */
> > +               gpu_warm_limit_cpu {
> > +                       trip = <&gpu_alert_warm>;
> > +                       cooling-device = <&cpu0 4 4>;
> 
> Shouldn't you list cpu1, cpu2, and cpu3 too?  That'd match what
> upstream did elsewhere in this file?

ack, should have noticed, I 'yelled' at others before for not doing this ...

Re: [PATCH 1/2] ARM: dts: rockchip: Limit GPU frequency on veyron mickey to 300 MHz when the CPU gets very hot

[Doug Anderson]

Hi,

On Mon, May 20, 2019 at 10:01 AM Matthias Kaehlcke <mka@chromium.org> wrote:
>
> On rk3288 the CPU and GPU temperatures are correlated. Limit the GPU
> frequency on veyron mickey to 300 MHz for CPU temperatures >= 85°C.
>
> This matches the configuration of the downstream Chrome OS 3.14 kernel,
> the 'official' kernel for mickey.
>
> Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
> ---
> Note: this patch depends on "ARM: dts: rockchip: Add #cooling-cells
> entry for rk3288 GPU" (https://lore.kernel.org/patchwork/patch/1075005/)
> ---
>  arch/arm/boot/dts/rk3288-veyron-mickey.dts | 6 ++++++
>  1 file changed, 6 insertions(+)
>
> diff --git a/arch/arm/boot/dts/rk3288-veyron-mickey.dts b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> index d889ab3c8235..f118d92a49d0 100644
> --- a/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> +++ b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> @@ -125,6 +125,12 @@
>                                          <&cpu2 8 THERMAL_NO_LIMIT>,
>                                          <&cpu3 8 THERMAL_NO_LIMIT>;
>                 };
> +
> +               /* At very hot, don't let GPU go over 300 MHz */
> +               cpu_very_hot_limit_gpu {
> +                       trip = <&cpu_alert_very_hot>;
> +                       cooling-device = <&gpu 2 2>;
> +               };

Two things:

A) If I'm reading things properly, you're actually limiting things to
400 MHz.  This is because you don't have <https://crrev.com/c/1574579>
which deletes the 500 MHz GPU operating point.  So on upstream the
available points are:

0: 600 MHz
1: 500 MHz
2: 400 MHz
3: 300 MHz
4: 200 MHz
5: 100 MHz

...and downstream:

0: 600 MHz
1: 400 MHz
2: 300 MHz
3: 200 MHz
4: 100 MHz

Thinking about it more, I bet Heiko would actually be OK deleting the
500 MHz GPU operating point for veyron.  Technically it's not needed
upstream because upstream doesn't have our hacks to allow re-purposing
NPLL for HDMI (so they _can_ make 500 MHz) but maybe we can make the
argument that these laptops have only ever been tested with the 500
MHz operating point removed and also that eventually someonje will
probably figure out a way to re-purpose NPLL for HDMI even upstream...


B) It seems like in the same patch you'd want to introduce
"cpu_warm_limit_gpu", AKA:

cpu_warm_limit_gpu {
  trip = <&cpu_alert_warm>;
  cooling-device =
  <&gpu 1 1>;
};


-Doug

Re: [PATCH 1/2] ARM: dts: rockchip: Limit GPU frequency on veyron mickey to 300 MHz when the CPU gets very hot

[Heiko Stübner]

Am Montag, 20. Mai 2019, 22:16:46 CEST schrieb Doug Anderson:
> Hi,
> 
> On Mon, May 20, 2019 at 10:01 AM Matthias Kaehlcke <mka@chromium.org> wrote:
> >
> > On rk3288 the CPU and GPU temperatures are correlated. Limit the GPU
> > frequency on veyron mickey to 300 MHz for CPU temperatures >= 85°C.
> >
> > This matches the configuration of the downstream Chrome OS 3.14 kernel,
> > the 'official' kernel for mickey.
> >
> > Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
> > ---
> > Note: this patch depends on "ARM: dts: rockchip: Add #cooling-cells
> > entry for rk3288 GPU" (https://lore.kernel.org/patchwork/patch/1075005/)
> > ---
> >  arch/arm/boot/dts/rk3288-veyron-mickey.dts | 6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> > diff --git a/arch/arm/boot/dts/rk3288-veyron-mickey.dts b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> > index d889ab3c8235..f118d92a49d0 100644
> > --- a/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> > +++ b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> > @@ -125,6 +125,12 @@
> >                                          <&cpu2 8 THERMAL_NO_LIMIT>,
> >                                          <&cpu3 8 THERMAL_NO_LIMIT>;
> >                 };
> > +
> > +               /* At very hot, don't let GPU go over 300 MHz */
> > +               cpu_very_hot_limit_gpu {
> > +                       trip = <&cpu_alert_very_hot>;
> > +                       cooling-device = <&gpu 2 2>;
> > +               };
> 
> Two things:
> 
> A) If I'm reading things properly, you're actually limiting things to
> 400 MHz.  This is because you don't have <https://crrev.com/c/1574579>
> which deletes the 500 MHz GPU operating point.  So on upstream the
> available points are:
> 
> 0: 600 MHz
> 1: 500 MHz
> 2: 400 MHz
> 3: 300 MHz
> 4: 200 MHz
> 5: 100 MHz
> 
> ...and downstream:
> 
> 0: 600 MHz
> 1: 400 MHz
> 2: 300 MHz
> 3: 200 MHz
> 4: 100 MHz
> 
> Thinking about it more, I bet Heiko would actually be OK deleting the
> 500 MHz GPU operating point for veyron.  Technically it's not needed
> upstream because upstream doesn't have our hacks to allow re-purposing
> NPLL for HDMI (so they _can_ make 500 MHz) but maybe we can make the
> argument that these laptops have only ever been tested with the 500
> MHz operating point removed and also that eventually someonje will
> probably figure out a way to re-purpose NPLL for HDMI even upstream...

Yeah. Dropping the opp sounds sensible ... for the npll-related thing
and also if you're really running into thermal constraints it might be
good to give the system a bit more breathing room?


Heiko


> B) It seems like in the same patch you'd want to introduce
> "cpu_warm_limit_gpu", AKA:
> 
> cpu_warm_limit_gpu {
>   trip = <&cpu_alert_warm>;
>   cooling-device =
>   <&gpu 1 1>;
> };
> 
> 
> -Doug

Re: [PATCH 1/2] ARM: dts: rockchip: Limit GPU frequency on veyron mickey to 300 MHz when the CPU gets very hot

[Matthias Kaehlcke]

On Mon, May 20, 2019 at 01:16:46PM -0700, Doug Anderson wrote:
> Hi,
> 
> On Mon, May 20, 2019 at 10:01 AM Matthias Kaehlcke <mka@chromium.org> wrote:
> >
> > On rk3288 the CPU and GPU temperatures are correlated. Limit the GPU
> > frequency on veyron mickey to 300 MHz for CPU temperatures >= 85°C.
> >
> > This matches the configuration of the downstream Chrome OS 3.14 kernel,
> > the 'official' kernel for mickey.
> >
> > Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
> > ---
> > Note: this patch depends on "ARM: dts: rockchip: Add #cooling-cells
> > entry for rk3288 GPU" (https://lore.kernel.org/patchwork/patch/1075005/)
> > ---
> >  arch/arm/boot/dts/rk3288-veyron-mickey.dts | 6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> > diff --git a/arch/arm/boot/dts/rk3288-veyron-mickey.dts b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> > index d889ab3c8235..f118d92a49d0 100644
> > --- a/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> > +++ b/arch/arm/boot/dts/rk3288-veyron-mickey.dts
> > @@ -125,6 +125,12 @@
> >                                          <&cpu2 8 THERMAL_NO_LIMIT>,
> >                                          <&cpu3 8 THERMAL_NO_LIMIT>;
> >                 };
> > +
> > +               /* At very hot, don't let GPU go over 300 MHz */
> > +               cpu_very_hot_limit_gpu {
> > +                       trip = <&cpu_alert_very_hot>;
> > +                       cooling-device = <&gpu 2 2>;
> > +               };
> 
> Two things:
> 
> A) If I'm reading things properly, you're actually limiting things to
> 400 MHz.  This is because you don't have <https://crrev.com/c/1574579>
> which deletes the 500 MHz GPU operating point.  So on upstream the
> available points are:
> 
> 0: 600 MHz
> 1: 500 MHz
> 2: 400 MHz
> 3: 300 MHz
> 4: 200 MHz
> 5: 100 MHz
> 
> ...and downstream:
> 
> 0: 600 MHz
> 1: 400 MHz
> 2: 300 MHz
> 3: 200 MHz
> 4: 100 MHz

Thanks spotting this!

> Thinking about it more, I bet Heiko would actually be OK deleting the
> 500 MHz GPU operating point for veyron.  Technically it's not needed
> upstream because upstream doesn't have our hacks to allow re-purposing
> NPLL for HDMI (so they _can_ make 500 MHz) but maybe we can make the
> argument that these laptops have only ever been tested with the 500
> MHz operating point removed and also that eventually someonje will
> probably figure out a way to re-purpose NPLL for HDMI even upstream...

Looks like Heiko is indeed ok with it, so let's remove the OPP and be
in sync with downstream on this.

> B) It seems like in the same patch you'd want to introduce
> "cpu_warm_limit_gpu", AKA:
> 
> cpu_warm_limit_gpu {
>   trip = <&cpu_alert_warm>;
>   cooling-device =
>   <&gpu 1 1>;
> };

Makes sense to do it in the same patch, will add it in v2.