ARM: dts: bcm283x: Fix critical trip point

Commit Message

Stefan Wahren Nov. 30, 2019, 12:31 p.m. UTC

During definition of the CPU thermal zone of BCM283x SoC family there was
a misunderstanding of the meaning "criticial trip point" and the thermal
throttling range of the VideoCore firmware. The latter one takes effect
when the core temperature is between 80 and 85 degree celsius. So the
current critical trip point doesn't make sense, because the
thermal shutdown appears before the firmware has a chance to throttle the
ARM core(s).

Fix these unwanted shutdowns by increasing the critical trip point
to a value which shouldn't be reached with working thermal throttling.

Fixes: 0fe4d2181cc4 ("ARM: dts: bcm283x: Add CPU thermal zone with 1 trip point")
Signed-off-by: Stefan Wahren <[e-mail]>
---
 arch/arm/boot/dts/bcm283x.dtsi | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

--
2.7.4

Comments

Florian Fainelli Dec. 9, 2019, 6:17 p.m. UTC | #1

On 11/30/19 5:54 AM, Lukas Wunner wrote:
> On Sat, Nov 30, 2019 at 01:31:13PM +0100, Stefan Wahren wrote:
>> During definition of the CPU thermal zone of BCM283x SoC family there was
>> a misunderstanding of the meaning "criticial trip point" and the thermal
>> throttling range of the VideoCore firmware. The latter one takes effect
>> when the core temperature is between 80 and 85 degree celsius. So the
>> current critical trip point doesn't make sense, because the
>> thermal shutdown appears before the firmware has a chance to throttle the
>> ARM core(s).
> 
> FWIW, the thermal throttling range goes way above 85°C.
> 
> At Kunbus we've performed numerous tests in a climate chamber
> and my recollection is that a CM3 starts throttling above 80°C
> with the frequency hovering between 600 and 1200 MHz.
> 
> Once 85°C is reached, the frequency is fixed at 600 MHz.
> 
> But there a additional tripping points above that when the frequency
> gets further reduced to 400 MHz and later 300 MHz.

Does that mean that you are okay with 90°C here, or would you rather see
this be change to 85°C?

Stefan Wahren Dec. 9, 2019, 7:31 p.m. UTC | #2

Am 09.12.19 um 19:17 schrieb Florian Fainelli:
> On 11/30/19 5:54 AM, Lukas Wunner wrote:
>> On Sat, Nov 30, 2019 at 01:31:13PM +0100, Stefan Wahren wrote:
>>> During definition of the CPU thermal zone of BCM283x SoC family there was
>>> a misunderstanding of the meaning "criticial trip point" and the thermal
>>> throttling range of the VideoCore firmware. The latter one takes effect
>>> when the core temperature is between 80 and 85 degree celsius. So the
>>> current critical trip point doesn't make sense, because the
>>> thermal shutdown appears before the firmware has a chance to throttle the
>>> ARM core(s).
>> FWIW, the thermal throttling range goes way above 85°C.
>>
>> At Kunbus we've performed numerous tests in a climate chamber
>> and my recollection is that a CM3 starts throttling above 80°C
>> with the frequency hovering between 600 and 1200 MHz.
>>
>> Once 85°C is reached, the frequency is fixed at 600 MHz.
>>
>> But there a additional tripping points above that when the frequency
>> gets further reduced to 400 MHz and later 300 MHz.
> Does that mean that you are okay with 90°C here, or would you rather see
> this be change to 85°C?

From my understanding the change is okay, but the wording in the patch
description is misleading.

Suggestion:

... The latter one takes effect when the core temperature is at least 85
degree celsius or higher.

Florian Fainelli Dec. 10, 2019, 8:37 p.m. UTC | #3

On 12/9/19 11:31 AM, Stefan Wahren wrote:
> Am 09.12.19 um 19:17 schrieb Florian Fainelli:
>> On 11/30/19 5:54 AM, Lukas Wunner wrote:
>>> On Sat, Nov 30, 2019 at 01:31:13PM +0100, Stefan Wahren wrote:
>>>> During definition of the CPU thermal zone of BCM283x SoC family there was
>>>> a misunderstanding of the meaning "criticial trip point" and the thermal
>>>> throttling range of the VideoCore firmware. The latter one takes effect
>>>> when the core temperature is between 80 and 85 degree celsius. So the
>>>> current critical trip point doesn't make sense, because the
>>>> thermal shutdown appears before the firmware has a chance to throttle the
>>>> ARM core(s).
>>> FWIW, the thermal throttling range goes way above 85°C.
>>>
>>> At Kunbus we've performed numerous tests in a climate chamber
>>> and my recollection is that a CM3 starts throttling above 80°C
>>> with the frequency hovering between 600 and 1200 MHz.
>>>
>>> Once 85°C is reached, the frequency is fixed at 600 MHz.
>>>
>>> But there a additional tripping points above that when the frequency
>>> gets further reduced to 400 MHz and later 300 MHz.
>> Does that mean that you are okay with 90°C here, or would you rather see
>> this be change to 85°C?
> 
> From my understanding the change is okay, but the wording in the patch
> description is misleading.
> 
> Suggestion:
> 
> ... The latter one takes effect when the core temperature is at least 85
> degree celsius or higher.

Applied to devicetree/fixes with this description fixed up, thanks Stefan.

Patch

diff --git a/arch/arm/boot/dts/bcm283x.dtsi b/arch/arm/boot/dts/bcm283x.dtsi
index 3caaa57..8394916 100644
--- a/arch/arm/boot/dts/bcm283x.dtsi
+++ b/arch/arm/boot/dts/bcm283x.dtsi
@@ -37,7 +37,7 @@ 

 			trips {
 				cpu-crit {
-					temperature	= <80000>;
+					temperature	= <90000>;
 					hysteresis	= <0>;
 					type		= "critical";
 				};