Re: [PATCH 1/2] thermal: db8500: Finalize device tree conversion

From: Daniel Lezcano <daniel.lezcano@linaro.org>
To: Linus Walleij <linus.walleij@linaro.org>,
	Vincent Guittot <vincent.guittot@linaro.org>
Cc: Zhang Rui <rui.zhang@intel.com>,
	Eduardo Valentin <edubezval@gmail.com>,
	Linux PM list <linux-pm@vger.kernel.org>,
	Lee Jones <lee.jones@linaro.org>
Subject: Re: [PATCH 1/2] thermal: db8500: Finalize device tree conversion
Date: Tue, 27 Aug 2019 15:54:44 +0200	[thread overview]
Message-ID: <f683466b-a554-87c4-0bfd-4d56bd1142fe@linaro.org> (raw)
In-Reply-To: <CACRpkdYO7zzrUg=7=4D7XMUwoa6XQggm0jB1PtPhKK=3EV9AyQ@mail.gmail.com>

On 27/08/2019 15:07, Linus Walleij wrote:
> On Thu, Aug 22, 2019 at 8:16 AM Daniel Lezcano
> <daniel.lezcano@linaro.org> wrote:
> 
>> Another point is there are too many trip points, two should be enough,
>> one for throttling and one for critical, the governor will handle that
>> properly by stepping the opps.
> 
> I looked a bit into this, and it seems that there are a number of
> trip points simply for inspecting the system (Vincent may know better
> because he worked directly with this): the thermal sensor can only
> set an IRQ for a certain temperature, there is no interface to just
> read the temperature. 

I see, it is a limitation of the sensor.

> By setting a number of trip points and
> first let the IRQ fire at the lowest you see how the system moves
> up in the temperature points and that makes it easier to inspect,
> so this sensor will set 70 degress, wait until the IRQ fires,
> then set 75 degrees, wait until IRQ fires, set 80 degrees...
> etc. It does the same with the IRQ for falling temperature so it
> moves up and down this scale, and interpolates the temperature
> between the trip points.
> 
> It is a bit crude but it gives some kind of granularity and control
> over the reported temperature from the sensor.
> 
> I think this is something that should be done inside the
> driver, not by abusing the trip points, so I try to rewrite the logic
> so that the driver fires these IRQs internally without using any
> trip points from the thermal subsystem.

In the comments, it is mentioned something is missing with the PRCMU but
when the interface will be available, the get_temp won't have to
interpolate. Wouldn't make sense to add the PRCMU interface instead ?

-- 
 <http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog