lkml.org 
[lkml]   [2015]   [Feb]   [25]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH] thermal: armada: read stable temp on Armada XP
Hi Gregory, Eduardo,

On Wed, Feb 25, 2015 at 05:10:14PM +0100, Gregory CLEMENT wrote:
> By using it by default do you mean removing marvell,armadaxp-thermal
> and adding armadaxp-filtered-thermal instead ?

Yes, replacing it in device tree. For me,
/sys/class/thermal/thermal_zone0/temp returns the same temperature but
with less variability between samples. My intent was to switch the
source of the data and not affect user space other than providing a
more stable reading.

> Tyler,
> In the meantime could you double check your values? The temperature on my board
> seemed broken on my board. If needed I can check on an other board. By the way
> on which board/product did you try it?

I'm on a custom board with the 4-core MV78460. In addition to my
patch, this is new device tree entry.

thermal@184c4 {
compatible = "marvell,armadaxp-filtered-thermal";
reg = <0x184c4 0x4
0x184d0 0x4>;
status = "okay";
};

I dumped some raw samples of the temperature fields in each of these
registers. This CSV contains the raw values converted to decimal.
http://pastebin.com/Umc3cy5L
The first column is the current register at 0x182b0 27:19 and the
second is the register at 0x184c4 9:1.

Here's a quick plot of a larger sample size while the temperature was rising.
https://imgur.com/E9HlSBx
The blue value is the current 0x182b0 register which seems to bounce
around the green value from 0x184c4.

I'll try a test of instantiating both at the same time and comparing
the final output of the driver.

On Wed, Feb 25, 2015 at 1:39 PM, Eduardo Valentin <edubezval@gmail.com> wrote:
>> Does that new thermal sensors only improve the stability or does it
>> also modify the value?
>>
>> In the second case it will more or less break the user space expectation.
>
> Yeah, I agree here. We need to understand if this is:
> (1) a fix of which register to use. In that case, the old dtbs are
> essentially wrong, and the driver would need to adapt, I would say.
> (2) a way to report better temperature extrapolations. then, this is
> essentially a new temp sensor, and we should have two separated
> compatible. in other words, just keep the patch the way it is.

This *might* be a different physical sensor, or it could be from the
same source with some averaging/filtering applied in hardware. The
conversion formula is the same, however, and I get similar raw values
from both.

> Yes. Typically I ask to see the complete series, even if I am not taking
> the DTS changes. That is, you send a complete series, with changes in
> the kernel code and in the DTS, copying the required audience (from
> kernel side and from DT side). Once the changes are accepted, the
> patches will be picked by the correct tree maintainer. It is common that
> DTS changes go via the platform tree, to avoid conflicts though.

Thanks for the clarification. I'll send both in the next version as I
suspect there will be a v2 of this change.


\
 
 \ /
  Last update: 2015-02-25 21:01    [W:0.056 / U:0.144 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site