Orange Pi Zero LTS Incorrect Temps Reported

[haxxa]

Armbianmonitor:

http://ix.io/1UAf

As stated in the topic the new Orange Pi Zero LTS reports incorrect system temps (likely due to 16bit overflow if I were to guess): 

~$ cat /sys/devices/virtual/thermal/thermal_zone0/temp
-2178

Model Purchased: https://www.aliexpress.com/item/4000070093365.html

[nvh]

Can confirm that, mine gives about the same (wrong) temperature readings

[guidol]

got "deeper" temps with a OPi One and kernel 5.3.0 - RC6 and RC7 (RC3 seems to be OK)

maybe the .dtb did change for both mainline and dev kernel?:

 

[guidol]

on my OPi Zero (non LTS) I got "working" temps with kernel 5.2.2:
 

Debian Buster with Armbian Linux 5.2.2-sunxi

root@opi-zero(192.168.6.99):~# armbianmonitor -m
Stop monitoring using [ctrl]-[c]
Time        CPU    load %cpu %sys %usr %nice %io %irq   CPU

18:11:14: 1008MHz  1.67   0%   0%   0%   0%   0%   0% 82.2°C
18:11:19: 1008MHz  1.69  17%   3%   4%   0%   9%   0% 79.6°C
18:11:24: 1008MHz  1.64  33%   8%  23%   0%   1%   0% 77.8°C
18:11:30: 1008MHz  1.59  21%   2%  13%   0%   5%   0% 80.5°C
18:11:35: 1008MHz  2.58  37%   8%  26%   0%   2%   0% 77.4°C
18:11:40: 1008MHz  2.45  26%   3%  21%   0%   1%   0% 83.5°C

but now after the update to kernel 5.3.0-RC6 it also has freezing temperatures:
 

Debian Buster with Armbian Linux 5.3.0-rc6-sunxi

root@opi-zero(192.168.6.99):~# armbianmonitor -m
Stop monitoring using [ctrl]-[c]
Time        CPU    load %cpu %sys %usr %nice %io %irq   CPU  C.St.

18:14:01: 1008MHz  1.21  28%  13%  13%   0%   1%   0% -172062°C  0/4
18:14:06:  480MHz  1.11   1%   0%   0%   0%   0%   0% -173998°C  0/4
18:14:11:  480MHz  1.02   0%   0%   0%   0%   0%   0% -174966°C  0/4
18:14:16:  480MHz  0.94   0%   0%   0%   0%   0%   0% -175208°C  0/4

its seems that there are in the newer dtb/dts some fewer cpu-temp stages (no warm - only cpu-hot and cpu-very-hot) and there is a thermal ths calibration, which I didnt see before in the dtb/dts like :
 

ths_calibration = "/soc/eeprom@1c14000/thermal-sensor-calibration@34";

 eeprom@1c14000 {
                        reg = < 0x1c14000 0x400 >;
                        compatible = "allwinner,sun8i-h3-sid";
                        #address-cells = < 0x01 >;
                        #size-cells = < 0x01 >;
                        phandle = < 0x3b >;

                        thermal-sensor-calibration@34 {
                                reg = < 0x34 0x04 >;
                                phandle = < 0x28 >;
                        };
                };

 

[Rakanskiy]

Can confirm that both of my Orange Pi Zero LTS reporting wrong low temperature with tested by me kernels (stock from vendor, armbian buster with 4.19.62-sunxi and 4.19.59-sunxi).

 

 

 

[guidol]

5.3.0 RC8 also has these low temperatures:
Debian Buster with Armbian Linux 5.3.0-rc8-sunxi
package bsp-kernel[5.96] u-boot[5.96] dtb[5.96] firmware[5.96] config[5.96]

 

root@nanopiduo(192.168.6.151):/sys/class/leds# armbianmonitor -m
Stop monitoring using [ctrl]-[c]
Time        CPU    load %cpu %sys %usr %nice %io %irq   CPU  C.St.

20:30:48: 1008MHz  0.35  12%   2%   8%   0%   1%   0% -203885°C  0/4
20:30:53:  480MHz  0.32   0%   0%   0%   0%   0%   0% -205337°C  0/4
20:30:58:  480MHz  0.29   0%   0%   0%   0%   0%   0% -205458°C  0/4
20:31:04:  480MHz  0.32   0%   0%   0%   0%   0%   0% -205700°C  0/4

@Igor @martinayotte Do you know something about some a commit/merge or else which did change some of the .DTBs for Allwinner-devices like NPi/OPi which apply after 5.3.0 RC3?

[Igor]

4 minutes ago, guidol said:

Do you know something about some a commit/merge or else which did change some of the .DTBs for Allwinner-devices like NPi/OPi which apply after 5.3.0 RC3?

No idea. This means some Allwinner related code landed to mainline kernel and there is possible some conflict with our patches. Or just a bug.

[guidol]

3 minutes ago, Igor said:

No idea. This means some Allwinner related code landed to mainline kernel and there is possible some conflict with our patches. Or just a bug.

the login-motd hasnt a temperature readout (because of?)

[   11.543638] thermal thermal_zone0: failed to read out thermal zone (-110)

 but armbianmonitor -m and htop does show this very low temperature:

root@npi-duo(192.168.6.113):~# more /sys/class/thermal/thermal_zone0/temp
-207152

 

[Igor]

5 minutes ago, guidol said:

but armbianmonitor -m and htop does show this very low temperature:

Kernel driver is reporting wrong temperature - input to those scripts and apps (which is the same) is wrong. I saw this bug. That's all I can do right now.

[martinayotte]

1 hour ago, guidol said:

these low temperatures:

I've some OPiPC+ that still at RC3, and the bug is not present, while some other with RC6, the issue is there ...

So, maybe @megous has committed something related to that ...

[guidol]

1 hour ago, martinayotte said:

I've some OPiPC+ that still at RC3, and the bug is not present, while some other with RC6, the issue is there ...

So, maybe @megous has committed something related to that ...

Hmmm - how much time has gone since RC3? - because thermal/THS before 8 days:

"remove a83t-OPP and H3-H5-THS patches since Megous 5.3.0-RC6 have them"
martinayotte committed 8 days ago

https://github.com/armbian/build/commit/b53556a5f1ada29169d08b6a106a01f87d763774

 

could something like this change it? H3/H5 and RC6 would make it a "good" nominee?

[martinayotte]

16 hours ago, guidol said:

martinayotte committed 8 days ago

My "delete" commit was due to duplicate nodes already present in @megous branch, but I didn't check if his commits was coming from mainstream and were healthy ...

It maybe not only DT issue, but from thermal driver, see here activities in his branch : https://github.com/megous/linux/commits/orange-pi-5.3/drivers/thermal/sun8i_thermal.c

 

[Igor]

I think this partially explains

https://github.com/megous/linux/commit/0376281ed8ef09e81dbc6ce72a7b3f7b675962b5#diff-445fc9c5ea648f3611cc61ce04a601ef

[guidol]

@Igor @martinayotte

the bad "freezeing"-temperatures seems to be fixed with the new kernel/dtb with date-code 190913:

Debian Buster with Armbian Linux 5.3.0-rc8-sunxi
package bsp-kernel[5.96.190913] u-boot[5.96] dtb[5.96.190913] firmware[5.96] config[5.96]

 

which I received via apt upgrade

 

Many thanks to who did fix this!

 

[martinayotte]

29 minutes ago, guidol said:

Many thanks to who did fix this!

Thanks to that mysterious dev !

[rain45]

 

On orange pi one plus does not work 

/usr/bin/armbianmonitor: line 385: read: read error: 0: Invalid argument
/usr/bin/armbianmonitor: line 386: [: -ge: unary operator expected

Edited September 13, 2019 by rain45
addition

[Igor]

1 hour ago, martinayotte said:

Thanks to that mysterious dev !

I believe its this change:

https://github.com/armbian/build/commit/bea1df9ed6e4d1661c981ed488cd17dc2cf3d0e2#diff-4e3f23ec46130644ad2268eda44b6e3eR3356

[guidol]

1 hour ago, rain45 said:

 

On orange pi one plus does not work 

/usr/bin/armbianmonitor: line 385: read: read error: 0: Invalid argument
/usr/bin/armbianmonitor: line 386: [: -ge: unary operator expected

 

before the freeze-temperatures we had this error-message on some boards.

So we are where we were before.

This error was in/with newer kernels (dtb?) 

 

[Igor]

1 minute ago, guidol said:

So we are where we were before.

 

You have this? On which boards?

 

Do we need to create a debug tool for our debugger? 

[rain45]

1 hour ago, guidol said:

This error was in/with newer kernels (dtb?) 

 

Yes

[guidol]

9 hours ago, Igor said:

 

You have this? On which boards?

 

Do we need to create a debug tool for our debugger? 

I had this some months ago, but now @rain45 has it - but I dont know with wich version...

[rain45]

linux-dtb-dev-sunxi64 5.96.190913

5.3.0-rc8-sunxi64

[guidol]

12 hours ago, Igor said:

 

You have this? On which boards?

Do we need to create a debug tool for our debugger? 

@Igormaybe we need ony a "old" fix-commit again for his device?

 

https://github.com/armbian/build/commit/76ac54aae381714cee7bf1b03d4bce097b1630de

[Rakanskiy]

The issue still present

[Igor]

15 hours ago, Rakanskiy said:

The issue still present

 

This is still an amateur project which is dealing with thousands of issues and thousands of people with a pinch of needed resources. When it's done it's done. Try upgrading to kernel 5.3.y (armbian-config -> system -> alternative kernels) since we will be ditching 4.19.y soon.

 

If you help around https://www.armbian.com/get-involved things go little faster.

[yuanyunlong]

I have encountered this problem before.
Later I dissect the armbian script file and found that it can be read in another way.

cat /etc/armbianmonitor/datasources/soctemp |awk '{printf("%d",$1/1000)}'

 

[sfx2000]

On 9/8/2019 at 5:04 AM, haxxa said:

likely due to 16bit overflow if I were to guess

 

Quick glance at the numbers, I'd say yes...

20:30:48: 1008MHz  0.35  12%   2%   8%   0%   1%   0% -203885°C  0/4

obviously not true - but one could test it out and run the clocks up, lol...

 

To paraphrase - what's cooler than being cool - ice cold...

 

 

[dolphs]

Can one verify with kernel 5.4 how things go in terms of temperature, please? 

I'm about to go ahead an purchase one ( see also here ),

but not sure if that would resist without heatsink in my meter cupboard ...

cheers

 

[Werner]

10 hours ago, dolphs said:

Can one verify with kernel 5.4 how things go in terms of temperature, please? 

I'm about to go ahead an purchase one ( see also here ),

but not sure if that would resist without heatsink in my meter cupboard ...

cheers

 

May I ask why it has to be an OPi Zero? OPi One has HDMI, better voltage regulation, full H3 instead of H2+, no broken-by-design wifi chip and is even cheaper than the OPi Zero in the 512MB version.

[dolphs]

1 hour ago, Werner said:

May I ask why it has to be an OPi Zero? OPi One has HDMI, better voltage regulation, full H3 instead of H2+, no broken-by-design wifi chip and is even cheaper than the OPi Zero in the 512MB version.

- [x] HDMI -> no need ( ssh and http enough )

 

Those caught my interest:

- [v] better voltage regulation -> that was my hunch but trying to verify how H2+ is nowadays ( kernel 5.3/ 5.4 )

- [v] full H3 instead of H2+

- [v] no broken by-design WiFi chip -> OPi One does not have WiFi imho

 

Zero LTS hopefully has no longer that "heat" problem , LTS states: " low running temperature and low power consumption. ", but I dunno.

Though I was not aware of crappy WiFi (xr819?), is mandatory though as I dont have cables in the meter cupboard, unless I'd use a HomePlug

Except for that have older H5 ( first Neo2 series ), but without WiFi and H2+ was the missing piece I did not have stocked ;-),

thus consider this more like a little "funny" project and might swap the board sooner or later, only stuff on there would be armbian ( buster ) with  domoticz

thanks!