Is there a reason for imx8mq thermal reading to get stuck and not represent actual SOC die temp

[trk_golf]

There have been a couple times we've noticed on the imx8mq that when reading the SOC die temperature it seems to get "stuck" and doesn't increase as we would typically see.

The value is ready this way:

cat /sys/class/thermal/thermal_zone0/temp

We've seen it get stuck at 36C and 39C and I think also 32 C on a different unit.

There doesn't seem to be a way to get it unstuck except to reboot the device.

We've confirmed it isn't responding properly because we'll exercise all 4 cores to generate heat and the temperature will remain the same.    This behavior is only rarely observed and most of the time it is reflecting the temperature properly.

Currently we are on 5.10.52 kernel.

Any explanation for this behavior or methods to get it unstuck without rebooting.

Thanks.

[Zhiming_Liu]

Hi @trk_golf 

Can you try more bsp version, like L5.10.72 and L5.15.5?

[trk_golf]

Does moving to a newer BSP solve the issue for sure?   If so, is there a specific patch for uboot or linux that can be applied to resolve the problem?  It is not so easy for us to upgrade to an entirely new BSP to try this out if there's no guarantee it will solve the issue.