NanoPi R2S/R4S warm boots often fail

[Djfe]

Bug report

What is the problem?
Updates or restarts of the rockchip based plattforms fail if it's a warm reboot (weeks of uptime).
Devices require a manual power-cycle after updating (more often than not)
I have never gotten my hands on one, yet. (when it happens)
So I can't tell you whether it's still reachable via it's nextnode address or whether the boot process fails at a certain point.

Previously we expected the update to be the cause. But now we had a case where a package rebooted a NanoPi R4S due to a missing gateway and it also stayed down till it was power-cycled.

This has happened atleast 10 times now on different R2S and R4S devices. We disabled autoupdate for some of them already to prevent them from going down (when placed in remote locations).

What is the expected behaviour?
Reboots should never fail

Gluon Version:
bd5c28e
This happened right from when we started using them. Even Gluon v2022.1 is likely affected. (it's just that we don't restart these often)

Site Configuration:
https://github.com/ffac/site/tree/v2023.2.3-6

Custom patches:
https://github.com/ffac/site/tree/v2023.2.3-6/patches

Next Steps:
Hopefully we can get this into a reproducible state.
We are going to let a device run and give it a cronjob to reboot daily, weekly, bi-weekly, monthly
Until we get it to fail and then take a look. And then check whether the uptime is related to the issue or not.

Maybe it's easily reproducible and it's related to some chip not being reset by the kernel on reboot. (reset pin missing in the dts)

[grische]

On the node where I saw the problem, replacing the sdcard with something better like a Sandisk Ultra/Extreme fixed any of such problems.