I’m having trouble with my Think Penguin TPE-R1100 wireless access points hanging after heavy Wi-Fi activity or the like. But, here’s the thing. I saw the hardware has support for a watchdog timer, and presumably LibreCMC/OpenWRT uses that to prevent device hanging. So, maybe I should check if something is wrong in that stack. But in the end, I found nothing, everything is swell.
Anyways, the tech learning sure was nice, though, so I’ll summarize what I’ve learned here.
The watchdog timer for this device is implemented here. Looking through the code, all is swell and nice, support for “magic close” is in there, so if the software crashes without a graceful close, the watchdog timer will still keep ticking and reset the system accordingly.
linux-VERSION/drivers/watchdog/ath79_wdt.c
-> ath79_wdt_keepalive
So, where is the code implementation in LibreCMC/OpenWRT? Until 2013,
OpenWRT previously used the watchdog daemon tool from Busybox. After
that, they merged the functionality into procd. However, although
procd opened and ticked the watchdog timer properly, it did not
implement magic close, so any time you stopped it, the watchdog timer
would keep ticking and reset the system since the driver required
magic close. Not until 2017 did we get a patch to implement the magic
close feature in procd, so then you could gracefully disable the
watchdog timer in procd with ease, for use and control with your own
software.
20200526/DuckDuckGo openwrt watchdog
20200526/http://kernelreloaded.com/manually-controlling-openwrt-hardware-watchdog/
Here is some great information on checking the status of the watchdog daemon on OpenWRT/LibreCMC.
20200526/https://openwrt.org/docs/guide-user/hardware/watchdog
Good point here. Having trouble with your device watchdog-reset? There are USB watchdog devices you can use, basically a microcontroller that controls relays on a pin header. You can then use that to control the reset pin on your board, the power, or both.
Now I imagine an extended watchdog serial protocol, send debug information to the watchdog device to save, which can then be retrieved and sent out over the network on successful reboot.