Hi,
I've noticed a strande behaviour of the internal watchdog on DM365. First of all, the linux driver is not very compliant to the standard: if the user application that opens the driver device then closes it "cleanly", the watchdog should disengage and not reset the system. Instead I've noticed that if I terminate (cleanly) the user space watchdog application, closing the device handle after writing "V", the system resets as if the watchdog is still enabled but nobody is updating it.
But this is only a missing feature.
What's really strange is that after the watchdog triggers and resets the DM365, the system resets again 60 seconds after the watchdog is enabled again, and never stops this cycle of reboots. This is very strange and I'd like to know if this happens on other boards too. This is how to try it:
Preparation:
1) Build your root system so that davinci_wdt driver is loaded as modules (but perhaps it happens if embedded in kernel too).
2) Build your root system so that it run you preferred "watchdog" keepalive app. There is an example app in linux kernel documentation.
Test:
1) Boot the system, watchdog should be enabled but userspace app should keep it alive.
2) Kill the app, so that after 60 seconds the watchdog reboots DM365
3) System reboots, this should reload the davinci_wdt driver and the app.
Now the system should not reboot again... But instead the system resets 60 seconds after point 3 happens. This is like if after the first time the wdt triggers, it is not reset/initialized correctly and so it keeps triggering even if it is updated by the app.
Very strange...
