Systemd shutdown, halt and reboot handling

egor.utsov

@SGaist It can happen when user login through ssh and cissue systemctl poweroff for example

SGaist

Then as I already suggested, dump a picture in your frame buffer as part of the procedure. Your application will be killed quickly so it won't have the time to show anything special in that case.

kshegunov

In addition to what @SGaist wrote, there are limitations of what you can and can't do in signal handlers. Not everything is allowed.

egor.utsov

@kshegunov it is done not from signal handler, but from slot, connected to qsocketnotifier.

SGaist

Depending on what you do, you will get a SIGTERM and a SIGKILL right after. The later you can't interrupt nor try to do some more stuff.

egor.utsov

@SGaist but as i remember systemd documentation stated, that process have 60 seconds until sigkill

SGaist

60 second per running process ?

egor.utsov

@SGaist i guessed so. Saw it somewhere in docs.

SGaist

Can you find that information again ?

egor.utsov

@SGaist No. Maybe i am mistaken. I found some info here: http://manpages.ubuntu.com/manpages/cosmic/en/man5/systemd.service.5.html

Check section about TimeoutStopSec=. On my system it is commented out, but here stated, that it is even 90 s. As i understand it is applied per-unit.

SGaist

Is your application started as a systemd service ?

egor.utsov

@SGaist Yes, it is started as systemd service. Finally i found the reason, why it hangs. It is happened because I made blocking syscall from event loop. So this thread can be closed and marked as resolved. Appproach with QSocketNotifier and signalfd works.