Created attachment 298309 [details] [review]
worker: drop inaccurate FIXME

I think the problem is really just the terminal getting
stuck in KD_GRAPHICS/VT_AUTO mode by plymouth.

Created attachment 298310 [details] [review]
worker: split VT_SETMODE handling to helper function

We're going to call it in more than one place in the code,
so it's better to keep it in a separate function.

Created attachment 298311 [details] [review]
worker: don't use SIGUSR1 for VT switch signal

That signal is blocked when the worker is forked,
because the main daemon uses it to notice when
the X server is ready.

This commit changes the worker to use real-time
signals instead, which are better anyway, since
they queue.

Created attachment 298312 [details] [review]
worker: fix hang in VT_WAITACTIVE

When plymouth deactivates it correctly leaves the display in
GRAPHICS mode but incorrectly stops processing VT changes.  This
combination of VT_AUTO + KD_GRAPHICS makes VT switches hang.

Deal with the situation on our side until plymouth is fixed.

Attachment 298309 [details] pushed as 19f169e - worker: drop inaccurate FIXME
Attachment 298310 [details] pushed as 69847b9 - worker: split VT_SETMODE handling to helper function
Attachment 298311 [details] pushed as 2b02aad - worker: don't use SIGUSR1 for VT switch signal
Attachment 298312 [details] pushed as fe8b113 - worker: fix hang in VT_WAITACTIVE