Re: [RFC/PATCH] virtio-console: wait for console ports

[Amit Shah <amit.shah@redhat.com>]

On (Thu) 22 Sep 2011 [13:20:07], Christian Borntraeger wrote:
> On 22/09/11 12:08, Amit Shah wrote:
> >> +	/* If there was an early virtio console, assume that there are no
> >> +	 * other consoles. We need to wait until the hvc_alloc matches the
> >> +	 * hvc_instantiate, otherwise tty_open will complain, resulting in
> >> +	 * a "Warning: unable to open an initial console" boot failure.
> >> +	 * Without multiport this is done in add_port above. With multiport
> >> +	 * this might take some host<->guest communication - thus we have to
> >> +	 * wait. */
> > 
> > This file uses comments in the form
> >   /*
> >    * ...
> >    */
> 
> Will fix.
> 
> > 
> >> +	if (multiport && early)
> >> +		wait_for_completion(&port_added);
> >> +
> > 
> > Can there be a problem to not timeout this wait?  Maybe it's not a
> > real problem; just thinking out aloud.
> 
> I had the same thoughts. I then asked myself, how big the timeout has to be - and
> the answer was it really depends on the host load. So we can certainly use
> wait_for_completion_timeout(&port_added, HZ*<x>);
> which will work in 99.9% of all cases. It might still cause spurious boot failures,
> if for some reasons it takes too long.

Yes; there's no deterministic way.

> So the big question is, is there a case were virtio is used as an early console
> but virtio_console does not register a console during probe. Ideas?

Currently only ppc and s390 use early_console.  And since you're
sending this patch now, I guess you've started using multiport in the
host recently.  So it's really upto you to decide :-)  I think this is
benign as of now.

		Amit