[MPlayer-users] Re: [XFree86] mga_vid VT switching hangs

[Billy Biggs]

Monty (xiphmont at xiph.org):

> On Wed, Nov 05, 2003 at 08:40:45AM -0600, Billy Biggs wrote:
> > Ryan Underwood (nemesis-lists at icequake.net):
> > 
> > > Just a quick and silly question:  does anyone know if there is a
> > > limitation in mga_vid's design, that causes a dead system when
> > > switching to a VT from X while mplayer is using it?  Of course,
> > > this is one of those things where the obvious solution is "Don't
> > > do that then", but I'm curious if I'm the only one that has that
> > > problem. :)
> > > 
> > > I have a G400MAX running dualhead with Matrox HAL, kernel 2.4.22,
> > > XFree86 4.3.0 on Debian.
> > 
> > When you have two drivers for the same hardware in conflict like
> > that, I would expect problems.  It's pretty well known that this
> > causes hangs.
> 
> Um, what exactly is the conflict here?  That should certainly be a
> safe operation.

  Every time you draw a frame using mga_vid you're reprogramming the
Matrox card with the address of next frame, the overlay parameters, etc.
When you switch VTs, X is also programming the card.  These operations
are not atomic.  You have two drivers both programming the same hardware
resource.  If you think that VT switching should be safe, it's up to you
to put forward a more convincing argument I think :)

  Just consider that mga_vid currently uses the end of video RAM and
just assumes that hopefully X isn't using that for its pixmap cache or
anything, and you'll understand a little bit more of the design.

  -Billy