[PATCH] drm: make sure drm_vblank_init() has been called before touching vbl_lock

[Michel Dänzer]

On Don, 2011-07-07 at 09:10 +1000, Ben Skeggs wrote: 
> On Wed, 2011-07-06 at 09:38 +0200, Michel Dänzer wrote:
> > On Mon, 2011-07-04 at 13:16 +1000, skeggsb at gmail.com wrote:
> > > From: Ben Skeggs <bskeggs at redhat.com>
> > > 
> > > Signed-off-by: Ben Skeggs <bskeggs at redhat.com>
> > > ---
> > >  drivers/gpu/drm/drm_irq.c |   15 +++++++++------
> > >  1 files changed, 9 insertions(+), 6 deletions(-)
> > > 
> > > diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c
> > > index 2022a5c..c3c87a1 100644
> > > --- a/drivers/gpu/drm/drm_irq.c
> > > +++ b/drivers/gpu/drm/drm_irq.c
> > > @@ -397,13 +397,16 @@ int drm_irq_uninstall(struct drm_device *dev)
> > >  	/*
> > >  	 * Wake up any waiters so they don't hang.
> > >  	 */
> > > -	spin_lock_irqsave(&dev->vbl_lock, irqflags);
> > > -	for (i = 0; i < dev->num_crtcs; i++) {
> > > -		DRM_WAKEUP(&dev->vbl_queue[i]);
> > > -		dev->vblank_enabled[i] = 0;
> > > -		dev->last_vblank[i] = dev->driver->get_vblank_counter(dev, i);
> > > +	if (dev->num_crtcs) {
> > > +		spin_lock_irqsave(&dev->vbl_lock, irqflags);
> > > +		for (i = 0; i < dev->num_crtcs; i++) {
> > > +			DRM_WAKEUP(&dev->vbl_queue[i]);
> > > +			dev->vblank_enabled[i] = 0;
> > > +			dev->last_vblank[i] =
> > > +				dev->driver->get_vblank_counter(dev, i);
> > > +		}
> > > +		spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
> > >  	}
> > > -	spin_unlock_irqrestore(&dev->vbl_lock, irqflags);
> > >  
> > >  	if (!irq_enabled)
> > >  		return -EINVAL;
> > 
> > Makes sense, but according to commit
> > 7a1c2f6c8d8485af5ac6c2a313f6a7162207a4af ('vmwgfx: Enable use of the
> > vblank system'), the vblank ioctls will also access uninitialized data
> > if drm_vblank_init() hasn't been called. So, is this just for cases
> > where drm_irq_uninstall() is called before the driver gets a chance to
> > call drm_vblank_init()? 
> Heh, ouch.
> 
> Yes, it was specifically that case I was aiming for however.  If
> nouveau's running on a board without any outputs, it won't call
> drm_vblank_init(), but it *will* call drm_irq_*.  Just avoiding lockdep
> complaints on module unload.

I see.

Reviewed-by: Michel Dänzer <michel at daenzer.net>


-- 
Earthling Michel Dänzer           |                   http://www.amd.com
Libre software enthusiast         |          Debian, X and DRI developer