On Wed, 2011-04-27 at 17:50 +0200, Mario Kleiner wrote: > On Apr 27, 2011, at 10:58 AM, dri-devel-request@xxxxxxxxxxxxxxxxxxxxx > wrote: > > Message: 5 > > Date: Wed, 27 Apr 2011 10:38:14 +0200 > > From: Michel D?nzer <michel@xxxxxxxxxxx> > > Subject: Re: [PATCH 2/3] drm: Warn if vblank state has become > > inconsistent. > > To: christopher.halse.rogers@xxxxxxxxxxxxx > > Cc: dri-devel@xxxxxxxxxxxxxxxxxxxxx > > Message-ID: <1303893494.5633.129.camel@xxxxxxxxxx> > > Content-Type: text/plain; charset="UTF-8" > > > > On Mit, 2011-04-27 at 16:10 +1000, > > christopher.halse.rogers@xxxxxxxxxxxxx wrote: > >> From: Christopher James Halse Rogers > >> <christopher.halse.rogers@xxxxxxxxxxxxx> > >> > >> After emitting all the waiting vblank events no-one should hold > >> a vblank reference. Emit a warning if this is not the case. > >> > >> Signed-off-by: Christopher James Halse Rogers > >> <christopher.halse.rogers@xxxxxxxxxxxxx> > >> --- > >> drivers/gpu/drm/drm_irq.c | 1 + > >> 1 files changed, 1 insertions(+), 0 deletions(-) > >> > >> diff --git a/drivers/gpu/drm/drm_irq.c b/drivers/gpu/drm/drm_irq.c > >> index a1f12cb..72407fa 100644 > >> --- a/drivers/gpu/drm/drm_irq.c > >> +++ b/drivers/gpu/drm/drm_irq.c > >> @@ -960,6 +960,7 @@ void drm_vblank_off(struct drm_device *dev, > >> int crtc) > >> e->event.sequence); > >> } > >> > >> + WARN_ON(atomic_read(&dev->vblank_refcount[crtc]) != 0); > >> spin_unlock_irqrestore(&dev->vbl_lock, irqflags); > >> } > >> EXPORT_SYMBOL(drm_vblank_off); > > > > Reviewed-by: Michel D?nzer <michel@xxxxxxxxxxx> > > > > > Any pending kms pageflip will also hold a reference on the vblank of > a crtc, so having the refcount non-zero there is not really a sign of > inconsistency, so i'm not sure if a warning is appropriate there. But at this point the vblank interrupts have been disabled, or at least the driver's disable function has been called. Will that not mean that any pending pageflips will wait indefinitely for a vblank interrupt that's not going to come - ie: exactly the state we're warning against?
Attachment:
signature.asc
Description: This is a digitally signed message part
_______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/dri-devel
