[Intel-gfx] [PATCH] drm/i915: Correctly refcount connectors in hw state readou

Chris Wilson chris at chris-wilson.co.uk
Fri May 6 14:19:22 UTC 2016


On Fri, May 06, 2016 at 05:07:09PM +0300, Ville Syrjälä wrote:
> On Fri, May 06, 2016 at 03:39:03PM +0200, Daniel Vetter wrote:
> > This was forgotten when adding the the refcounting to
> > drm_connector_state.
> > 
> > v2: Don't forget to unreference existing connectors. This isn't
> > relevant on driver load, but this code also runs on resume, and there
> > we already have an atomic state. Spotted by Chris Wilson.
> > 
> > Cc: Gabriel Feceoru <gabriel.feceoru at intel.com>
> > Cc: Chris Wilson <chris at chris-wilson.co.uk>
> > Cc: Marius Vlad <marius.c.vlad at intel.com>
> > Cc: Dave Airlie <airlied at redhat.com>
> > Fixes: d2307dea14a4 ("drm/atomic: use connector references (v3)")
> > Signed-off-by: Daniel Vetter <daniel.vetter at intel.com>
> > ---
> >  drivers/gpu/drm/i915/intel_display.c | 5 +++++
> >  1 file changed, 5 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c
> > index 45c218db86be..6ba676f81bae 100644
> > --- a/drivers/gpu/drm/i915/intel_display.c
> > +++ b/drivers/gpu/drm/i915/intel_display.c
> > @@ -12023,11 +12023,16 @@ static void intel_modeset_update_connector_atomic_state(struct drm_device *dev)
> >  	struct intel_connector *connector;
> >  
> >  	for_each_intel_connector(dev, connector) {
> > +		if (connector->base.state->crtc)
> > +			drm_connector_unreference(&connector->base);
> > +
> >  		if (connector->base.encoder) {
> >  			connector->base.state->best_encoder =
> >  				connector->base.encoder;
> >  			connector->base.state->crtc =
> >  				connector->base.encoder->crtc;
> > +
> > +			drm_connector_reference(&connector->base);
> 
> Is it safe to unref before ref?

In this case, I believe we can assert that we hold a reference to the
connector (as part of the driver state) and the atomic state holds a
reference, i.e. the reference count here should be at least 2. If it was
freed, we would get a warn from kref as well as memdebug (if enabled).
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre


More information about the dri-devel mailing list