> > > tmp = I915_READ(GEN8_GT_IIR(0));
> > > if (tmp) {
> > > ret = IRQ_HANDLED;
> > > +
> > > rcs = tmp >> GEN8_RCS_IRQ_SHIFT;
> > > - bcs = tmp >> GEN8_BCS_IRQ_SHIFT;
> > > + ring = &dev_priv->ring[RCS];
> > > if (rcs & GT_RENDER_USER_INTERRUPT)
> > > - notify_ring(dev, &dev_priv->ring[RCS]);
> > > + notify_ring(dev, ring);
> > > + if (rcs & GEN8_GT_CONTEXT_SWITCH_INTERRUPT)
> > > + gen8_handle_context_events(ring);
> >
> > Handling the context events here can generate a new execlist submission,
> > which if a small enough workload, can finish and generate a new context event
> > interrupt before we ack this interrupt.
> >
> > When we ack this interrupt, we clear the new one too, loosing an interrupt.
> >
> > Moving the
> >
> > I915_WRITE(GEN8_GT_IIR(0), tmp);
> >
> > to just inside the if (tmp) { conditional (or anywhere before this call) fixes this
> > issue. There is no harm in acking the interrupt immediately as we have the
> > read stored in tmp.
> >
> -----Original Message-----
> From: Daniel, Thomas
> Sent: Monday, April 28, 2014 10:58 AM
> To: Beckett, Robert; Mateo Lozano, Oscar; Barbalho, Rafael; Ewins, Jon
> Subject: RE: Re: [PATCH 43/49] drm/i915/bdw: Handle context switch
> events
>
> Hi Bob,
>
> Looks like a good catch, and a sensible fix.
>
> Thomas.
I agree with Thomas. Will add to the next revision of the series.
Thanks!
Oscar
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx
