[Intel-gfx] [PATCH] drm/i915/bdw: The TLB invalidation mechanism has been removed from INSTPM

[Daniel Vetter]

On Wed, Mar 12, 2014 at 07:21:45PM -0700, Ben Widawsky wrote:
> On Thu, Mar 13, 2014 at 01:40:28AM +0000, Damien Lespiau wrote:
> > While wandering in the spec, I noticed that BDW removes those 2 bits
> > from INSTPM. I couldn't find any direct way to invalidate the TLB (ie
> > without the ring working already). Maybe someone will be more lucky.
> > 
> > At least, we now know we may be a problem.
> > 
> > Signed-off-by: Damien Lespiau <damien.lespiau at intel.com>
> > ---
> >  drivers/gpu/drm/i915/intel_ringbuffer.c | 10 ++++++++--
> >  1 file changed, 8 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/i915/intel_ringbuffer.c b/drivers/gpu/drm/i915/intel_ringbuffer.c
> > index c50388a..4eb3e06 100644
> > --- a/drivers/gpu/drm/i915/intel_ringbuffer.c
> > +++ b/drivers/gpu/drm/i915/intel_ringbuffer.c
> > @@ -981,8 +981,14 @@ void intel_ring_setup_status_page(struct intel_ring_buffer *ring)
> >  	I915_WRITE(mmio, (u32)ring->status_page.gfx_addr);
> >  	POSTING_READ(mmio);
> >  
> > -	/* Flush the TLB for this page */
> > -	if (INTEL_INFO(dev)->gen >= 6) {
> > +	/*
> > +	 * Flush the TLB for this page
> > +	 *
> > +	 * FIXME: These two bits have disappeared on gen8, so a question
> > +	 * arises: do we still need this and if so how should we go about
> > +	 * invalidating the TLB?
> > +	 */
> > +	if (INTEL_INFO(dev)->gen >= 6 && INTEL_INFO(dev)->gen < 8) {
> >  		u32 reg = RING_INSTPM(ring->mmio_base);
> >  
> >  		/* ring should be idle before issuing a sync flush*/
> 
> I'm missing something on the original patch,
> 884020bf3d2a3787a1cc6df902e98e0eec60330b. How were we emitting
> breadcrumbs without flushing the TLB? All bathcbuffers should be
> bookended by a TLB invalidate already, so I'm not sure the logic holds.
> Chris could explain that one a bit further?
> 
> The only reason I bring this up is I'd like to rip this out completely
> and have Thiago retest, or at least change the comment/commit message to
> be to reflect whatever light Chris sheds on the matter.
> 
> Anyway, the bits are definitely gone, and I also can't find a non-ring
> based replacement.
> Reviewed-by: Ben Widawsky <ben at bwidawsk.net>

Queued for -next, thanks for the patch.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch