[Intel-gfx] [PATCH fixes] drm/i915: Fix unfenced alignment on pre-G33 hardware

[Chris Wilson]

On Mon, 06 Jun 2011 09:03:30 -0700, Keith Packard <keithp at keithp.com> wrote:
> On Mon,  6 Jun 2011 15:18:44 +0100, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> 
> >  	case I915_PARAM_HAS_RELAXED_FENCING:
> > -		value = 1;
> > +		value = 2;
> 
> This looks like a change in ABI to me. I think this means you want a new
> ioctl so that applications using the existing HAS_RELAXED_FENCING ioctl
> and expecting a boolean will continue to work.

Hah. Anyway it is actually irrelevant as it turns out, the kernel is broken
with any per-surface tiling on gen2/gen3.

> >  uint32_t
> > -i915_gem_get_unfenced_gtt_alignment(struct drm_i915_gem_object *obj);
> > +i915_gem_get_unfenced_gtt_alignment(struct drm_i915_gem_object *obj,
> > +				    int tiling_mode);
> ...
> > -				i915_gem_get_unfenced_gtt_alignment(obj);
> > +				i915_gem_get_unfenced_gtt_alignment(obj,
> > +								    args->tiling_mode);
> 
> This combination looks like a simple bug fix -- not using the new tiling
> mode when computing the required alignment. Can you separate out this
> From the power-of-two change?

Yes, I can. But the simple bug fix doesn't fix anything without the other
chunk. Do you still want it split?
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre