[Intel-gfx] [PATCH] drm/i915: prevent out of range pt in the PDE macros (take 3)

[Chris Wilson]

On Fri, Oct 02, 2015 at 01:47:03PM +0100, Michel Thierry wrote:
> On 10/1/2015 5:09 PM, Chris Wilson wrote:
> >On Thu, Oct 01, 2015 at 04:59:35PM +0100, Michel Thierry wrote:
> >>---
> >>  drivers/gpu/drm/i915/i915_gem_gtt.h | 14 ++++++++++----
> >>  1 file changed, 10 insertions(+), 4 deletions(-)
> >>
> >>diff --git a/drivers/gpu/drm/i915/i915_gem_gtt.h b/drivers/gpu/drm/i915/i915_gem_gtt.h
> >>index 9fbb07d..94f8344 100644
> >>--- a/drivers/gpu/drm/i915/i915_gem_gtt.h
> >>+++ b/drivers/gpu/drm/i915/i915_gem_gtt.h
> >>@@ -394,7 +394,9 @@ struct i915_hw_ppgtt {
> >>   */
> >>  #define gen6_for_each_pde(pt, pd, start, length, temp, iter) \
> >>  	for (iter = gen6_pde_index(start); \
> >>-	     pt = (pd)->page_table[iter], length > 0 && iter < I915_PDES; \
> >>+	     pt = (length > 0 && iter < I915_PDES) ? \
> >>+			(pd)->page_table[iter] : NULL, \
> >>+	     length > 0 && iter < I915_PDES; \
> >
> >length > 0 && iter < I915_PDES ? (pt = (pd)->page_table[iter]) : 0,
> >
> >as the compiler wouldn't be able to CSE it otherwise (I think).
> 
> Even after that change, the compiler keeps doing an optimization
> when page_table[iter] is null (takes the null assignment as the
> break condition).
> 
> I've been playing with these examples
> http://paste.ubuntu.com/12638106/
> 
> Only the 1st example (a) iterates over all elements, b & c stop
> after the 1st run.

Forgot that was the condition you wanted to change.

length > 0 && iter < I915_PDES ? (pt = (pd)->page_table[iter]), 1 : 0,

Would be nice to prove that length > 0 implies iter < I915_PDES. If only
we had smart tools :)
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre