Re: [PATCH 2/2] Revert "drm/i915: Workaround incoherence between fences and LLC across multiple CPUs"

[Daniel Vetter <daniel@ffwll.ch>]

On Tue, Jul 09, 2013 at 05:54:40PM +0100, Chris Wilson wrote:
> This reverts commit 25ff119 and the follow on for Valleyview commit 2dc8aae.
> 
> commit 25ff1195f8a0b3724541ae7bbe331b4296de9c06
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Thu Apr 4 21:31:03 2013 +0100
> 
>     drm/i915: Workaround incoherence between fences and LLC across multiple CPUs
> 
> commit 2dc8aae06d53458dd3624dc0accd4f81100ee631
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Wed May 22 17:08:06 2013 +0100
> 
>     drm/i915: Workaround incoherence with fence updates on Valleyview
> 
> Jon Bloomfield came up with a plausible explanation and cheap fix for the
> race condition, so lets run with it.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
> Cc: Jon Bloomfield <jon.bloomfield@intel.com>
> Cc: Carsten Emde<C.Emde@osadl.org>
> Cc: stable@vger.kernel.org

Can you please add  reference for the various bug reports we've gotten
from the realtime linux folks? Also we have some bug report from QA iirc
about a perf regression on UXA due to this.
-Daniel

> ---
>  drivers/gpu/drm/i915/i915_gem.c | 47 ++++-------------------------------------
>  1 file changed, 4 insertions(+), 43 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index ce46e777..96ede8c 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -2924,56 +2924,17 @@ static inline int fence_number(struct drm_i915_private *dev_priv,
>  	return fence - dev_priv->fence_regs;
>  }
>  
> -struct write_fence {
> -	struct drm_device *dev;
> -	struct drm_i915_gem_object *obj;
> -	int fence;
> -};
> -
> -static void i915_gem_write_fence__ipi(void *data)
> -{
> -	struct write_fence *args = data;
> -
> -	/* Required for SNB+ with LLC */
> -	wbinvd();
> -
> -	/* Required for VLV */
> -	i915_gem_write_fence(args->dev, args->fence, args->obj);
> -}
> -
>  static void i915_gem_object_update_fence(struct drm_i915_gem_object *obj,
>  					 struct drm_i915_fence_reg *fence,
>  					 bool enable)
>  {
>  	struct drm_i915_private *dev_priv = obj->base.dev->dev_private;
> -	struct write_fence args = {
> -		.dev = obj->base.dev,
> -		.fence = fence_number(dev_priv, fence),
> -		.obj = enable ? obj : NULL,
> -	};
> -
> -	/* In order to fully serialize access to the fenced region and
> -	 * the update to the fence register we need to take extreme
> -	 * measures on SNB+. In theory, the write to the fence register
> -	 * flushes all memory transactions before, and coupled with the
> -	 * mb() placed around the register write we serialise all memory
> -	 * operations with respect to the changes in the tiler. Yet, on
> -	 * SNB+ we need to take a step further and emit an explicit wbinvd()
> -	 * on each processor in order to manually flush all memory
> -	 * transactions before updating the fence register.
> -	 *
> -	 * However, Valleyview complicates matter. There the wbinvd is
> -	 * insufficient and unlike SNB/IVB requires the serialising
> -	 * register write. (Note that that register write by itself is
> -	 * conversely not sufficient for SNB+.) To compromise, we do both.
> -	 */
> -	if (INTEL_INFO(args.dev)->gen >= 6)
> -		on_each_cpu(i915_gem_write_fence__ipi, &args, 1);
> -	else
> -		i915_gem_write_fence(args.dev, args.fence, args.obj);
> +	int reg = fence_number(dev_priv, fence);
> +
> +	i915_gem_write_fence(obj->base.dev, reg, enable ? obj : NULL);
>  
>  	if (enable) {
> -		obj->fence_reg = args.fence;
> +		obj->fence_reg = reg;
>  		fence->obj = obj;
>  		list_move_tail(&fence->lru_list, &dev_priv->mm.fence_list);
>  	} else {
> -- 
> 1.8.3.2
> 

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