[Intel-gfx] [PATCH 1/3] drm/i915: Retire the VMA's fence tracker before unbinding

[Tvrtko Ursulin]

On 20/06/2017 13:43, Chris Wilson wrote:
> Since we may track unfenced access (GPU access to the vma that
> explicitly requires no fence), vma->last_fence may be set without any

Is this the gen < 4 code path? There is no actual fence in this case?

> attached fence (vma->fence) and so will not be flushed when we call
> i915_vma_put_fence(). Since we stopped doing a full retire of the
> activity trackers for unbind, we need to explicitly retire each tracker.
> 
> Fixes: b0decaf75bd9 ("drm/i915: Track active vma requests")
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com>
> ---
>   drivers/gpu/drm/i915/i915_vma.c | 5 +++++
>   1 file changed, 5 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/i915_vma.c b/drivers/gpu/drm/i915/i915_vma.c
> index 532c709febbd..1cfe137cdc32 100644
> --- a/drivers/gpu/drm/i915/i915_vma.c
> +++ b/drivers/gpu/drm/i915/i915_vma.c
> @@ -672,6 +672,11 @@ int i915_vma_unbind(struct i915_vma *vma)
>   				break;
>   		}
>   
> +		if (!ret) {
> +			ret = i915_gem_active_retire(&vma->last_fence,
> +						     &vma->vm->i915->drm.struct_mutex);
> +		}
> +
>   		__i915_vma_unpin(vma);
>   		if (ret)
>   			return ret;
> 

Looks safe anyway, but I'd like to understand how exactly it happens and 
if it is still possible after patch 2/3.

Regards,

Tvrtko