[Intel-gfx] [PATCH 2/2] drm/i915: do not make assumptions on GGTT VMA sizes

[Chris Wilson]

On Fri, Apr 17, 2015 at 05:00:07PM +0300, Joonas Lahtinen wrote:
> GGTT VMA sizes might be smaller than the whole object size due to
> different GGTT views.
> 
> Signed-off-by: Joonas Lahtinen <joonas.lahtinen at linux.intel.com>
> ---
>  drivers/gpu/drm/i915/i915_gem.c     | 32 +++++++++++++++++++-------------
>  drivers/gpu/drm/i915/i915_gem_gtt.c | 23 +++++++++++++++++++++++
>  drivers/gpu/drm/i915/i915_gem_gtt.h |  4 ++++
>  3 files changed, 46 insertions(+), 13 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index f7b8766..77116bd 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -3507,7 +3507,7 @@ i915_gem_object_bind_to_vm(struct drm_i915_gem_object *obj,
>  {
>  	struct drm_device *dev = obj->base.dev;
>  	struct drm_i915_private *dev_priv = dev->dev_private;
> -	u32 size, fence_size, fence_alignment, unfenced_alignment;
> +	u32 size, vma_size, fence_size, fence_alignment, unfenced_alignment;
>  	unsigned long start =
>  		flags & PIN_OFFSET_BIAS ? flags & PIN_OFFSET_MASK : 0;
>  	unsigned long end =
> @@ -3518,15 +3518,18 @@ i915_gem_object_bind_to_vm(struct drm_i915_gem_object *obj,
>  	if(WARN_ON(i915_is_ggtt(vm) != !!ggtt_view))
>  		return ERR_PTR(-EINVAL);
>  
> +	size = ggtt_view ? i915_ggtt_view_size(obj, ggtt_view) :
> +			   obj->base.size;
> +
>  	fence_size = i915_gem_get_gtt_size(dev,
> -					   obj->base.size,
> +					   size,
>  					   obj->tiling_mode);
>  	fence_alignment = i915_gem_get_gtt_alignment(dev,
> -						     obj->base.size,
> +						     size,
>  						     obj->tiling_mode, true);
>  	unfenced_alignment =
>  		i915_gem_get_gtt_alignment(dev,
> -					   obj->base.size,
> +					   size,
>  					   obj->tiling_mode, false);

This is completely wrong. These calculations are purely about the fence
representation of the whole object, not of the view.

After computing the fence_size etc, you then want to deduce the actual
size to bind, based on partial view, fencing, padded or just normal.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre