On la, 2017-02-25 at 23:25 +0000, Chris Wilson wrote:
> As we track whether a vma has been inserted into the drm_mm using the
> vma->flags, if we fail to bind the vma into the GTT we do not update
> those bits and will attempt to reinsert the vma into the drm_mm on
> future passes. To prevent that, we want to unwind i915_vma_insert() if
> we fail in our attempt to bind.
>
> Fixes: 59bfa1248e22 ("drm/i915: Start passing around i915_vma from execbuffer")
> Testcase: igt/drv_selftest/live_gtt
> Signed-off-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
> Cc: Matthew Auld <matthew.william.auld@xxxxxxxxx>
> Cc: Joonas Lahtinen <joonas.lahtinen@xxxxxxxxxxxxxxx>
> Cc: <stable@xxxxxxxxxxxxxxx> # v4.9+
Reviewed-by: Joonas Lahtinen <joonas.lahtinen@xxxxxxxxxxxxxxx>
One note below.
> @@ -541,7 +566,12 @@ int __i915_vma_do_pin(struct i915_vma *vma,
> GEM_BUG_ON(i915_vma_misplaced(vma, size, alignment, flags));
> return 0;
>
> -err:
> +err_remove:
> + if ((vma->flags & I915_VMA_BIND_MASK) == 0) {
This condition could be more symmetric.
Regards, Joonas
> + GEM_BUG_ON(vma->pages);
> + i915_vma_remove(vma);
> + }
> +err_unpin:
> __i915_vma_unpin(vma);
> return ret;
> }
--
Joonas Lahtinen
Open Source Technology Center
Intel Corporation