Quoting Matthew Auld (2020-12-23 17:19:41)
> The shadow batch is an internal object, which doesn't have any page
> clearing, and since the batch_len might be within a page, we should take
> care to clear any unused dwords before making it visible in the ppGTT.
>
> Testcase: igt/gen9_exec_parse/shadow-peek
> Signed-off-by: Matthew Auld <matthew.auld@xxxxxxxxx>
> ---
> drivers/gpu/drm/i915/i915_cmd_parser.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
> index 8d88402387bd..6506ed30b444 100644
> --- a/drivers/gpu/drm/i915/i915_cmd_parser.c
> +++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
> @@ -1139,6 +1139,7 @@ static u32 *copy_batch(struct drm_i915_gem_object *dst_obj,
> struct drm_i915_gem_object *src_obj,
> unsigned long offset, unsigned long length)
> {
> + unsigned long orig_length = length;
Hmm. This is too short, since this is shorter than the vma->size we
retrieved from the pool. It does feel like we need to clear first.
Anyway well spotted, and this should be cc'ed stable when we push.
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
