drm/amdgpu/ttm: handle tt moves properly.
authorDave Airlie <airlied@redhat.com>
Thu, 24 Sep 2020 05:18:03 +0000 (15:18 +1000)
committerDave Airlie <airlied@redhat.com>
Thu, 24 Sep 2020 19:48:00 +0000 (05:48 +1000)
The core move code currently handles use_tt moves, for amdgpu
this was being handled also in the driver, but not using the same
paths.

If moving between TT/SYSTEM (all the use_tt paths on amdgpu) use
the core move function.

Eventually the core will be flipped over to calling the driver.

Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200924051845.397177-4-airlied@gmail.com
drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c

index db5f761..d3bd2fd 100644 (file)
@@ -671,14 +671,16 @@ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict,
                ttm_bo_move_null(bo, new_mem);
                return 0;
        }
-       if ((old_mem->mem_type == TTM_PL_TT &&
-            new_mem->mem_type == TTM_PL_SYSTEM) ||
-           (old_mem->mem_type == TTM_PL_SYSTEM &&
-            new_mem->mem_type == TTM_PL_TT)) {
-               /* bind is enough */
+       if (old_mem->mem_type == TTM_PL_SYSTEM &&
+           new_mem->mem_type == TTM_PL_TT) {
                ttm_bo_move_null(bo, new_mem);
                return 0;
        }
+
+       if (old_mem->mem_type == TTM_PL_TT &&
+           new_mem->mem_type == TTM_PL_SYSTEM)
+               return ttm_bo_move_ttm(bo, ctx, new_mem);
+
        if (old_mem->mem_type == AMDGPU_PL_GDS ||
            old_mem->mem_type == AMDGPU_PL_GWS ||
            old_mem->mem_type == AMDGPU_PL_OA ||