drm/ttm: fix leaking fences via ttm_buffer_object_transfer

Set the drm_device to NULL, so that the newly created buffer object
doesn't appear to use the embedded gem object.

This is necessary, because otherwise no corresponding dma_resv_fini for
the dma_resv_init is called, resulting in a memory leak.

The dma_resv_fini in ttm_bo_release_list is only called if the embedded
gem object is not used, which is determined by checking if the
drm_device is NULL.

Bug: https://gitlab.freedesktop.org/drm/amd/issues/958
Fixes: 1e053b10ba60 ("drm/ttm: use gem reservation object")
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Ahzo <Ahzo@tutanota.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/355089/

diff --git a/drivers/gpu/drm/ttm/ttm_bo_util.c b/drivers/gpu/drm/ttm/ttm_bo_util.c
index 49ed55779128e3cf5954041499fc534e20fa2dfd..953c82a4f5736afc22484a14296976319aa1a5e0 100644 (file)
--- a/drivers/gpu/drm/ttm/ttm_bo_util.c
+++ b/drivers/gpu/drm/ttm/ttm_bo_util.c
@@ -515,6 +515,7 @@ static int ttm_buffer_object_transfer(struct ttm_buffer_object *bo,
                fbo->base.base.resv = &fbo->base.base._resv;
 
        dma_resv_init(&fbo->base.base._resv);
+       fbo->base.base.dev = NULL;
        ret = dma_resv_trylock(&fbo->base.base._resv);
        WARN_ON(!ret);