drm/nouveau: prime: fix refcount underflow

Calling nouveau_bo_ref() on a nouveau_bo without initializing it (and
hence the backing ttm_bo) leads to a refcount underflow.

Instead of calling nouveau_bo_ref() in the unwind path of
drm_gem_object_init(), clean things up manually.

Fixes: ab9ccb96a6e6 ("drm/nouveau: use prime helpers")
Reviewed-by: Ben Skeggs <bskeggs@nvidia.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Danilo Krummrich <dakr@kernel.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20240718165959.3983-2-dakr@kernel.org
(cherry picked from commit 1b93f3e89d03cfc576636e195466a0d728ad8de5)
Signed-off-by: Danilo Krummrich <dakr@kernel.org>

diff --git a/drivers/gpu/drm/nouveau/nouveau_prime.c b/drivers/gpu/drm/nouveau/nouveau_prime.c
index b58ab595faf8278c661ce2c9988c09c98f47a515..cd95446d6851113995f8723cdd3a8a022d05f356 100644 (file)
--- a/drivers/gpu/drm/nouveau/nouveau_prime.c
+++ b/drivers/gpu/drm/nouveau/nouveau_prime.c
@@ -64,7 +64,8 @@ struct drm_gem_object *nouveau_gem_prime_import_sg_table(struct drm_device *dev,
         * to the caller, instead of a normal nouveau_bo ttm reference. */
        ret = drm_gem_object_init(dev, &nvbo->bo.base, size);
        if (ret) {
-               nouveau_bo_ref(NULL, &nvbo);
+               drm_gem_object_release(&nvbo->bo.base);
+               kfree(nvbo);
                obj = ERR_PTR(-ENOMEM);
                goto unlock;
        }