drm/nouveau/svm: Fix refcount leak bug and missing check against null bug

The reference counting issue happens in one exception handling path of
nouveau_svmm_bind(). When cli->svm.svmm is null, the function forgets
to decrease the refcount of mm increased by get_task_mm(), causing a
refcount leak.

Fix this issue by using mmput() to decrease the refcount in the
exception handling path.

Also, the function forgets to do check against null when get mm
by get_task_mm().

Fix this issue by adding null check after get mm by get_task_mm().

Signed-off-by: Chenyuan Mi <cymi20@fudan.edu.cn>
Signed-off-by: Xiyu Yang <xiyuyang19@fudan.edu.cn>
Signed-off-by: Xin Tan <tanxin.ctf@gmail.com>
Fixes: 822cab6150d3 ("drm/nouveau/svm: check for SVM initialized before migrating")
Reviewed-by: Lyude Paul <lyude@redhat.com>
Reviewed-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210907122633.16665-1-cymi20@fudan.edu.cn
Link: https://gitlab.freedesktop.org/drm/nouveau/-/merge_requests/14

1 files changed, 4 insertions, 0 deletions

diff --git a/drivers/gpu/drm/nouveau/nouveau_svm.c b/drivers/gpu/drm/nouveau/nouveau_svm.c
index b0c3422cb01f..9985bfde015a 100644
--- a/drivers/gpu/drm/nouveau/nouveau_svm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_svm.c
@@ -162,10 +162,14 @@ nouveau_svmm_bind(struct drm_device *dev, void *data,
 	 */
 
 	mm = get_task_mm(current);
+	if (!mm) {
+		return -EINVAL;
+	}
 	mmap_read_lock(mm);
 
 	if (!cli->svm.svmm) {
 		mmap_read_unlock(mm);
+		mmput(mm);
 		return -EINVAL;
 	}