[PATCH 38/73] drm/amd/display: Fix warnings on S3 resume

[Harry Wentland]

From: "Leo (Sunpeng) Li" <sunpeng.li at amd.com>

This is a followup to the following revert:

Rex Zhu    Revert "drm/amd/display: Match actual state during S3
           resume."

Three things needed to be addressed:

1. Potential memory leak on dc_state creation in atomic_check during
   s3 resume
2. Warnings are now seen in dmesg during S3 resume
3. Since dc_state is now created in atomic_check, what the reverted
   patch was addressing needs to be reevaluated.

This change addresses the above:

1. Since the suspend procedure calls drm_atomic_state_clear, our hook
   for releasing the dc_state is called. This frees it before
   atomic_check creates it during resume. The leak does not occur.

2. The dc_crtc/plane_state references kept by the atomic states need to
   be released before calling atomic_check, which warns if they are
   non-null. This is because atomic_check is responsible for creating
   the dc_*_states. This is a special case for S3 resume, since the
   atomic state duplication that occurs during suspend also copies a
   reference to the dc_*_states.

3. See 2. comments are also updated to reflect this.

Change-Id: I6e342bf8134f0e5dc32888a8d894c2cd20d28296
Signed-off-by: Leo (Sunpeng) Li <sunpeng.li at amd.com>
Reviewed-by: Harry Wentland <Harry.Wentland at amd.com>
---
 drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 28 +++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
index 1c7f22146bc9..bdef1ed0dfac 100644
--- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
+++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c
@@ -643,6 +643,11 @@ int amdgpu_dm_display_resume(struct amdgpu_device *adev)
 	struct drm_connector *connector;
 	struct drm_crtc *crtc;
 	struct drm_crtc_state *new_crtc_state;
+	struct dm_crtc_state *dm_new_crtc_state;
+	struct drm_plane *plane;
+	struct drm_plane_state *new_plane_state;
+	struct dm_plane_state *dm_new_plane_state;
+
 	int ret = 0;
 	int i;
 
@@ -685,6 +690,29 @@ int amdgpu_dm_display_resume(struct amdgpu_device *adev)
 	for_each_new_crtc_in_state(adev->dm.cached_state, crtc, new_crtc_state, i)
 		new_crtc_state->active_changed = true;
 
+	/*
+	 * atomic_check is expected to create the dc states. We need to release
+	 * them here, since they were duplicated as part of the suspend
+	 * procedure.
+	 */
+	for_each_new_crtc_in_state(adev->dm.cached_state, crtc, new_crtc_state, i) {
+		dm_new_crtc_state = to_dm_crtc_state(new_crtc_state);
+		if (dm_new_crtc_state->stream) {
+			WARN_ON(kref_read(&dm_new_crtc_state->stream->refcount) > 1);
+			dc_stream_release(dm_new_crtc_state->stream);
+			dm_new_crtc_state->stream = NULL;
+		}
+	}
+
+	for_each_new_plane_in_state(adev->dm.cached_state, plane, new_plane_state, i) {
+		dm_new_plane_state = to_dm_plane_state(new_plane_state);
+		if (dm_new_plane_state->dc_state) {
+			WARN_ON(kref_read(&dm_new_plane_state->dc_state->refcount) > 1);
+			dc_plane_state_release(dm_new_plane_state->dc_state);
+			dm_new_plane_state->dc_state = NULL;
+		}
+	}
+
 	ret = drm_atomic_helper_resume(ddev, adev->dm.cached_state);
 
 	drm_atomic_state_put(adev->dm.cached_state);
-- 
2.14.1