Previously, we assumed that allow_modeset=false => page flip. This
assumption breaks when an atomic commit is submitted with allow_modeset
set to false, since the legacy flip code is never called (the legacy
code grabs the vblank reference).
Fix: Move drm_vblank_get() from amdgpu_atomic_helper_page_flip() to
amdgpu_dm_commit_surfaces().
Signed-off-by: Leo (Sunpeng) Li <sunpeng.li@amd.com>
Reviewed-by: Andrey Grodzovsky <Andrey.Grodzovsky@amd.com>
Acked-by: Harry Wentland <Harry.Wentland@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
                        if (!con_state)
                                continue;
 
-
-
                        add_surface(dm->dc, crtc, plane,
                                    &dc_surfaces_constructed[planes_count]);
                        if (dc_surfaces_constructed[planes_count] == NULL) {
                                acrtc_attach->flip_flags & DRM_MODE_PAGE_FLIP_ASYNC ?
                                false : true;
 
+                       /* TODO: Needs rework for multiplane flip */
+                       if (plane->type == DRM_PLANE_TYPE_PRIMARY)
+                               drm_crtc_vblank_get(crtc);
+
                        amdgpu_dm_do_flip(
                                crtc,
                                fb,