[Intel-gfx] [PATCH 35/37] drm/radeon: fix fence locking in the pageflip callback

[Daniel Vetter]

We need to hold bdev->fence_lock while grabbing a reference to
the fence, to prevent concurrent clearing/changing of the
ttm_bo->sync_obj field.

Signed-off-by: Daniel Vetter <daniel.vetter at ffwll.ch>
---
 drivers/gpu/drm/radeon/radeon_display.c |    4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c
index 8724196..069d5cc 100644
--- a/drivers/gpu/drm/radeon/radeon_display.c
+++ b/drivers/gpu/drm/radeon/radeon_display.c
@@ -378,8 +378,12 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc,
 	work->old_rbo = rbo;
 	obj = new_radeon_fb->obj;
 	rbo = gem_to_radeon_bo(obj);
+
+	spin_lock(&rbo->tbo.bdev->fence_lock);
 	if (rbo->tbo.sync_obj)
 		work->fence = radeon_fence_ref(rbo->tbo.sync_obj);
+	spin_unlock(&rbo->tbo.bdev->fence_lock);
+
 	INIT_WORK(&work->work, radeon_unpin_work_func);
 
 	/* We borrow the event spin lock for protecting unpin_work */
-- 
1.7.10.4