[PATCH v2 1/2] radeon: Don't generate new fence for page flip.

[Michel Dänzer]

From: Michel Dänzer <michel.daenzer at amd.com>

Use the fence of the new frontbuffer, if any.

Generating a new fence could cause us to wait for completely unrelated
rendering to finish before performing the flip.

Signed-off-by: Michel Dänzer <michel.daenzer at amd.com>
---

v2:

* Rebase against drm-core-next.
* Leave radeon_fence_unref() call in error path, or we leak the fence there.

 drivers/gpu/drm/radeon/radeon_display.c |   33 +++++-------------------------
 1 files changed, 6 insertions(+), 27 deletions(-)

diff --git a/drivers/gpu/drm/radeon/radeon_display.c b/drivers/gpu/drm/radeon/radeon_display.c
index 292f73f..ffce85a 100644
--- a/drivers/gpu/drm/radeon/radeon_display.c
+++ b/drivers/gpu/drm/radeon/radeon_display.c
@@ -282,7 +282,7 @@ void radeon_crtc_handle_flip(struct radeon_device *rdev, int crtc_id)
 	spin_lock_irqsave(&rdev->ddev->event_lock, flags);
 	work = radeon_crtc->unpin_work;
 	if (work == NULL ||
-	    !radeon_fence_signaled(work->fence)) {
+	    (work->fence && !radeon_fence_signaled(work->fence))) {
 		spin_unlock_irqrestore(&rdev->ddev->event_lock, flags);
 		return;
 	}
@@ -348,7 +348,6 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc,
 	struct radeon_framebuffer *new_radeon_fb;
 	struct drm_gem_object *obj;
 	struct radeon_bo *rbo;
-	struct radeon_fence *fence;
 	struct radeon_unpin_work *work;
 	unsigned long flags;
 	u32 tiling_flags, pitch_pixels;
@@ -359,16 +358,9 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc,
 	if (work == NULL)
 		return -ENOMEM;
 
-	r = radeon_fence_create(rdev, &fence);
-	if (unlikely(r != 0)) {
-		kfree(work);
-		DRM_ERROR("flip queue: failed to create fence.\n");
-		return -ENOMEM;
-	}
 	work->event = event;
 	work->rdev = rdev;
 	work->crtc_id = radeon_crtc->crtc_id;
-	work->fence = radeon_fence_ref(fence);
 	old_radeon_fb = to_radeon_framebuffer(crtc->fb);
 	new_radeon_fb = to_radeon_framebuffer(fb);
 	/* schedule unpin of the old buffer */
@@ -377,6 +369,10 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc,
 	drm_gem_object_reference(obj);
 	rbo = gem_to_radeon_bo(obj);
 	work->old_rbo = rbo;
+	obj = new_radeon_fb->obj;
+	rbo = gem_to_radeon_bo(obj);
+	if (rbo->tbo.sync_obj)
+		work->fence = radeon_fence_ref(rbo->tbo.sync_obj);
 	INIT_WORK(&work->work, radeon_unpin_work_func);
 
 	/* We borrow the event spin lock for protecting unpin_work */
@@ -391,9 +387,6 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc,
 	spin_unlock_irqrestore(&dev->event_lock, flags);
 
 	/* pin the new buffer */
-	obj = new_radeon_fb->obj;
-	rbo = gem_to_radeon_bo(obj);
-
 	DRM_DEBUG_DRIVER("flip-ioctl() cur_fbo = %p, cur_bbo = %p\n",
 			 work->old_rbo, rbo);
 
@@ -461,25 +454,11 @@ static int radeon_crtc_page_flip(struct drm_crtc *crtc,
 		goto pflip_cleanup1;
 	}
 
-	/* 32 ought to cover us */
-	r = radeon_ring_lock(rdev, 32);
-	if (r) {
-		DRM_ERROR("failed to lock the ring before flip\n");
-		goto pflip_cleanup2;
-	}
-
-	/* emit the fence */
-	radeon_fence_emit(rdev, fence);
 	/* set the proper interrupt */
 	radeon_pre_page_flip(rdev, radeon_crtc->crtc_id);
-	/* fire the ring */
-	radeon_ring_unlock_commit(rdev);
 
 	return 0;
 
-pflip_cleanup2:
-	drm_vblank_put(dev, radeon_crtc->crtc_id);
-
 pflip_cleanup1:
 	r = radeon_bo_reserve(rbo, false);
 	if (unlikely(r != 0)) {
@@ -501,7 +480,7 @@ pflip_cleanup:
 unlock_free:
 	drm_gem_object_unreference_unlocked(old_radeon_fb->obj);
 	spin_unlock_irqrestore(&dev->event_lock, flags);
-	radeon_fence_unref(&fence);
+	radeon_fence_unref(&work->fence);
 	kfree(work);
 
 	return r;
-- 
1.7.5.4