drm/i915: Unset legacy_cursor_update early in intel_atomic_commit, v3.

Commit b44d5c0c ("drm/i915: Always wait for flip_done, v2.") removed
the call to wait_for_vblanks and replaced it with flip_done.

Unfortunately legacy_cursor_update was unset too late, and the
replacement call drm_atomic_helper_wait_for_flip_done() was
a noop. Make sure that its unset before setup_commit() is
called to fix this issue.

Changes since v1:
- Force vblank wait for watermarks not yet converted to atomic too. (Ville)
- Use for_each_new_intel_crtc_in_state. (Ville)
Changes since v2:
- Move the optimization to a separate commit. (Ville)
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Fixes: b44d5c0c ("drm/i915: Always wait for flip_done, v2.")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=102675
Testcase: kms_cursor_crc
Cc: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Reported-by: Marta Löfstedt <marta.lofstedt@intel.com>
Cc: Marta Löfstedt <marta.lofstedt@intel.com>
Tested-by: Marta Löfstedt <marta.lofstedt@intel.com>
Cc: Ville Syrjälä <ville.syrjala@linux.intel.com>
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20170919121419.13708-1-maarten.lankhorst@linux.intel.com

drivers/gpu/drm/i915/intel_display.c

@@ -12541,21 +12541,10 @@ static int intel_atomic_commit(struct drm_device *dev,
 	struct drm_i915_private *dev_priv = to_i915(dev);
 	int ret = 0;
 
-	ret = drm_atomic_helper_setup_commit(state, nonblock);
-	if (ret)
-		return ret;
-
 	drm_atomic_state_get(state);
 	i915_sw_fence_init(&intel_state->commit_ready,
 			   intel_atomic_commit_ready);
 
-	ret = intel_atomic_prepare_commit(dev, state);
-	if (ret) {
-		DRM_DEBUG_ATOMIC("Preparing state failed with %i\n", ret);
-		i915_sw_fence_commit(&intel_state->commit_ready);
-		return ret;
-	}
-
 	/*
 	 * The intel_legacy_cursor_update() fast path takes care
 	 * of avoiding the vblank waits for simple cursor
@@ -12564,11 +12553,11 @@ static int intel_atomic_commit(struct drm_device *dev,
 	 * updates happen during the correct frames. Gen9+ have
 	 * double buffered watermarks and so shouldn't need this.
 	 *
-	 * Do this after drm_atomic_helper_setup_commit() and
-	 * intel_atomic_prepare_commit() because we still want
-	 * to skip the flip and fb cleanup waits. Although that
-	 * does risk yanking the mapping from under the display
-	 * engine.
+	 * Unset state->legacy_cursor_update before the call to
+	 * drm_atomic_helper_setup_commit() because otherwise
+	 * drm_atomic_helper_wait_for_flip_done() is a noop and
+	 * we get FIFO underruns because we didn't wait
+	 * for vblank.
 	 *
 	 * FIXME doing watermarks and fb cleanup from a vblank worker
 	 * (assuming we had any) would solve these problems.
@@ -12576,7 +12565,17 @@ static int intel_atomic_commit(struct drm_device *dev,
 	if (INTEL_GEN(dev_priv) < 9)
 		state->legacy_cursor_update = false;
 
-	ret = drm_atomic_helper_swap_state(state, true);
+	ret = intel_atomic_prepare_commit(dev, state);
+	if (ret) {
+		DRM_DEBUG_ATOMIC("Preparing state failed with %i\n", ret);
+		i915_sw_fence_commit(&intel_state->commit_ready);
+		return ret;
+	}
+
+	ret = drm_atomic_helper_setup_commit(state, nonblock);
+	if (!ret)
+		ret = drm_atomic_helper_swap_state(state, true);
+
 	if (ret) {
 		i915_sw_fence_commit(&intel_state->commit_ready);