Our atomic transaction maintains a bitmask of planes that we need to update sprite watermarks for once vblank evasion is complete. When we actually go to make use of that bitmask, we've been comparing against the plane index rather than the plane mask; we need to update our comparison to check '(1 << index)' rather than 'index' directly. This bug was introduced by commit 32b7eeec4d1e861230b09d437e95d76c86ff4a68 Author: Matt Roper <matthew.d.roper@xxxxxxxxx> Date: Wed Dec 24 07:59:06 2014 -0800 drm/i915: Refactor work that can sleep out of commit (v7) However we've been "lucky" and haven't actually run into problems caused by this yet due to another bug in the code...we forgot to actually remove the calls to intel_update_sprite_watermarks() from the low-level plane programming functions that run under evasion. We'll fix that problem in a subsequent patch. Signed-off-by: Matt Roper <matthew.d.roper@xxxxxxxxx> --- drivers/gpu/drm/i915/intel_display.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_display.c b/drivers/gpu/drm/i915/intel_display.c index 5c2047b..2878924 100644 --- a/drivers/gpu/drm/i915/intel_display.c +++ b/drivers/gpu/drm/i915/intel_display.c @@ -13460,7 +13460,8 @@ static void intel_finish_crtc_commit(struct drm_crtc *crtc) intel_post_enable_primary(crtc); drm_for_each_legacy_plane(p, &dev->mode_config.plane_list) - if (intel_crtc->atomic.update_sprite_watermarks & drm_plane_index(p)) + if (intel_crtc->atomic.update_sprite_watermarks & + (1 << drm_plane_index(p))) intel_update_sprite_watermarks(p, crtc, 0, 0, 0, false, false); -- 1.8.5.1 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/intel-gfx