This series is based on top of [1] floated by Maarten. To fix regressions seen in CI, following changes were made on top of the original series. 1. add a check in the plane state destroy hook to not move forward if the vblank worker is scheduled. 2. add checks before accessing frame buffer object (we are not sure yet how much this helps but we have found that this operation causes dump stacks) 3. do not defer the intel atomic cleanup into a work queue, instead execute it at the end of atomic commit tail. While this series is in noway a complete or proper solution it is meant to trigger a discussion to arrive at one. [1] https://patchwork.freedesktop.org/series/126934/ v2: Add missing patch v3: Remove misleading error log Change condition to access fb object v4: Remove unused variables Chaitanya Kumar Borah (3): drm/i915: do not destroy plane state if cursor unpin worker is scheduled drm/i915: Add sanity check before accessing fb buffer object drm/i915: do not defer cleanup work Maarten Lankhorst (2): drm: Add drm_vblank_work_flush_all(). drm/i915: Use the same vblank worker for atomic unpin Ville Syrjälä (1): drm/i915: Use vblank worker to unpin old legacy cursor fb safely drivers/gpu/drm/drm_vblank_work.c | 22 ++++++++ .../gpu/drm/i915/display/intel_atomic_plane.c | 52 ++++++++++++++++++- .../gpu/drm/i915/display/intel_atomic_plane.h | 4 ++ drivers/gpu/drm/i915/display/intel_crtc.c | 27 ++++++++++ drivers/gpu/drm/i915/display/intel_cursor.c | 26 +++++++++- drivers/gpu/drm/i915/display/intel_cursor.h | 3 ++ drivers/gpu/drm/i915/display/intel_display.c | 11 ++-- .../drm/i915/display/intel_display_types.h | 3 ++ include/drm/drm_vblank_work.h | 2 + 9 files changed, 141 insertions(+), 9 deletions(-) -- 2.25.1
