Two fixes to address the two issues found in: https://lore.kernel.org/dri-devel/TfhgV1W0W5LI6RWUO6J35B3R8QIYH_FN3Eihzdo_9PH39hfn1AVByT-QBPHmWiAR-L0Kqi2sppM_EhFPKwZPGmb5pFgpF2MrzeFzkZDmpG8=@proton.me/
Many thanks to Luke for reporting, sharing one AI proposed fix, and the super quick testing turnaround! After these two we can hopefully work at promoting the "fair" policy back to the default in an upcoming kernel release, followed by removing the FIFO and round-robin, and then re-applying the code base simplifications which we had to temporarily revert. Cc: Christian König <[email protected]> Cc: Danilo Krummrich <[email protected]> Cc: [email protected] Cc: Matthew Brost <[email protected]> Cc: Philipp Stanner <[email protected]> Cc: Pierre-Eric Pelloux-Prayer <[email protected]> Cc: Vitaly Prosyak <[email protected]> Tvrtko Ursulin (2): drm/sched: Do not restore unsaved virtual runtime drm/sched: Ensure monotonic min_vruntime drivers/gpu/drm/scheduler/sched_entity.c | 8 ++- drivers/gpu/drm/scheduler/sched_rq.c | 85 ++++++++---------------- include/drm/gpu_scheduler.h | 2 + 3 files changed, 37 insertions(+), 58 deletions(-) -- 2.54.0
