Changes since v1:

* Reworked QUEUE_RESET handling based on review feedback.
  Use the doorbell xarray to look up affected queues instead of
  scanning all queues. Move queue reset handling into amdgpu_userq.c
  and consolidate queue state updates, EVENTFD signaling, and fence
  completion in a single helper.

* Reworked GPU_RESET notification placement.
  Move signaling out of the scheduler resume path and notify only
  after successful hardware reset completion. Avoid notifications
  from reset-skip and recovery-only paths.

* Dropped SCRATCH EVENTFD notifications.
  Queue restore failures due to missing VA mappings require separate
  per-queue error handling and are not equivalent to SCRATCH events.

* Dropped MEMORY_EXCEPTION EVENTFD notifications.
  The VM fault interrupt path requires further discussion around
  PASID ownership and event routing before notification support can
  be added.
  
Only compilaion tested.

Srinivasan Shanmugam (3):
  drm/amdgpu: Signal USERQ_EOP EVENTFD notifications
  drm/amdgpu: Signal QUEUE_RESET EVENTFD notifications
  drm/amdgpu: Signal GPU_RESET EVENTFD notifications

 drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 51 +++++++++++++++++++++-
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.c  | 45 +++++++++++++++++++
 drivers/gpu/drm/amd/amdgpu/amdgpu_userq.h  |  2 +
 drivers/gpu/drm/amd/amdgpu/mes_userqueue.c | 21 +++------
 4 files changed, 103 insertions(+), 16 deletions(-)

-- 
2.34.1

Reply via email to