linux/drivers/gpu
Prike Liang e0153b94a1 drm/amdgpu: unmap userq for evicting user queue
If the driver only preempts queues, there can still be inflight waves,
pending dispatch state, or resume/redispatch possibility tied to the
same queue. Then the VM/TTM side may proceed to move/unmap queue related
BOs during evicting userq objects while shader TCP clients still need to
access them.

So for eviction, unmap is safer because it makes the queue nonrunnable
before memory backing is invalidated. Meanwhile, for a idle queue it's
more sutiable for unmapping it rather preempt and unmapping also can
save more processing time than preempt.

Signed-off-by: Prike Liang <Prike.Liang@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
(cherry picked from commit d87c9d8672)
History 2026-06-03 14:48:49 -04:00
..
drm drm/amdgpu: unmap userq for evicting user queue 2026-06-03 14:48:49 -04:00
host1x gpu: host1x: convert MIPI to use operation function pointers 2026-03-19 08:18:36 +01:00
ipu-v3 Convert 'alloc_obj' family to use the new default GFP_KERNEL argument 2026-02-21 17:09:51 -08:00
nova-core drm for v7.1-rc1 2026-04-15 08:45:00 -07:00
tests gpu/tests/gpu_buddy: Add gpu_test_buddy_alloc_range for exact-range allocation 2026-03-09 13:58:12 +05:30
trace gpu/trace: make TRACE_GPU_MEM configurable 2025-07-08 09:34:47 -07:00
vga Convert 'alloc_obj' family to use the new default GFP_KERNEL argument 2026-02-21 17:09:51 -08:00
buddy.c drm/buddy: Improve offset-aligned allocation handling 2026-03-09 12:36:10 +05:30
Kconfig gpu: Move DRM buddy allocator one level up (part two) 2026-02-06 11:38:35 +10:00
Makefile gpu: Fix uninitialized buddy for built-in drivers 2026-02-24 13:40:59 +10:00