Revert "drm/amdgpu: Avoid extra evict-restore process."

This reverts commit a3201e3b7c which is
commit 1f02f2044b upstream.

This commit introduced a regression, however the fix for the regression:
aa5fc4362f ("drm/amdgpu: fix task hang from failed job submission
during process kill") depends on things not yet present in 6.12.y and
older kernels.  Since this commit is more of an optimization, just
revert it for 6.12.y and older stable kernels.

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org # 6.1.x - 6.12.x
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This commit is contained in:
Alex Deucher 2025-08-29 15:36:52 -04:00 committed by Greg Kroah-Hartman
parent b9cf42a8f0
commit 8d7ee56bf5

View File

@ -2125,11 +2125,13 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t min_vm_size,
*/
long amdgpu_vm_wait_idle(struct amdgpu_vm *vm, long timeout)
{
timeout = drm_sched_entity_flush(&vm->immediate, timeout);
timeout = dma_resv_wait_timeout(vm->root.bo->tbo.base.resv,
DMA_RESV_USAGE_BOOKKEEP,
true, timeout);
if (timeout <= 0)
return timeout;
return drm_sched_entity_flush(&vm->delayed, timeout);
return dma_fence_wait_timeout(vm->last_unlocked, true, timeout);
}
/**