The MES has special handling to track unmap
failures for KFD processes when they occur during
queue removal. This is because the driver is expected to call remove_queue
again after a reset.
Normally, unmap failures aren't remembered during queue removal as cleanup
is performed. However, for KFD processes we need to maintain this state.
This change sets the is_kfd_process flag to 1 when creating user queues
to ensure proper handling of unmap failures during queue removal scenarios.
Suggested-by: Chau Kyle-haiChau <kyle-hai.c...@amd.com>
Signed-off-by: Jesse Zhang <jesse.zh...@amd.com>
---
drivers/gpu/drm/amd/amdgpu/mes_userqueue.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
b/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
index d6f50b13e2ba..64aa749c58c5 100644
--- a/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
+++ b/drivers/gpu/drm/amd/amdgpu/mes_userqueue.c
@@ -143,7 +143,7 @@ static int mes_userq_map(struct amdgpu_userq_mgr *uq_mgr,
queue_input.doorbell_offset = userq_props->doorbell_index;
queue_input.page_table_base_addr =
amdgpu_gmc_pd_addr(queue->vm->root.bo);
queue_input.wptr_mc_addr = queue->wptr_obj.gpu_addr;
-
+ queue_input.is_kfd_process = 1;
amdgpu_mes_lock(&adev->mes);
r = adev->mes.funcs->add_hw_queue(&adev->mes, &queue_input);
amdgpu_mes_unlock(&adev->mes);
--
2.49.0
