[PATCH v2 2/3] drm/xe: Use device, gt ordered work queues for resource cleanup

[Matthew Brost <matthew.brost@intel.com>]

Resource cleanup is a device private operations with no expectation of
performance. Use device, gt ordered work queues to cleanup resources to
avoid grabbing locks on shared work queues.

Signed-off-by: Matthew Brost <matthew.brost@intel.com>
---
 drivers/gpu/drm/xe/xe_device_types.h | 5 ++++-
 drivers/gpu/drm/xe/xe_execlist.c     | 2 +-
 drivers/gpu/drm/xe/xe_gt_types.h     | 5 ++++-
 drivers/gpu/drm/xe/xe_guc_submit.c   | 2 +-
 drivers/gpu/drm/xe/xe_vm.c           | 4 ++--
 5 files changed, 12 insertions(+), 6 deletions(-)

diff --git a/drivers/gpu/drm/xe/xe_device_types.h b/drivers/gpu/drm/xe/xe_device_types.h
index c710cec835a7..d696aa2de8cc 100644
--- a/drivers/gpu/drm/xe/xe_device_types.h
+++ b/drivers/gpu/drm/xe/xe_device_types.h
@@ -366,7 +366,10 @@ struct xe_device {
 	/** @preempt_fence_wq: used to serialize preempt fences */
 	struct workqueue_struct *preempt_fence_wq;
 
-	/** @ordered_wq: used to serialize compute mode resume */
+	/**
+	 * @ordered_wq: used to serialize compute mode resume, cleanup
+	 * resources
+	 */
 	struct workqueue_struct *ordered_wq;
 
 	/** @unordered_wq: used to serialize unordered work, mostly display */
diff --git a/drivers/gpu/drm/xe/xe_execlist.c b/drivers/gpu/drm/xe/xe_execlist.c
index dece2785933c..1ae922509f05 100644
--- a/drivers/gpu/drm/xe/xe_execlist.c
+++ b/drivers/gpu/drm/xe/xe_execlist.c
@@ -393,7 +393,7 @@ static void execlist_exec_queue_kill(struct xe_exec_queue *q)
 static void execlist_exec_queue_fini(struct xe_exec_queue *q)
 {
 	INIT_WORK(&q->execlist->fini_async, execlist_exec_queue_fini_async);
-	queue_work(system_unbound_wq, &q->execlist->fini_async);
+	queue_work(q->gt->ordered_wq, &q->execlist->fini_async);
 }
 
 static int execlist_exec_queue_set_priority(struct xe_exec_queue *q,
diff --git a/drivers/gpu/drm/xe/xe_gt_types.h b/drivers/gpu/drm/xe/xe_gt_types.h
index 2143dffcaf11..cd22ad6e881a 100644
--- a/drivers/gpu/drm/xe/xe_gt_types.h
+++ b/drivers/gpu/drm/xe/xe_gt_types.h
@@ -268,7 +268,10 @@ struct xe_gt {
 		} acc_queue[NUM_ACC_QUEUE];
 	} usm;
 
-	/** @ordered_wq: used to serialize GT resets and TDRs */
+	/**
+	 * @ordered_wq: used to serialize GT resets and TDRs, clean up
+	 * resources
+	 */
 	struct workqueue_struct *ordered_wq;
 
 	/** @uc: micro controllers on the GT */
diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
index 13b7e195c7b5..e30ad9fccf6c 100644
--- a/drivers/gpu/drm/xe/xe_guc_submit.c
+++ b/drivers/gpu/drm/xe/xe_guc_submit.c
@@ -1033,7 +1033,7 @@ static void guc_exec_queue_fini_async(struct xe_exec_queue *q)
 	if (q->flags & EXEC_QUEUE_FLAG_PERMANENT)
 		__guc_exec_queue_fini_async(&q->guc->fini_async);
 	else
-		queue_work(system_wq, &q->guc->fini_async);
+		queue_work(q->gt->ordered_wq, &q->guc->fini_async);
 }
 
 static void __guc_exec_queue_fini(struct xe_guc *guc, struct xe_exec_queue *q)
diff --git a/drivers/gpu/drm/xe/xe_vm.c b/drivers/gpu/drm/xe/xe_vm.c
index 8b32aa5003df..7808b540c013 100644
--- a/drivers/gpu/drm/xe/xe_vm.c
+++ b/drivers/gpu/drm/xe/xe_vm.c
@@ -1005,7 +1005,7 @@ static void vma_destroy_cb(struct dma_fence *fence,
 	struct xe_vma *vma = container_of(cb, struct xe_vma, destroy_cb);
 
 	INIT_WORK(&vma->destroy_work, vma_destroy_work_func);
-	queue_work(system_unbound_wq, &vma->destroy_work);
+	queue_work(xe_vma_vm(vma)->xe->ordered_wq, &vma->destroy_work);
 }
 
 static void xe_vma_destroy(struct xe_vma *vma, struct dma_fence *fence)
@@ -1625,7 +1625,7 @@ static void xe_vm_free(struct drm_gpuvm *gpuvm)
 	struct xe_vm *vm = container_of(gpuvm, struct xe_vm, gpuvm);
 
 	/* To destroy the VM we need to be able to sleep */
-	queue_work(system_unbound_wq, &vm->destroy_work);
+	queue_work(vm->xe->ordered_wq, &vm->destroy_work);
 }
 
 struct xe_vm *xe_vm_lookup(struct xe_file *xef, u32 id)
-- 
2.34.1