[PATCH 4/4] drm/amdgpu: further optimize amdgpu_vm_handle_moved

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



I failed to git am the patch to test, the patch format has problem?

 From the patch self, is it possible that the vm->moved list is updated 
when you use local one?


Regards,

David Zhou


On 2018å¹´04æ??19æ?¥ 17:29, Christian König wrote:
> Splice the moved list to a local one to avoid taking the lock over and
> over again.
>
> Signed-off-by: Christian König <christian.koenig at amd.com>
> ---
>   drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 27 ++++++++++++++-------------
>   1 file changed, 14 insertions(+), 13 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> index 9da476637574..9d4ab7f70b1f 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
> @@ -1761,19 +1761,18 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev,
>   int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
>   			   struct amdgpu_vm *vm)
>   {
> +	struct amdgpu_bo_va *bo_va, *tmp;
> +	struct list_head moved;
>   	bool clear;
> -	int r = 0;
> +	int r;
>   
> +	INIT_LIST_HEAD(&moved);
>   	spin_lock(&vm->moved_lock);
> -	while (!list_empty(&vm->moved)) {
> -		struct amdgpu_bo_va *bo_va;
> -		struct reservation_object *resv;
> -
> -		bo_va = list_first_entry(&vm->moved,
> -			struct amdgpu_bo_va, base.vm_status);
> -		spin_unlock(&vm->moved_lock);
> +	list_splice_init(&vm->moved, &moved);
> +	spin_unlock(&vm->moved_lock);
>   
> -		resv = bo_va->base.bo->tbo.resv;
> +	list_for_each_entry_safe(bo_va, tmp, &moved, base.vm_status) {
> +		struct reservation_object *resv = bo_va->base.bo->tbo.resv;
>   
>   		/* Per VM BOs never need to bo cleared in the page tables */
>   		if (resv == vm->root.base.bo->tbo.resv)
> @@ -1786,17 +1785,19 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev,
>   			clear = true;
>   
>   		r = amdgpu_vm_bo_update(adev, bo_va, clear);
> -		if (r)
> +		if (r) {
> +			spin_lock(&vm->moved_lock);
> +			list_splice(&moved, &vm->moved);
> +			spin_unlock(&vm->moved_lock);
>   			return r;
> +		}
>   
>   		if (!clear && resv != vm->root.base.bo->tbo.resv)
>   			reservation_object_unlock(resv);
>   
> -		spin_lock(&vm->moved_lock);
>   	}
> -	spin_unlock(&vm->moved_lock);
>   
> -	return r;
> +	return 0;
>   }
>   
>   /**



[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux