Pass cuda array to host function

Question

I'm not able to find any answer to this. Is it possible to pass a device array (pointer) to a host function? Inside this host function a kernel will be launched which will use the device array. I don't want to copy the data from host to device every time I call the (host) function, so that's why I copy the data (from host to device) before calling the (host) function. The kernel within this function will always use the same device array.