Michael> You seem to assume that each completion is for exactly Michael> one wqe. How does this work for selective signaling? Oh crap, back to the drawing board again... - R.