David Hildenbrand <david@xxxxxxxxxx> wrote: > The first DIO messes up the refcount (double-unpin?) such that the second one > detects the refcount underflow when trying to pin the page again? It seems to be something to do with the zero page:-/ David