>>> On 13.08.15 at 12:14, <ian.campb...@citrix.com> wrote: > There isn't a race here is there? What if the reference were dropped after > this check but before the guest_physmap_add_page (which takes new > references)? We are implicitly relying on a guarantee made elsewhere that a > direct mapped guest never drops the final ref until it is destroyed (which > can't be happening in this window I think), but it would be obviously safer > against such bugs if we just held the ref until after the p2m was updated.
Yeah, we depend on this, and what you suggest would indeed be an improvment. > Another thing for a future/4.7 cleanup I guess. Right. But as bottom line I take it that your objections to the patch going in soon have been eliminated? Jan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel