On Tue, 22 Mar 2016 10:46:58 +0100 Paolo Bonzini <pbonz...@redhat.com> wrote:
> On 22/03/2016 10:07, Cornelia Huck wrote: > > So far, we had the best results with my refactoring + the mutex/bh > > change. Two problems: > > > > - We don't really understand yet why my refactoring helps, but passing > > the assign value is a good canditate (and it's agreed that this fixes a > > bug, I think.) > > - There's some problem with the bh, if I understood Stefan correctly. > > They can be fixed with just an extra object_ref/object_unref. > > I didn't understand that Tu Bo also needed the BH fix, and with that > information it makes sense. Passing the assign value ensures that > ioeventfd remains always assigned. With the CPU threads out of the > picture, the BH becomes enough to make everything thread-safe. Yes, this makes sense. Might we still have a hole somewhere in dataplane teardown? Probably not, from reading the code, even if it runs in cpu thread context.