> What you can do is change notify_guest to something like
>
> qemu_bh_schedule(req->dev->dataplane->notify_guest_bh);
>
> and do the actual notification in the bottom half. This should ensure that
> multiple notifications are coalesced, but it may also introduce new
> aio_notify calls even with my patch (a BH scheduled from a BH currently does
> an aio_notify; this can be fixed).
I think we can do better than above because one aio IO completes lots of
requests, and we just need to notify guest for all these requests.