On Thu, May 20, 2021, Jim Mattson wrote: > Initially, I added a new kvm request for a userspace exit on the next guest > entry, but Sean hated that approach. Hey! You could at least say _why_ I hated it. :-D My argument is that the only way to guarantee that vcpu->run->exit_reason isn't crushed between making the request and servicing the request is by guaranteeing that KVM makes it back to the run loop immediately after making the request. Since the only way to accomplish that is by forcing a return up the stack, at that point we can simply return the "request" directly.
