On Mon, Sep 19, 2022 at 09:58:10AM +1000, Gavin Shan wrote: > I think Marc's explanation makes sense. It won't make difference in terms > of performance. We need to explicitly handle barrier when kvm_test_request() > is used. So I prefer to keep the code if Peter agrees. No strong opinion here. I keep thinking clear+set look awkward even if it's unlikely path to trigger (ring should be recycled when reaching here for any sane user app), but if it's already 2:1 then please go ahead. :) Thanks, -- Peter Xu
