On 04/11/2018 11:17 AM, Dario Faggioli wrote: > On Wed, 2018-04-11 at 12:00 +0200, Olaf Hering wrote: >> On Wed, Apr 11, Dario Faggioli wrote: >> >>> Olaf, can you give it a try? It should be fine to run it on top of >>> the >>> last debug patch (the one that produced this crash). >> >> Yes, with both changes it did >4k iterations already. Thanks. >> > That's great to hear! :-D > > Now, I think I'll submit it as a proper patch in the variant that > Juergen suggested, and that I also were thinking to use. > > George, any opinion? I'm going somewhere now. If I don't hear any > pushback, I'll do that as soon as back.
I think for simplicity / reliability of backporting, we should start with a patch like the one you gave to Olaf (i.e., adding the "missing" vcpu_sleep_nosync()). Moving forward we should definitely move things around so that there's no risk of accidentally forgetting to take the vcpu off the runqueue, but there are some other changes it might be nice to make as well; for instance, it looks like on a busy system there may be a fair amount of duplicate cpu_pick() calculations; it would be nice to avoid that. But those probably shouldn't be done during the feature freeze. -George _______________________________________________ Xen-devel mailing list Xen-devel@lists.xenproject.org https://lists.xenproject.org/mailman/listinfo/xen-devel
