At Thu, 28 Mar 2024 08:09:27 -0400, Brad Spencer <brad%anduin.eldar.org@localhost> wrote: Subject: Re: timekeeping regression? > > For me personally, I only have problems with PV/PVSHIM guests and only > if they are using more than one vcpu. The PVH guests appear to be fine. > But to mention again, I don't have the same problem as Greg. In my > case, the monotonic clock goes backwards, but the guest can keep time > sync just fine with ntpd or chronyd until that event. PVH for FreeBSD, PV for NetBSD. For FreeBSD I don't see that there should be any difference whatsoever in terms of timekeeping, clocks, and timecounters between PV and PVH. In all cases the Xen kernel is the source of authority. I do have one machine, a Dell R510 with Intel E5645 CPUs, which is currently booted with: dom0_max_vcpus=4 dom0_vcpus_pin=true and all the domUs (FreeBSD PVH and NetBSD PV) keep good time there. That CPU does not have a reliable TSC either, according to Xen, where it also uses HPET as its platform timer. The only difference with the R510 is that all the domUs there run NTP with multiple remote servers, while my local machines always use the one local NTP server (which is running in a dom0) (partly because I've had trouble trying to run multiple NTPd's with the same set of multiple remote servers behind a single-address NAT, as might be expected). -- Greg A. Woods <gwoods%acm.org@localhost> Kelowna, BC +1 250 762-7675 RoboHack <woods%robohack.ca@localhost> Planix, Inc. <woods%planix.com@localhost> Avoncote Farms <woods%avoncote.ca@localhost>
Attachment:
pgpAi_751xTw0.pgp
Description: OpenPGP Digital Signature