On Fri, Mar 23, 2018 at 12:10:47PM -0700, Mike Larkin wrote: > On Fri, Mar 23, 2018 at 09:22:26AM +0100, Andreas Kusalananda Kähäri wrote: > > Hi, > > > > I recently set up a VM on my Lenovo Thinkpad x230 to build ports. With > > yesterday's sources, the clock in the VM seems to run at about 50% the > > speed of the clock on the host. I don't think this was an issue at the > > start of the week. > > > > After 21 minutes of uptime on the host, the VM reports 12 minutes of > > uptime and the daemon log contains > > > > Mar 23 08:54:38 ports ntpd[8242]: ntp engine ready > > Mar 23 08:54:39 ports ntpd[83538]: set local clock to Fri Mar 23 > > 08:54:39 CET 2018 (offset 1.083490s) > > Mar 23 08:56:03 ports ntpd[89983]: adjusting local clock by 31.117545s > > Mar 23 08:58:04 ports ntpd[89983]: adjusting local clock by 84.676093s > > Mar 23 08:59:39 ports ntpd[89983]: adjusting local clock by 153.445943s > > Mar 23 09:01:09 ports ntpd[89983]: adjusting local clock by 239.135899s > > Mar 23 09:02:39 ports ntpd[89983]: adjusting local clock by 328.310449s > > Mar 23 09:04:10 ports ntpd[89983]: adjusting local clock by 417.339908s > > Mar 23 09:05:41 ports ntpd[89983]: adjusting local clock by 506.461867s > > > > ntpctl -s all on the VM reports > > > > 0/1 peers valid, 1/1 sensors valid, clock unsynced, clock offset is > > 506151.866ms > > > > peer > > wt tl st next poll offset delay jitter > > not resolved from pool pool.ntp.org > > 1 2 - 14s 15s ---- peer not valid ---- > > > > sensor > > wt gd st next poll offset correction > > vmmci0 > > 1 1 0 14s 15s 97005.318ms 0.000ms > > > > The host and VM runs exactly the same base system (as I said, with > > yesterday's sources). > > > > The following is the dmesg of the host followed by the dmesg of the VM. > > > > OpenBSD 6.3 (GENERIC.MP) #39: Thu Mar 22 17:04:01 CET 2018 [cut] > > WARNING: / was not properly unmounted > > > > -- > > Andreas Kusalananda Kähäri, > > National Bioinformatics Infrastructure Sweden (NBIS), > > Uppsala University, Sweden. > > > > What does sysctl kern.timecounter.hardware say in the VM? > > If it's not "tsc", set it to that and see if that helps. And if it does, > toss that into /etc/sysctl.conf. > > -ml >
This seems to solve my issue. Thanks, -- Andreas Kusalananda Kähäri, National Bioinformatics Infrastructure Sweden (NBIS), Uppsala University, Sweden.