I have a host system running a recent 2.6.33-rc8 kernel, with
constant_tsc, and a bunch of guests. One of them, running rawhide, sees
constant clock drifts on the order of 4-12 hours at a time.
I'm going to play with switching to the acpi_pm timer, but I am
interested in what's happening with the Fedora bugs on this. It seems
that over the past year there have been bugs (e.g. 449187, 475598) that
mention this problem and the issue has been confused with non-constant
tsc host clocksources. The fact that there is the "Bug Zapper" closing
perfectly useful bugs also does not help the situation.
I know the KVM folks are poking at TSC issues, etc. but has anyone got
anything to note on this particular issue?
Show replies by date