>>> Keir Fraser <Keir.Fraser@xxxxxxxxxxxx> 12.06.06 12:21 >>>
>
>On 12 Jun 2006, at 10:35, Jan Beulich wrote:
>
>> I continue to see occasional 'time went backwards' messages on 
>> 3.0-testing on larger systems. Is anyone else seeing
>> such? Have there been any improvements to -unstable that haven't been 
>> moved over? If so, can anyone point out the
>> respective changesets? Thanks, Jan
>
>That would mean cross-cpu jitter of >10ms, when TSCs are resynced every 
>1s. That's a lot of jitter. I would guess that either we have a bug or 
>something is changing CPU clock frequencies every once in a while 
>(perhaps automatically?). Any idea if the messages correlate with lots 
>of system activity or anything like that?
I cannot entirely exclude clock freq changes - while we have all CPUFREQ stuff 
disabled in our Xen kernels, the
messages only appear during late system boot about the time when the power save 
daemon (which would normally load and
configure cpufreq modules) gets started (but other things started at that time 
are e.g. the xen daemon and the network
interfaces); I'd have to check what things the power save stuff might be doing 
behind our backs.
Heavy system load doesn't appear to be triggering this.
Manually (re-)starting the power save daemon doesn't trigger this either.
Nor does it happen every time the system gets booted.
It seems to get harder to reproduce the longer the system had been up (running 
Xen or native Linux) prior to the reboot
into Xen. But I can see this for instance when booting into Xen twice in a row 
(without doing a lot after the system was
up for first time) after turning on the previously cold system.
Jan
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel