Re: [PATCH] per-cpu timer changes

[Don Fry <brazilnut@us.ibm.com>]

I am adding some debugging to figure this out.  Right now, I have
nothing running on the system except dom0.  Xwindows is not running.
The system has 8 cpus, (4 HT) and I see this on dom0 when it is compiled
with and without SMP support.  The only thing running on dom0 is
tail -f /var/log/messages and that is very quiet.

For example, between 11:34:01 and 11:54:01 I had a 5 second pause occur
3 times.  Between 11:59:00 and 12:14:01 it occurred 4 times.

Since May 25 16:53:31 and now it has happened about 180 times.

On Thu, May 26, 2005 at 11:03:52AM -0700, Nivedita Singhvi wrote:
> Don Fry wrote:
> 
> >The time going backwards was only occasionally, and it was a BIG jump
> >backwards.  I tracked it down yesterday to a problem with doing 32-bit
> >arithmetic in Linux on the tsc values.  For some reason, every 5-20
> >minutes xen seems to pause for about 5 seconds.  This causes the tsc to
> >wrap if only 32-bits are used, and the 'time went backwards' message is
> 
> What happens when there is activity on the system? Do you
> still see the pause? 5 seconds is an inordinately long time,
> and will impact networking if it happens during normal
> workloads too..
> 
> thanks,
> Nivedita
-- 
Don Fry
brazilnut@us.ibm.com