On Fri, Jan 04, 2008 at 01:05:57AM -0500, Thor Lancelot Simon wrote:
> On Thu, Jan 03, 2008 at 08:56:20PM -0500, Thor Lancelot Simon wrote:
> > 
> > A kernel from Gentoo's package for the XenSource 2.6.20-xen-r6 sources
> > (which seem to be end-of-the-line for the separate Xen patches?) has
> > working net when run with a NetBSD dom0, but has this problem instead:
> > 
> > 	madcow dev # date
> > 	Thu Jan  3 16:51:01 EST 2008
> > 	madcow dev # ntpdate pool.ntp.org
> >  	3 Jan 16:51:08 ntpdate[3761]: step time server 69.31.13.210 offset 14495.938724 sec
> > 	madcow dev # date
> > 	Thu Jan  3 16:51:10 EST 2008
> > 	madcow dev # date 01031951
> > 	Thu Jan  3 19:51:00 EST 2008
> > 	madcow dev # date
> > 	Thu Jan  3 16:51:26 EST 2008
> 
> I don't know why the kernel in question is syncing to the dom0 clock
> except with an offset of several hours.  I'd love to.  However, the
> workaround is to echo 1 > /proc/sys/xen/independent_wallclock before
> running ntpdate or ntpd.

I think it behaves this way because it assumes dom0/hypervisor has the right
time, and there's no need to have an independant, local clock. But it seems
to not handle timezones right. Maybe it has a confusion about what timezone
the hypervisor time should be in. I note that you have a 3h difference,
while EST offset is 5 hours, so it's more complex than just local vs GMT time

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--