Re: proposal: stop using the xen_system_time timecounter in dom0

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Fri, Jun 21, 2024 at 09:26:55AM +0200, Martin Husemann wrote:
> On Fri, Jun 21, 2024 at 08:59:01AM +0200, Manuel Bouyer wrote:
> > On Thu, Jun 20, 2024 at 05:16:38PM -0400, Greg Troxel wrote:
> > > Manuel Bouyer <bouyer%antioche.eu.org@localhost> writes:
> > > 
> > > > On Thu, Jun 20, 2024 at 01:53:34PM -0700, Greg A. Woods wrote:
> > > >> I would like to propose that we stop using the xen_system_time
> > > >> timecounter in NetBSD dom0, i.e. in XEN3_DOM0.
> > > >
> > > > Why can't you just set kern.timecounter.hardware to something else
> > > > if xen_system_time isn't working for you ?
> > > 
> > > I suppose he can, but I read that as an argument that xen_system_time is
> > > not reliable, and thus should be withdrawn.  People should not have to
> > > add workarounds.
> > 
> > it is perfectly reliable for me. In my case I'm not sure (really)
> > that other sources would be better.
> 
> This perfectly matches the real world on many amd64 machines where not
> all timecounter sources work as expected (due to firmware bugs or whatever).
> 
> However, in this case an easy workaround would be to lower the quality
> estimate of the timecounter if the cpu in use does not have an
> invariant TSC - I think we do something like that (with a bad list and
> some cpu ident magic) on some other timecounters, maybe that logic could
> be reused here?

That is certainly more reasonable than removing it completely

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