RE: [Xen-devel] Re: [PATCH 3/5] x86/pvclock: add vsyscall implementation

[Dan Magenheimer <dan.magenheimer@oracle.com>]

Is there any way for an application to conclusively determine
programmatically if the "fast vsyscall" pvclock is functional
vs the much much slower gettimeofday/clock_gettime equivalents?

If not, might it be possible to implement some (sysfs?)
way to determine this, that would also be backwards compatible
to existing OS's that don't have pvclock+vsyscall supported?

Thanks,
Dan

> -----Original Message-----
> From: Avi Kivity [mailto:avi@redhat.com]
> Sent: Wednesday, October 14, 2009 6:33 AM
> To: Jeremy Fitzhardinge
> Cc: Dan Magenheimer; Jeremy Fitzhardinge; Kurt Hackel; the arch/x86
> maintainers; Linux Kernel Mailing List; Glauber de Oliveira Costa;
> Xen-devel; Keir Fraser; Zach Brown; Chris Mason; Ingo Molnar
> Subject: Re: [Xen-devel] Re: [PATCH 3/5] x86/pvclock: add vsyscall
> implementation
> 
> 
> On 10/14/2009 05:00 AM, Jeremy Fitzhardinge wrote:
> >
> >> So it's broken or disabled when that assumption is wrong?  We could
> >> easily fix that now.  Might even reuse the pvclock structures.
> >>      
> > Well, the kernel internally makes more or less the same 
> assumption; the
> > vsyscall clocksource is the same as the kernel's internal 
> one.  I think
> > idea is that it just drops back to something like hpet if the tsc
> > doesn't have very simple SMP characteristics.
> >
> > If the kernel could characterize the per-cpu properties of 
> the tsc more
> > accurately, then it could use the pvclock mechanism on native.
> >    
> 
> It does - that's how kvm implements pvclock on the host side.  See 
> kvmclock_cpufreq_notifier() in arch/x86/kvm/x86.c.
> 
> -- 
> I have a truly marvellous patch that fixes the bug which this
> signature is too narrow to contain.
> 
> 
>