Martti Kuparinen wrote:

> > > For i386, amd64, sparc, sparc64 and prep the time keeping code has 
> been
> > > changed to
> > > FreeBSD's timecounters. Could you send me the /var/run/dmesg.boot 
> file
> > > and the output
> > > of "sysctl kern.timecounter" ? There might be something different
> > > depending on which
> > > counter is picked and what counter frequency is assumed.
>
> VMware 3.2 on Linux, NetBSD 4beta. Full dmesg.boot is below.
>
>     # sysctl kern.timecounter
>     kern.timecounter.choice = clockinterrupt(q=0, f=100 Hz) 
> TSC(q=-100, f=1813623160 Hz) piixpm0(q=1000, f=3579545 Hz) 
> i8254(q=100, f=1193182 Hz) dummy(q=-1000000, f=1000000 Hz)
>     # sysctl -w kern.timecounter.hardware=i8254
>     kern.timecounter.hardware: piixpm0 -> i8254
>
> This makes "sleep 1" consistent with system time.
>
>     # while :
>     > do
>     > date
>     > sleep 1
>     > done
>     Fri Sep  1 16:23:53 EDT 2006
>     Fri Sep  1 16:23:54 EDT 2006
>     Fri Sep  1 16:23:55 EDT 2006
>     ^C
>     #
>
> With piixpm0, "sleep 1" is correct but system time is 30+ times
> too slow.
>
>     # while :
>     > do
>     > date
>     > sleep 1
>     > done
>     Fri Sep  1 16:27:38 EDT 2006
>     Fri Sep  1 16:27:38 EDT 2006
>     Fri Sep  1 16:27:38 EDT 2006

[...]

Well, looks like VMware emulates the powermanagement controller not like
the real hardware behaves. Try a kernel where piixpm and iic at piixpm 
is unconfigured.
I see actually no added benefit in using piixpm in a VMware environment.

I'll make a mental note of piixpm counter misbehaving in VMware 
environments.

Thanks for testing.

Frank