Re: NetBSD/sparc running on qemu

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

On Sat, Jan 16, 2010 at 11:06:55AM +0900, Izumi Tsutsui wrote:
> > One oddity is that the clock and TOD in the virtual machine seems to run
> > too slow by a factor of 2.
> > A 'date; sleep 10; date' in the sparc virtual machine show 10s between
> > the 2 dates commands, but it really took 20s to run.
> > The same 'date; sleep 10; date' in a NetBSD/i386 qemu virtual machine
> > also needs 20s to run (which looks like an issue with qemu in the
> > first place) but the time of day printed by date is correct (there is
> > a 20s difference between the 2 dates). Anyone have an idea what could
> > cause this difference of behavior between NetBSD/i386 and
> > NetBSD/sparc ? I guess i386 reads the TOD for each date command
> > while sparc reads it only at boot ?
> 
> No, TOD clock is only read in inittodr(8) at boot on all port, I think.
> The problem could happen if the frequency of timecounter source
> (timer0 at obio0 on sun4m?) is not correct value.

It's indeed incorrect as sleep 10 sleeps for 20s, on both i386 and
sparc virtual machines (both running NetBSD). What I understand is why
date(1) shows the correct value (that it, it really took 20s to
run sleep 10) on i386, and not on sparc ?

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