Re: NetBSD/sparc running on qemu

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

On Sun, Jan 17, 2010 at 02:30:10AM +0900, Izumi Tsutsui wrote:
> > > I tried it, it doesn't change behavior.
> > > 
> > > The difference in timecounter wouldn't understand the difference in
> > > behavior:
> > > on a i386 guest, sleep(10) takes 20s, and date says it took 20s
> > > on a sparc guest, sleep(10) takes 20s, and date says it took 10s
> > 
> > Ah, I thought sleep used hardclock(9), but I was wrong.
> > 
> > I'll try to build the latest pkgsrc/emulators/qemu to test it.
> > (current qemu-system-sparc I have dumps core right after mountroot)
> 
> qemu-system-sparc still dumps core,
> but I test it on qemu (emulating i386).
> 
> nanosleep(2) in kern_time.c still uses tstohz() and kpause(9),
> and I think kpause(9) depends on hz(9). On sparc, timer0 at obio0
> is used not only for timecounter(9) but also for hz(9),
> so they show the same count.
> 
> On i386, it uses ACPI-Safe for timercounter(9) but hz(9) uses
> a different source so they show different count, I guess.
> 
> ---
> # uname -mrs
> NetBSD 5.0.1 i386
> # sysctl kern.timecounter.hardware
> kern.timecounter.hardware = ACPI-Safe
> # date;sleep 10;date
> Sun Jan 17 02:20:52 JST 2010
> Sun Jan 17 02:21:12 JST 2010
> # sysctl -w kern.timecounter.hardware=clockinterrupt
> kern.timecounter.hardware: ACPI-Safe -> clockinterrupt
> # date;sleep 10;date
> Sun Jan 17 02:22:04 JST 2010
> Sun Jan 17 02:22:14 JST 2010
> # 

And I guess, in your second test, sleep 10 still takes
20 real seconds, right ?
That would mean that on qemu-i386, the acpi source is right, but the
clockinterrupt one is wrong, just like as timer0 at obio0 on sparc.

I'll try to see how acpi is different from other clock sources in qemu.

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