Re: [Qemu-devel] Re: Timing problems

[Sven Zenker <Sven.Zenker@gmx.net>]

Hi,
people have had this problem with SpeedStep machines, where it is
related to qemu timing being based on CPU cycle count, which is thereby
messed up. 
Since it seems to occur with frequency scaling disabled as well, could
this be related to running on a HyperThreading CPU? Not sure what
happens there with the cycle counts, but to verify, you could try to
apply my previously posted patch that substituted the built-in mechanism
with realtime clock calls (was for 0.7.1, though, I think).
Best,
Sven

Am Montag, den 07.11.2005, 22:46 +0000 schrieb Michael Smith:
> Alexander Toresson <alexander.toresson <at> gmail.com> writes:
> 
> > time flies by at 5x the speed it should.
> 
> > PS. I'm susprised nobody has seen this problem before. Is it just me
> > who experience it?
> 
> No, I experience this as well.
> 
> I'm running Qemu 0.7.2 with Linux 2.6.14 (vanilla kernel) as host OS and Windows
> XP Professional as guest OS.  If the VM is running and actively using CPU, time
> runs much faster in the guest than in the host.  On the other hand, if the VM is
> mostly idle, time is slower in the guest than in the host OS.
> 
> This odd clock behavior seems to cause other problems in the guest OS,
> particularly during boot/service initialization.
> 
> I can recreate this behavior with or without KQemu.
> 
> I can recreate this behavior with or without CPU frequency scaling enabled.
> 
> It is frustrating, and I would be very happy if someone could suggest a fix or
> workaround.
> 
> Thanks,
> Mike Smith
> mas3f@alumni.virginia.edu
> 
> 
> 
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel
>