Re: [Qemu-devel] BSOD when using Hyper-V extensions with Win2k8R2 guests

[Anthony Liguori]

On 03/15/2012 02:54 PM, Vadim Rozenfeld wrote:
> On Thursday, March 15, 2012 09:14:39 PM Michael Roth wrote:
> > Hi Vadim,
> >
> > I've been trying to use -cpu qemu64,hv_relaxed as a workaround for a
> > CLOCK_WATCHDOG_TIMEOUT BSOD on Windows 2008 R2 guests when a vcpu's
> > execution is delayed for an extended period. When I do this however I get
> > the following the BSOD:
> >
> > http://msdn.microsoft.com/en-us/library/windows/hardware/ff559206(v=vs.85).
> > aspx
> >
> > I've also tried -cpu qemu64,hv_spinlocks=4 and -cpu qemu64,hv_vapic as a
> > sanity check and this results in the same error being thrown.
> >
> > Host is RHEL6.1 using latest qemu.git, guests are win2k8 server r2 (same
>
> I'm pretty much sure that hyper-v support was not backported into RHEL 6.1
> KVM.

So is there a hyper-v capability for KVM then?  Or a CPUID feature?

Should QEMU refuse to enable hv_relaxed if the capability isn't present?

Regards,

Anthony Liguori

> Best regards,
> Vadim.
>
> > behavior for both SP0 and SP1). Here's the full command-line:
> >
> > qemu-upstream -L ../qemu-build/pc-bios -cpu qemu64,hv_relaxed --enable-kvm
> > \ -M pc-1.1 -enable-kvm -m 4096 -smp 4,sockets=4,cores=1,threads=1 \ -name
> > ichigo-dom45 -nodefconfig -nodefaults -rtc base=utc,driftfix=slew \ -drive
> > file=/scratch/mdroth/win2k8r2sp1.raw,if=none,id=drive0,format=raw,cache=wr
> > itethrough,snapshot=on \ -device
> > virtio-blk-pci,bus=pci.0,drive=drive0,id=disk0,bootindex=1 \ -netdev
> > tap,id=hostnet0,script=/etc/qemu-ifup \
> >      -device
> > virtio-net-pci,netdev=hostnet0,id=net0,mac=00:FF:FE:00:00:2d,bus=pci.0 \
> > -device virtio-balloon-pci,id=balloon0,bus=pci.0 \
> >      -usb -device usb-tablet,id=input0 -vnc 0.0.0.0:45 -vga cirrus
> >
> > Unfortunately the BSOD is too early to get a kernel dump, but here are
> > the error params in case that's useful:
> >
> > 0xFC0000096, 0xF800027F6200, 0x0, 0x0
> >
> > Let me know if you need any other details.
> >
> > Thanks!