HWInfo is crashing my VM while reading CPU info

[marcosscriven]

Please see full attached debug report.

However, the error seems to be at the end:

Log CPUs: 12, cores: 6
RdMSR[#0000001B] = 00000000:FEE00900
APIC ID[00]: PackageID=0, CoreID=0, ThreadID=0, M, BSP
ExtCPUname: "AMD EPYC Processor (with IBPB)"
RdMSR[#C0010293] ERROR
RdMSRm: #C0010064 #C0010065 #C0010066 #C0010067 #C0010068 #C0010069 #C001006A #C001006B
APIC_CLK = 1001.216 (took 4 ms)

This is in the context of a VM - however, I'd still like to use HWINFO as I'm passing through two components natively: the SSD and the GPU, and both have sensors I'd like to monitor.

I'd also be curious what exactly HWINFO is doing that is able to completely crash the VM?

Thanks for the great software.

 

[Martin]

So this not just an application crash, but entire VM is crashing? In such case I'd say the VMM is not OK. What VM is that, Microsoft Hyper-V?

 

[marcosscriven]

Sorry, it‘s not correct to for me to say it crashes the VM.

When I check the Windows events afterwards it says something is crashing the kernel. From the host perspective everything is fine with the VM.

I totally get it’s not the intended use case for HWiNFO here, however it really would be great to get an idea what it’s doing at this point so I could use that to figure out what’s wrong.

There‘s various options in the VM setup on passing through CPU topology, so wondering if it’s that?

EDIT The VM is Qemu/kvm in a Linux host

 

[Martin]

Yes, it might be due to some features not being passed thru, but I'm not sure what exact it might be. Something related to MSRs if there's such...
Anyway, I will try to add a workaround for this in the next build, but not sure if that will be sufficient.

 

[marcosscriven]

Thank you @Martin - I’ll keep a lookout.

 

[marcosscriven]

@Martin - I just tried beta v6.25-4135 and it worked perfectly (without any changes to my setup).

May I ask what you did please?

And thanks!

 

[Martin]

Thanks for your feedback.
I noticed that a certain MSR (which is present on all Zen CPUs) is not available under the VM, so I check for that and skip access in such case.

 

[marcosscriven]

May I ask which register? It may be useful feedback for the QEMU devs.

I Googled “Ryzen FEE00900” and could only find something called “APIC base”

Thanks again

 

[Martin]

MSR 0xC0010293

 

[marcosscriven]

Doh - it’s right there in the error log, sorry.

looks like something to do with boost state:https://github.com/cyring/CoreFreq/blob/master/amdmsr.h

 

[marcosscriven]

Thanks for your feedback.
I noticed that a certain MSR (which is present on all Zen CPUs) is not available under the VM, so I check for that and skip access in such case.

Just an idea - as I’m not sure how the code works. But could you have a function that wraps all such register accesses, and debug logs if they aren’t there?