Host system BSOD/Reboot when virtualizing a 64 bit guest on a Sandy Bridge setup

[sinni800]

Hey guys, before I thoroughly explain this problem again, please visit this page

http://superuser.com/questions/355177/host-system-resets-crashes-when-using-vmware-or-virtualbox

This is a question I posted on a different board. It applies to VMWare and Virtualbox and possible VirtualPC by Microsoft, though.

In essence, I get a complete HOST system crash when attempting to virtualize something. When VT-X is activated it crashes right in the BIOS of VMWare. When it is inactive it does not crash when virtualizing an XP system.

I think it is as follows: Activated VT-x crashes the Host machine right in the BIOS of the virtual one. Deactivated VT-x crashes 64-bit systems, seemingly when they turn the CPU into long (64bit) mode. 32 bit seems to run fine with no vt-x...

Very strange problem. Memtest, Prime95 all return positive results. The Intel CPU diagnostic tool returns no errors at all.

[satya1]

HI what is exact problem and why you want to activate VTX ?

Yours,

Satya

[RDPetruska]

satya wrote:

HI what is exact problem and why you want to activate VTX ?

So that he can run a 64-bit guest.

[admin]

You mention a cryptic error message.  Have you been able to reproduce that?

[sinni800]

Well I have gotten a few of those, ALL right in the BIOS of the guest system:

- 0x00000101 bluescreen (WATCHDOG TIMEOUT) using VirtualBox

- 0x00000124 WHEA_UNCORRECTABLE_ERROR using VmWare

- unknown Kernel Crash Panic using Linux VMware

Sometimes VMware manages to catch the error and close the VM. Same as VirtualBox.

I can always reproduce them

[admin]

One interesting experiment would be to disable EPT.  Go to the processors tab and select "VT-x or AMD-V" as the preferred execution mode.  See if that makes any difference.

[sinni800]

I tried it with both execution modes and both crash to BIOS (no bluescreen this time. It's random if I get one or not). Just now even binary translation made VMWare error (with "virtualize ... EPT " or something activated... Its the checkbox).

It gave me a log which had an error somewhere in the mdidle of the file ans also a few dumpfiles this time, here is the error:

Vmware monitor panic: vcpu-0:VT resume failed (error 5)

Pretty strange, huh?

[admin]

sinni800 wrote:

Vmware monitor panic: vcpu-0:VT resume failed (error 5)

This can be a sign of non-maskable interrupts occurring where we can't handle them.  If you boot your host into Linux, does /proc/interrupts show a high rate of NMI activity?

[sinni800]

I won't have access to my PC until monday, so bear with me until then. I will edit this post on my findings on monday!

EDIT:

I have had time to check. Here is the output of /proc/interrupts after about 5 minutes of using the pc on Linux.

http://pastebin.com/SJrit8JL

I don't know what you need here... Non-maskable interrupts?