Re: [Bug 104091] New: [bisected] Starting a VM causes the host to halt and create Machine Check Exceptions

[Xiao Guangrong <guangrong.xiao@linux.intel.com>]

On 09/06/2015 06:03 PM, bugzilla-daemon@bugzilla.kernel.org wrote:
> https://bugzilla.kernel.org/show_bug.cgi?id=104091
>
>              Bug ID: 104091
>             Summary: [bisected] Starting a VM causes the host to halt and
>                      create Machine Check Exceptions
>             Product: Virtualization
>             Version: unspecified
>      Kernel Version: 4.2
>            Hardware: All
>                  OS: Linux
>                Tree: Mainline
>              Status: NEW
>            Severity: normal
>            Priority: P1
>           Component: kvm
>            Assignee: virtualization_kvm@kernel-bugs.osdl.org
>            Reporter: harn-solo@gmx.de
>          Regression: No
>
> Created attachment 186851
>    --> https://bugzilla.kernel.org/attachment.cgi?id=186851&action=edit
> dmesg output after starting the VM
>
> With kernel 4.2, starting one of my VMs instantly freezes the host system and
> creates Machine Check Exceptions on CPUs dedicated to that particula VM:
>
> [12316.171917] mce: [Hardware Error]: CPU 3: Machine Check Exception: 5 Bank
> 17: be2000000003110a
> [12316.171917] mce: [Hardware Error]: RIP !INEXACT! 10:<ffffffff813217fd>
> {intel_idle+0xbd/0x120}
> [12316.171917] mce: [Hardware Error]: TSC 76fd7352bf6 ADDR fa137140 MISC
> 30f0083884509086
> [12316.171917] mce: [Hardware Error]: PROCESSOR 0:306f2 TIME 1441130705 SOCKET
> 0 APIC 6 microcode 2d
> [12316.171917] mce: [Hardware Error]: Run the above through 'mcelog --ascii'
> ...
>

Could you please decode it by mcelog?

> A bisection revealed that commit fd717f11015f673487ffc826e59b2bad69d20fe5
> introduced the problem:
>
> KVM: x86: apply guest MTRR virtualization on host reserved pages
>
> Currently guest MTRR is avoided if kvm_is_reserved_pfn returns true.
> However, the guest could prefer a different page type than UC for
> such pages. A good example is that pass-throughed VGA frame buffer is
> not always UC as host expected.
>
> This patch enables full use of virtual guest MTRRs.
>
> One could argue that the following warning is an obvious hint
> [12311.584431] pmd_set_huge: Cannot satisfy [mem 0x383fe0000000-0x383fe0200000]
> with a huge-page mapping due to MTRR override.

Just confirm: it happened on host, right?

>
> but I'm able to run another VM without problems despite that warning.
>
> Please let me know I you need additional information.
>