Re: HVM crash system on AMD APU A8-6600K

[Vitaliy Tomin <highwaystar.ru@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 2476 bytes --]

Ok, after 3 weeks waiting for reply from Asrock they still no reply
and I've tried bios downgrading. After I downrade BIOS from P2.40
(latest) to P2.20 xen and linux kernel showed all CPU cores have same
microcode version:

(XEN) masked ExtINT on CPU#1
(XEN) microcode: CPU1 collect_cpu_info: patch_id=0x6001119
(XEN) masked ExtINT on CPU#2
(XEN) microcode: CPU2 collect_cpu_info: patch_id=0x6001119
(XEN) masked ExtINT on CPU#3
(XEN) microcode: CPU3 collect_cpu_info: patch_id=0x6001119
(XEN) Brought up 4 CPUs

But still no message about CPU#0. With Linux kernel all 4 cores now
have same microcode version.
Next i've tested xen hvm with hvm guests and it looks stable now, at
least no crashes after 2 hours.

>(XEN) mtrr: your CPUs had inconsistent variable MTRR settings
Still appears in console log. Log with downgraded BIOS attached.

On Fri, Apr 11, 2014 at 10:30 PM, Jan Beulich <JBeulich@suse.com> wrote:
>>>> On 11.04.14 at 13:57, <highwaystar.ru@gmail.com> wrote:
>>>  Can you get logs from a debug and non debug build of Xen, with at least
>>> "loglvl=all guest_loglvl=all hap=0 iommu=0" on the command line.  For
>>> the non debug version, can you boot and shut down the HVM guest.
>>>
>>> Given the (hopefully correct) disabling of hap and the iommu, we are
>>> left with AMD SVM left as the likely suspect for problems.
>>
>> Logs as you requested attached. Both logs made with 3-lines hvm guest.
>
> There's a very obvious problem here, whether or not related to your
> issue:
>
> (XEN) microcode: CPU1 collect_cpu_info: patch_id=0x6001119
> (XEN) masked ExtINT on CPU#2
> (XEN) microcode: CPU2 collect_cpu_info: patch_id=0
> (XEN) masked ExtINT on CPU#3
> (XEN) microcode: CPU3 collect_cpu_info: patch_id=0
> (XEN) Brought up 4 CPUs
>
> I.e.
> - CPU0 isn't being mentioned at all (this would normally be immediately
>   after the "Allocated console ring ..." message)
> - CPU2 and CPU3 run a different microcode update that CPU1
>
> Please find a suitable microcode update blob for your system and
> pass it to xen (making use of the "ucode=" option).
>
> And this
>
> (XEN) mtrr: your CPUs had inconsistent variable MTRR settings
> (XEN) mtrr: probably your BIOS does not setup all CPUs.
> (XEN) mtrr: corrected configuration.
>
> suggests further firmware problems. I'm afraid we need to rely
> on you to get the vendor to supply you something at least half
> way sane before it makes sense to further investigate this rater
> irritating problem.
>
> Jan
>

[-- Attachment #2: screenlog.0-p2.20 --]
[-- Type: application/octet-stream, Size: 78021 bytes --]

[-- Attachment #3: Type: text/plain, Size: 126 bytes --]

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel