Windows HVM cannot auto reboot after cashdump with more then 1 VCPU

Windows HVM cannot auto reboot after cashdump with more then 1 VCPU

[Wayne Gong]

[-- Attachment #1.1: Type: text/plain, Size: 447 bytes --]

Hi folks

I meet a problem when implement crash dump feature of winpv drives for 
xen. If a windows HVM (without winpv driver) use more then 1 VCPU, it 
cannot reboot automatically even I set on_crash = 'restart' and 
on_reboot = 'restart' in vm.cfg file.  If I set VCPU to 1, it works 
fine. Any suggestion?

http://communities.vmware.com/thread/188897 will tell you how to course 
a BSOD and crash dump manually.

I use Xen 3.1.4.

Thanks
Wayne

[-- Attachment #1.2: Type: text/html, Size: 746 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

RE: Windows HVM cannot auto reboot after cashdump with morethen 1 VCPU

[James Harper]

> Hi folks
> 
> I meet a problem when implement crash dump feature of winpv drives for
> xen. If a windows HVM (without winpv driver) use more then 1 VCPU, it
> cannot reboot automatically even I set on_crash = 'restart' and
on_reboot
> = 'restart' in vm.cfg file.  If I set VCPU to 1, it works fine. Any
> suggestion?

Does it write out a crash dump file without winpv drivers loaded and
with VCPU > 1

> http://communities.vmware.com/thread/188897 will tell you how to
course a
> BSOD and crash dump manually.
> 
> I use Xen 3.1.4.
> 

My drivers can use 'xm sysrq B' to cause a crash dump :)

The crash dump path is pretty fiddly to get right, but is that the
problem you are trying to solve right now?

James

Re: Windows HVM cannot auto reboot after cashdump with more then 1 VCPU

[Tim Deegan]

At 06:04 -0500 on 11 Feb (1234332247), Wayne Gong wrote:
> Hi folks
> 
> I meet a problem when implement crash dump feature of winpv drives for xen. If a windows HVM (without winpv driver) use more then 1 VCPU, it cannot reboot automatically even I set on_crash = 'restart' and on_reboot = 'restart' in vm.cfg file.  If I set VCPU to 1, it works fine. Any suggestion?

ISTR Xen 3.1 didn't handle inter-vcpu NMIs properly, so multi-vcpu
windows guests would fail to shut down in some cases (where Windows was
using NMIs to halt its APs for restart); instead the guest would hang.

Cheers,

Tim.
 
> http://communities.vmware.com/thread/188897 will tell you how to course a BSOD and crash dump manually.
> 
> I use Xen 3.1.4.
> 
> Thanks
> Wayne

Content-Description: ATT00001.txt
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel


-- 
Tim Deegan <Tim.Deegan@citrix.com>
Principal Software Engineer, Citrix Systems (R&D) Ltd.
[Company #02300071, SL9 0DZ, UK.]

Re: Windows HVM cannot auto reboot after cashdump with morethen 1 VCPU

[Wayne Gong]

[-- Attachment #1.1: Type: text/plain, Size: 266 bytes --]

>
> Does it write out a crash dump file without winpv drivers loaded and
> with VCPU > 1
>
>   

Yes.
> The crash dump path is pretty fiddly to get right, but is that the
> problem you are trying to solve right now?
>   

Yes, I think I will if I can.

Thanks
Wayne

[-- Attachment #1.2: Type: text/html, Size: 748 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Windows HVM cannot auto reboot after cashdump with more then 1 VCPU

[Sönke Schwardt]

[-- Attachment #1.1: Type: text/plain, Size: 1426 bytes --]

Hi!

Am Mittwoch, 11. Februar 2009 12:09:56 schrieb Tim Deegan:
> At 06:04 -0500 on 11 Feb (1234332247), Wayne Gong wrote:
> > Hi folks
> >
> > I meet a problem when implement crash dump feature of winpv drives for
> > xen. If a windows HVM (without winpv driver) use more then 1 VCPU, it
> > cannot reboot automatically even I set on_crash = 'restart' and
> > on_reboot = 'restart' in vm.cfg file.  If I set VCPU to 1, it works
> > fine. Any suggestion?
>
> ISTR Xen 3.1 didn't handle inter-vcpu NMIs properly, so multi-vcpu
> windows guests would fail to shut down in some cases (where Windows was
> using NMIs to halt its APs for restart); instead the guest would hang.

Is there any workaround on guest or host side? 
We can reproduce a hanging win2k3 guest on normal windows reboot if 4 vcpus 
are used. The guest is switched off correctly after windows shutdown and 
reboots without problem if only 2 vcpus are assigned to this guest.

Thanks

 Sönke

-- 
**Besuchen Sie uns vom 03.03.-08.03.2009 auf der CeBIT in Hannover**
                        *Halle 6 Stand G41*

Sönke Schwardt
Open Source Software Engineer

Univention GmbH
Linux for your business
Mary-Somerville-Str.1
28359 Bremen
Tel. : +49 421 22232-40
Fax : +49 421 22232-99

schwardt@univention.de
http://www.univention.de

Geschäftsführer: Peter H. Ganten
HRB 20755 Amtsgericht Bremen
Steuer-Nr.: 71-597-02876

[-- Attachment #1.2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Windows HVM cannot auto reboot after cashdump with more then 1 VCPU

[Tim Deegan]

At 14:50 +0000 on 04 Mar (1236178249), S?nke Schwardt wrote:
> > ISTR Xen 3.1 didn't handle inter-vcpu NMIs properly, so multi-vcpu
> > windows guests would fail to shut down in some cases (where Windows was
> > using NMIs to halt its APs for restart); instead the guest would hang.
> 
> Is there any workaround on guest or host side? 

Not that I know of.  Upgrade to Xen 3.2? :)

> We can reproduce a hanging win2k3 guest on normal windows reboot if 4 vcpus 
> are used. The guest is switched off correctly after windows shutdown and 
> reboots without problem if only 2 vcpus are assigned to this guest.

Huh.  I would have expeceted it to use the same reboot code on all SMP
configurations.  But the ways of Windows are indeed strange.

Cheers,

Tim.

-- 
Tim Deegan <Tim.Deegan@citrix.com>
Principal Software Engineer, Citrix Systems (R&D) Ltd.
[Company #02300071, SL9 0DZ, UK.]

Re: Windows HVM cannot auto reboot with more than 2 VCPU

[Sönke Schwardt]

[-- Attachment #1.1: Type: text/plain, Size: 2214 bytes --]

Hi!

Am Donnerstag, 5. März 2009 15:20:19 schrieb Tim Deegan:
> > We can reproduce a hanging win2k3 guest on normal windows reboot if 4
> > vcpus are used. The guest is switched off correctly after windows
> > shutdown and reboots without problem if only 2 vcpus are assigned to
> > this guest.
>
> Huh.  I would have expeceted it to use the same reboot code on all SMP
> configurations.  But the ways of Windows are indeed strange.

Yes, indeed.

In the meantime I had access to that specific system and gathered some 
additional information:
- xen 3.2 / kernel 2.6.18
- problem happens with Win2003 Enterprise Edition if 3 or 4 vcpus 
  are assigned and reboot has been initiated via start menu.
- happens with and without windows updates
- happens with and without Windows ACPI HAL (acpi=1 / acpi=0)
- no virus scanners installed
- problem does NOT happen if restart has been initiated via external
  tools like daemontools (using daemontools is no real workaround since 
  this system has to install new windows updates during night hours and
  related automatic reboots causing that hanging guest)

In case of error one vcpu is busy (80% to 120% cpu usage):
xm> vcpu-list
Name      ID  VCPU CPU State Time(s) CPU Affinity
Domain-0   0  0    4   r--   58905.6 any cpu
xen01     38  0    7   r--     180.2 any cpu
xen01     38  1    3   -b-      17.3 any cpu
xen01     38  2    2   -b-      15.7 any cpu
xen01     38  3    -   --p      12.9 any cpu
xm> top
   NAME  STATE   CPU(sec) CPU(%)     MEM(k) MEM(%)  MAXMEM(k) MAXMEM(%) VCPUS NETS NETTX(k) NETRX(k) VBDS   VBD_OO   VBD_RD   VBD_WR SSID
xnts008 -----r        276  112.7    1056640    3.1    1064960      3.2     4    1        0        0    1        0        0        0  2149957416

Any ideas for a solution? Can I provide additional information to 
track this down?

Regards

 Sönke
-- 
Sönke Schwardt
Open Source Software Engineer

Univention GmbH
Linux for your business
Mary-Somerville-Str.1
28359 Bremen
Tel. : +49 421 22232-40
Fax : +49 421 22232-99

schwardt@univention.de
http://www.univention.de

Geschäftsführer: Peter H. Ganten
HRB 20755 Amtsgericht Bremen
Steuer-Nr.: 71-597-02876

[-- Attachment #1.2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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