This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


RE: [Xen-users] How do I find out why Dom0 crashes on guest startup?

To: "Schober Walter" <Walter.Schober@xxxxxxxxx>, xen-users@xxxxxxxxxxxxxxxxxxx
Subject: RE: [Xen-users] How do I find out why Dom0 crashes on guest startup?
From: "Petersson, Mats" <Mats.Petersson@xxxxxxx>
Date: Fri, 18 May 2007 14:36:25 +0200
Delivery-date: Fri, 18 May 2007 05:35:01 -0700
Envelope-to: www-data@xxxxxxxxxxxxxxxxxx
In-reply-to: <30FA315AB19B5F49B7B7C5CA912B081104F646@xxxxxxxxxxxxx>
List-help: <mailto:xen-users-request@lists.xensource.com?subject=help>
List-id: Xen user discussion <xen-users.lists.xensource.com>
List-post: <mailto:xen-users@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=unsubscribe>
Sender: xen-users-bounces@xxxxxxxxxxxxxxxxxxx
Thread-topic: [Xen-users] How do I find out why Dom0 crashes on guest startup?

> -----Original Message-----
> From: Schober Walter [mailto:Walter.Schober@xxxxxxxxx] 
> Sent: 18 May 2007 13:10
> To: xen-users@xxxxxxxxxxxxxxxxxxx
> Cc: Petersson, Mats
> Subject: RE: [Xen-users] How do I find out why Dom0 crashes 
> on guest startup?
> Tested:
> 1. apic=0, vcpus=4     OK, no crash
> 2. apic=1, vcpus=1     OK, no crash
> 3. apic=1, vcpus=2     OK (?!?!?) But:
> (XEN) This hvm_vlapic is for P4, no work for De-assert init
> (XEN) (file=hvm.c, line=665) AP 1 bringup suceeded.
> (XEN) vmx_do_launch(): GUEST_CR3<=00fbd7a0, HOST_CR3<=cc3ae000
> (XEN) (GUEST: 4) Start AP 1 from 00006000 ...
> (XEN) (GUEST: 4) Starting emulated 16-bit real-mode: ip=0600:0000
> 4. apic=1, vcpus=4     NOK, but no crash anymore ?!?!?
> A lot of 
> (XEN) <ioapic_inj_irq> error delivery mode 7
> and on the guest:
> ..MP-BIOS bug: 8254 timer not connected to IO-APIC
> Kenrel panic - not syncing: IO-APIC + timer doesn't work! Try 
> using the
> noapic kernel parameter
> 5. apic=1, vcpus=4, 'xm create' right after reboot    NOK, Dom0
> crashes as before
> 6. starting at least one another VM with apic=0 or commented 
> out apic ->
> See 3. + 4., but no Dom0 crash.

Interesting behaviour. 

There was a change between 3.0.3 and 3.0.4 that changed the way that the
vlapic info is stored in connection to the VCPU, and I suspect that the
problem is related to a race of "adding the vlapic to the VCPU" and
"using vlapic from the vcpu". Not sure why that should be, but I suspect
that's the problem. 

> => Dom0 crashes only, if the hvm wasn't started with apic=0 
> once before.
> That brings me back to time when I installed that VM: Installation was
> possible only with vcpus=1 (stale after Centos 5 Install 
> Screen). After
> that I set vcpus=4 and rebootet the VM.
> On Wednesday I had to reboot the Dom0 for that HP management thing and
> started the VM the first time directly with apic=1, vcpus=4 
> causing the
> crash. So it had nothing to do with the HP drivers nor QEMU package.
> I guess, I have to wait for a fix in the Xen kernel or maybe try to
> compile my own 3.0.4.

I'd try the xen-3.1-testing.hg repository - it's closed for anything but
really critical updates now, so it's pretty much the same as final
version. That's the latest availble (aside from "unstable", which I
wouldn't recommend unless you find that 3.1 doesn't work either). 

> Thanks, Mats for you help!
> br
> Walter
> -----Original Message-----
> From: Petersson, Mats [mailto:Mats.Petersson@xxxxxxx] 
> Sent: Friday, May 18, 2007 1:22 PM
> To: Schober Walter; xen-users@xxxxxxxxxxxxxxxxxxx
> Subject: RE: [Xen-users] How do I find out why Dom0 crashes on guest
> startup?
> [snip big log]
> > (XEN) Xen call trace:
> > (XEN)    [<ffff830000142da7>] vlapic_range+0x7/0x30
> So, the code here does very few things (at least in the 
> current version,
> but I doubt that the older code is MUCH different). This in turn means
> that it's pretty easy to determine what could be wrong:
> 1. VCPU = NULL - quite likely, as it's one of the first steps. 
> 2. vlapic field in vcpu = NULL - even more likely. 
> I'd try running your guest with "apic=0" to see if that makes any
> difference. [That will of course screw up SMP on your guest, but at
> least it will hopefully tell us if it's APIC that is a problem or not]
> I doubt that having installed a newer QEMU would make any difference. 
> --
> Mats

Xen-users mailing list