Re: xen-hypervisor-4.3-amd64 4.3.0-3+b1 -> Intel M/B, I/O virt. enabled, start vm -> Kernel panic

[David Vrabel <david.vrabel@citrix.com>]

On 20/05/14 19:41, Jo Mills wrote:
> On Tue, May 20, 2014 at 05:20:10PM +0100, David Vrabel wrote:
>> On 20/05/14 11:18, Ian Campbell wrote:
>>> On Tue, 2014-05-20 at 11:14 +0100, David Vrabel wrote:
>>>> On 17/05/14 18:46, Jo Mills wrote:
>>>>>
>>>>>     [  498.336430] ---[ end trace b9630577ecf84cd8 ]---
>>>>>     [  498.340053] Kernel panic - not syncing: Fatal exception in interrupt
>>>>>     (XEN) Domain 0 crashed: 'noreboot' set - not rebooting.
>>>>
>>>> We need the full backtrace from the kernel.
>>>
>>> It was in the attached serial console log. I've pasted what I think is
>>> the relevant bit below.
>>
>> Ah. I hadn't noticed the attachment.
>>
>>> [  485.048066] d-con vm-13-disk2: PingAck did not arrive in time.
>>> [  486.876191] e1000 0000:04:02.0 eth2: Detected Tx Unit Hang
>>> [  486.876191]   Tx Queue             <0>
>>> [  486.876191]   TDH                  <a5>
>>> [  486.876191]   TDT                  <b5>
>>> [  486.876191]   next_to_use          <b5>
>>> [  486.876191]   next_to_clean        <a5>
>>> [  486.876191] buffer_info[next_to_clean]
>>> [  486.876191]   time_stamp           <10000b439>
>>> [  486.876191]   next_to_watch        <a6>
>>> [  486.876191]   jiffies              <10000b67f>
>>> [  486.876191]   next_to_watch.status <0>
>>
>> This looks like an e1000 driver/hardware problem.  It doesn't
>> immediately look Xen specific.
>>
>> David
> 
> Hi David et al,
> 
> Many thanks for your replies and for your time spent looking into this 
> problem.  
> 
> I probably should have mentioned in my report that when BIOS settings:
> 
>     Intel(R) Virtualization Technology
>     Intel(R) VT for Directed I/O
>     
> are enabled, creating VMs that do not use a pci passthrough device 
> works just fine.  It's only when I try and create a VM that does use a 
> pci passthrough device that the crash happens.  Whether this is 
> relevant or not I'm not competent enough to say.

It's not obviously related since it's a device/driver in dom0 that's
broken.  I would suggest asking the e1000 maintainers if the Tx Unit
Hang debug output hints at anything.

David