Re: [xen-4.0.1-rc5-pre] [pvops 2.6.32.16] Complete freeze within 2 days, no info in serial log

[Jeremy Fitzhardinge <jeremy@goop.org>]

  On 08/03/2010 08:45 AM, Konrad Rzeszutek Wilk wrote:
> On Tue, Aug 03, 2010 at 05:30:57PM +0200, Sander Eikelenboom wrote:
>> Hi All,
>>
>> I'm experiencing for what it seems a random freeze with current xen-4.0-testing, pvops dom0 2.6.32.16 kernel, most of the time within 2 days after rebooting.
>>
> You did not experience the freeze with 2.6.32.15?

There have been a few updates to the .32.16 kernel too (and now its 
.17...).  But it would be very useful to identify which the last working 
kernel was.

>> Symptoms:
>> - Complete freeze, only power cycle does work.
>> - No bug output/stacktrace in serial log / on screen.
>> - Not able to get into hypervisor with ctrl-a (doesn't react to keyboard)
>> - No info in syslog.
>>
>> Are there any more boot options I could give a try in the hope it will give some debug output ?
> The Linux kernel has some of those 'DETECT_SPINLOCK_HANG' or
> 'DETECT_WORK..something' flags. It might be a good idea to compile those
> and see when your machine freezes if after 2 minutes the kernel starts
> spitting out what is hung. That could give some idea.
>

If Xen doesn't respond then it isn't a kernel spinlock problem; it looks 
more system-wide than that.  I notice the kernel command line has lots 
of hidden PCI devices.  Sander, is there any particular activity (esp 
passthrough device activity) which might correspond to the hang?

     J