Re: [Qemu-devel] Re: irq problems after live migration with 0.12.4

[Peter Lieven <pl@dlh.net>]

Michael Tokarev wrote:
> 25.05.2010 15:03, Peter Lieven wrote:
>> Michael Tokarev wrote:
>>> 23.05.2010 13:55, Peter Lieven wrote:
> []
>>>> [64442.298521] irq 10: nobody cared (try booting with the "irqpoll" 
>>>> option)
> []
>>>> [64442.299433] handlers:
>>>> [64442.299840] [<ffffffffa0000b80>] (e1000_intr+0x0/0x190 [e1000])
>>>> [64442.300046] Disabling IRQ #10
>
> Apparently, for some reason, e1000_intr decided it's not
> interesting IRQ or somehow wrong or not for that NIC.  I
> dunno.  But something fishy is going on with IRQs here.
>
>>> See also LP bug #584131 (https://bugs.launchpad.net/bugs/584131)
>>> and original Debian bug#580649 (http://bugs.debian.org/580649)
>
>>> Not sure if they're related...
>
> It looks they are actually the same thing, but happens with
> different devices and/or IRQs.  Either spurious, or unwanted,
> or unrecognized or somesuch IRQ which is not recognized by
> the irq handler, which results in disabling that IRQ by the
> kernel, which is a bad thing (In your case it works because
> e1000 works in 2 modes, interrupts and polling).
>
>> michael, do you have any ideas what i got do to debug whats happening?
>
> Unfortunately, no idea.  I don't know neither kernel nor kvm
> internals.
I would be very greatful if someone with deeper knowledge would hook up.
I'm also not familiar with internals, unfortunately.
>
>> looking at launchpad and debian bug tracker i found other bugs also
>> with a maybe related problem. so this issue might be greater...
>
> Can you share your findings?  I don't know other debian bugs which
> are similar to this one.
I suspect that other reports regarding crashed VMs after migration
might be related. If I take my test VM with that I can trigger the bug
and change the Network Adapter from e1000 to rtl8139 and leave
everything else untouched the VM hangs at 100% CPU..



>
> Thanks!
>
> /mjt
>
>