[libvirt] [PATCH 0/8] Various KVM PCI device assignment improvements

[Mark McLoughlin]

On Fri, 2009-08-14 at 16:15 +0100, Daniel P. Berrange wrote:
> On Fri, Aug 14, 2009 at 04:09:32PM +0100, Mark McLoughlin wrote:
> > On Fri, 2009-08-14 at 16:41 +0200, Mirko Raasch wrote:
> > > Hi,
> > > 
> > > thanks for you improvements.
> > > 
> > > I updated libvirt with git and tried to passthrough two pci devices to a 
> > > linux guest.
> > > 
> > >  <hostdev mode='subsystem' type='pci' managed='yes'>
> > >       <source>
> > >         <address domain='0x0000' bus='0x05' slot='0x01' function='0x0'/>
> > >       </source>
> > >     </hostdev>
> > >     <hostdev mode='subsystem' type='pci' managed='yes'>
> > >       <source>
> > >         <address domain='0x0000' bus='0x05' slot='0x03' function='0x0'/>
> > >       </source>
> > >     </hostdev>
> > > 
> > > In the guest, only one of them works and after the shutdown of the 
> > > guest, i got a segfault on the host.
> > > 
> > > libvirtd[15363]: segfault at 38 ip 0000000000421164 sp 00007fff59d1dfb0 
> > > error 4 in libvirtd[400000+6d000]
> > > 
> > > I use kernel 2.6.30-4 on the host.
> > > 
> > > What did i wrong?
> > 
> > It's more a question of what *I* did wrong, probably :-)
> > 
> > Did this work before? With what version? Is there any chance you could
> > use git-bisect to find out what commit introduced the regression?
> > 
> > If not, please run libvirtd from the command line with LIBVIRT_DEBUG and
> > post the log file somewhere. Also, the guest log file
> > from /var/log/libvirt/qemu would help
> 
> For segfault crashes, your best bet is actually to just run libvirtd
> under valgrind. That usually narrows down the problem more quickly
> than the debug modes

Good point - I actually meant to ask Mirko to run it in gdb and get a
stack trace.

Cheers,
Mark.