Re: [Qemu-devel] [PATCH 2/6] pci-assign: Add support for no-route

[Alex Williamson]

On Tue, 2012-10-02 at 14:55 -0500, Anthony Liguori wrote:
> Alex Williamson <address@hidden> writes:
> 
> > In the event that a pci-assign device is added to a chipset that
> > hasn't yet implemented the INTx routing interface, exit gracefully
> > instead of killing the VM.  I'm sure we'll implement this for q35,
> > but there's no reason for such a harsh response.
> >
> > Signed-off-by: Alex Williamson <address@hidden>
> > ---
> >
> >  hw/kvm/pci-assign.c |    6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> > diff --git a/hw/kvm/pci-assign.c b/hw/kvm/pci-assign.c
> > index 05b93d9..7ce0f37 100644
> > --- a/hw/kvm/pci-assign.c
> > +++ b/hw/kvm/pci-assign.c
> > @@ -887,6 +887,12 @@ static int assign_intx(AssignedDevice *dev)
> >  
> >      intx_route = pci_device_route_intx_to_irq(&dev->dev, dev->intpin);
> >      assert(intx_route.mode != PCI_INTX_INVERTED);
> > +    if (intx_route.mode == PCI_INTX_NOROUTE) {
> > +        error_report("pci-assign: chipset provides no INTx routing "
> > +                     "information, but device supports INTx interrupt 
> > mode.\n");
> > +        pci_device_set_intx_routing_notifier(&dev->dev, NULL);
> > +        return -ENOTSUP;
> > +    }
> 
> Please don't use error_report() in new code.  Propagate an Error object
> using error_setg() and friends.

That doesn't really seem to be an option since this function is called
both from the driver initfn and the interrupt routing change notifier.
Neither of those give me an Error object... suggestions?  Thanks,

Alex