Re: [Qemu-devel] [PATCH v1] s390x/pci: Warn when adding PCI devices without the 'zpci' feature

[Cornelia Huck <cohuck@redhat.com>]

On Tue, 22 Jan 2019 14:25:07 +0100
David Hildenbrand <david@redhat.com> wrote:

> On 22.01.19 14:23, Cornelia Huck wrote:
> > On Tue, 22 Jan 2019 14:20:27 +0100
> > David Hildenbrand <david@redhat.com> wrote:
> >   
> >> On 22.01.19 14:13, Cornelia Huck wrote:  
> >>> On Tue, 22 Jan 2019 11:06:46 +0100
> >>> David Hildenbrand <david@redhat.com> wrote:
> >>>     
> >>>> On 22.01.19 10:50, Thomas Huth wrote:    
> >>>>> On 2019-01-22 10:41, David Hildenbrand wrote:      
> >>>>>> We decided to always create the PCI host bridge, even if 'zpci' is not
> >>>>>> enabled (due to migration compatibility).      
> >>>>>
> >>>>> Couldn't we disable the host bridge for newer machine types, and just
> >>>>> create it on the old ones for migration compatibility?      
> >>>
> >>> I very dimly remember some problems with that approach.
> >>>     
> >>>>
> >>>> I think we can with a compat property. However I somewhat dislike that
> >>>> the error/warning will then be "no bus" vs. "zpci CPU feature not
> >>>> enabled". Somebody who has no idea about that will think he somehow has
> >>>> to create a PCI bus on the QEMU comandline.    
> >>>
> >>> Agreed, "zpci cpu feature not enabled" gives a much better clue.
> >>>     
> >>>>
> >>>> ... however
> >>>>    
> >>>>>       
> >>>>>> This however right now allows
> >>>>>> to add zPCI/PCI devices to a VM although the guest will never actually see
> >>>>>> them, confusing people that are using a simple CPU model that has no
> >>>>>> 'zpci' enabled - "Why isn't this working" (David Hildenbrand)
> >>>>>>
> >>>>>> Let's check for 'zpci' and at least print a warning that this will not
> >>>>>> work as expected. We could also bail out, however that might break
> >>>>>> existing QEMU commandlines.
> >>>>>>
> >>>>>> Signed-off-by: David Hildenbrand <david@redhat.com>
> >>>>>> ---
> >>>>>>  hw/s390x/s390-pci-bus.c | 5 +++++
> >>>>>>  1 file changed, 5 insertions(+)
> >>>>>>
> >>>>>> diff --git a/hw/s390x/s390-pci-bus.c b/hw/s390x/s390-pci-bus.c
> >>>>>> index b86a8bdcd4..e7d4f49611 100644
> >>>>>> --- a/hw/s390x/s390-pci-bus.c
> >>>>>> +++ b/hw/s390x/s390-pci-bus.c
> >>>>>> @@ -863,6 +863,11 @@ static void s390_pcihost_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
> >>>>>>  {
> >>>>>>      S390pciState *s = S390_PCI_HOST_BRIDGE(hotplug_dev);
> >>>>>>  
> >>>>>> +    if (!s390_has_feat(S390_FEAT_ZPCI)) {
> >>>>>> +        warn_report("Adding PCI or zPCI devices without the 'zpci' CPU feature."
> >>>>>> +                    " The guest will not be able to see/use these devices.");
> >>>>>> +    }      
> >>>>>
> >>>>> I think it would be better to bail out. The hotplug clearly can not work
> >>>>> in this case, and the warn report might go unnoticed, so blocking the
> >>>>> hotplug process is likely better to get the attention of the user.      
> >>>>
> >>>> ... we could also create the bus but bail out here in case the compat
> >>>> property strikes (a.k.a. new QEMO machine type).    
> >>>
> >>> Now you confused me... why should failing be based on a compat property?
> >>>     
> >>
> >> Otherwise, a QEMU comandline that used to work (which could be created
> >> by libvirt) would now fail. Are we ok with that?
> >>  
> > 
> > I think we should not fail at all in that case, then. Or only for
> > hotplug, not for coldplug.
> >   
> 
> We could fail on hotplug and warn on coldplug. This would keep existing
> setups running.
> 

Ok with me.