Re: [Qemu-devel] [PATCH 2/3] hw/virtio-net.c: set config size using host features

[Anthony Liguori <aliguori@us.ibm.com>]

"Michael S. Tsirkin" <mst@redhat.com> writes:

> On Thu, Mar 07, 2013 at 10:44:18AM -0600, Anthony Liguori wrote:
>> Andreas Färber <afaerber@suse.de> writes:
>> 
>> > Am 07.03.2013 17:27, schrieb Christian Borntraeger:
>> >>> It's a bug in both virtio-ccw that features=0 when get_features is
>> >>> called.  You can also tell this with:
>> >>>
>> >>> [10:02 AM] anthony@titi:~/git/qemu/hw/s390x$ grep DEFINE_VIRTIO_NET_FEATURES *
>> >>> virtio-ccw.c:    DEFINE_VIRTIO_NET_FEATURES(VirtioCcwDevice, host_features[0]),
>> >>>
>> >>> So virtio-s390 is doing it wrong, but virtio-ccw looks like its doing it
>> >>> right.
>> >> 
>> >> At least, this patch seems to work. (That also implies, that a transport
>> >> must not hide virtio feature bits).
>> >
>> > To me it indicates that the use of the old qdev property setters is
>> > hiding errors resulting from trying to set not-existing properties.
>> > If we would set the properties in a way that gets us an Error* on
>> > failure like the object_property_set_*() do, we would notice on machine
>> > creation (or device_add).
>> 
>> Hrm, I don't understand your statement.
>> 
>> Can you elaborate?
>> 
>> Regards,
>> 
>> Anthony Liguori
>
>
> Well to me this indicates that s390 virtio is buggy.
> It's not supposed to crash whatever set of features
> we specify.

Ack.

Regards,

Anthony Liguori