Re: [Qemu-devel] [PATCH 2/4] s390-virtio-bus: move common virtio properties to virtio s390 device class

[Cornelia Huck]

On Tue, 17 Jun 2014 09:07:41 +0200
Cornelia Huck <address@hidden> wrote:

> On Tue, 17 Jun 2014 10:44:11 +0800
> Ming Lei <address@hidden> wrote:
> 
> > On Tue, Jun 17, 2014 at 12:04 AM, Cornelia Huck
> > <address@hidden> wrote:
> > > On Mon, 16 Jun 2014 23:40:50 +0800
> > > Ming Lei <address@hidden> wrote:
> > >
> > >> The two common virtio features can be defined per bus, so move all
> > >> into virtio-s390 class device to make code more clean.
> > >>
> > >> Suggested-by: Paolo Bonzini <address@hidden>
> > >> Signed-off-by: Ming Lei <address@hidden>
> > >> ---
> > >>  hw/s390x/s390-virtio-bus.c |   15 ++++++---------
> > >>  1 file changed, 6 insertions(+), 9 deletions(-)
> > >
> > > This one breaks for me:
> > >
> > > qemu-system-s390x: -device 
> > > virtio-blk-s390,scsi=off,config-wce=off,drive=drive-virtio-disk0,id=virtio-disk0,bootindex=1:
> > >  Property '.scsi' not found
> > >
> > >>
> > >> diff --git a/hw/s390x/s390-virtio-bus.c b/hw/s390x/s390-virtio-bus.c
> > >> index 9c71afa..ab9758e 100644
> > >> --- a/hw/s390x/s390-virtio-bus.c
> > >> +++ b/hw/s390x/s390-virtio-bus.c
> > >> @@ -526,18 +526,11 @@ static const TypeInfo s390_virtio_net = {
> > >>      .class_init    = s390_virtio_net_class_init,
> > >>  };
> > >>
> > >> -static Property s390_virtio_blk_properties[] = {
> > >> -    DEFINE_VIRTIO_BLK_PROPERTIES(VirtIOBlkS390, blk),
> > >> -    DEFINE_PROP_END_OF_LIST(),
> > >> -};
> > >> -
> > >>  static void s390_virtio_blk_class_init(ObjectClass *klass, void *data)
> > >>  {
> > >> -    DeviceClass *dc = DEVICE_CLASS(klass);
> > >>      VirtIOS390DeviceClass *k = VIRTIO_S390_DEVICE_CLASS(klass);
> > >>
> > >>      k->init = s390_virtio_blk_init;
> > >> -    dc->props = s390_virtio_blk_properties;
> > >
> > > ...which is probably because you removed the block properties here.
> > 
> > You are right, DEFINE_VIRTIO_BLK_PROPERTIES() should have been
> > kept.
> > 
> > Could you test attached patch?
> > 
> Hm, with the attached patch qemu starts, but the guest will not come up
> (same commandline + kernel comes up fine on master). Let me dig around
> a bit.

And the winner is: event_idx

s390-virtio wants event_idx=false or we can't get I/O through. I guess
we can't use DEFINE_VIRTIO_COMMON_FEATURES for this transport.

(Probably nobody tested virtio-rng and vhost-scsi on this transport,
either :)