Re: [Qemu-devel] [PATCH v4 0/2] virtio: Move host features to backends

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Apr 28, 2015 at 03:20:54PM +0200, Cornelia Huck wrote:
> On Tue, 28 Apr 2015 15:05:47 +0200
> "Michael S. Tsirkin" <mst@redhat.com> wrote:
> 
> > On Tue, Apr 28, 2015 at 01:52:40PM +0100, Peter Maydell wrote:
> > > On 28 April 2015 at 13:48, Michael S. Tsirkin <mst@redhat.com> wrote:
> > > > On Tue, Apr 28, 2015 at 07:51:11PM +0800, shannon.zhao@linaro.org wrote:
> > > >> From: Shannon Zhao <shannon.zhao@linaro.org>
> > > >> Here we move the host features to backends, involving
> > > >> DEFINE_VIRTIO_NET_FEATURES, DEFINE_VIRTIO_SCSI_FEATURES. So the
> > > >> virtio-mmio devices could have the host freatures, and this has a great
> > > >> performance improvement to virtio-mmio, especially to virtio-net-device.
> > > >
> > > > Can you move COMMON_FEATURES too please?
> > > 
> > > I think that would be wrong -- COMMON_FEATURES are transport features,
> > > not backend-specific features, so they belong on the transports
> > > (and on the convenience wrappers), not the backends.
> > > 
> > > thanks
> > > -- PMM
> > 
> > Hmm you are right. The problem is with s390 which
> > puts this DEFINE_VIRTIO_COMMON_FEATURES in all devices.
> 
> All save virtio-blk. There were some problems when switching on
> event_idx on virtio-blk, IIRC.

Specifically on s390?

> The other transport add DEFINE_VIRTIO_COMMON FEATURES in their base
> class. Because of the not-quite-understood problem with virtio-blk, we
> couldn't do that with s390-virtio.

-- 
MST