Re: [virtio-dev] [PATCH] content: Introduce VIRTIO_NET_F_STANDBY feature

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

On Mon, Jul 23, 2018 at 11:03:11AM +0200, Cornelia Huck wrote:
> On Sun, 22 Jul 2018 18:37:42 +0300
> "Michael S. Tsirkin" <mst@redhat.com> wrote:
> 
> > On Fri, Jul 20, 2018 at 09:44:50AM +0200, Cornelia Huck wrote:
> > > On Thu, 19 Jul 2018 14:29:40 -0700
> > > Sridhar Samudrala <sridhar.samudrala@intel.com> wrote:
> > >   
> > > > VIRTIO_NET_F_STANDBY feature enables hypervisor to indicate virtio_net
> > > > driver to act as a standby for another device with the same MAC address.
> > > > 
> > > > Signed-off-by: Sridhar Samudrala <sridhar.samudrala@intel.com
> > > > ---
> > > >  content.tex | 10 ++++++++++
> > > >  1 file changed, 10 insertions(+)
> 
> > > > @@ -2636,6 +2639,13 @@ If the driver negotiates VIRTIO_NET_F_MTU, it MUST NOT transmit packets of
> > > >  size exceeding the value of \field{mtu} (plus low level ethernet header length)
> > > >  with \field{gso_type} NONE or ECN.
> > > >  
> > > > +A driver SHOULD negotiate VIRTIO_NET_F_STANDBY feature if the device offers it.  
> > > 
> > > I'm not sure that this is the right section of the spec. Maybe we need
> > > a new normative driver section for "cross-device features" (better
> > > names wanted :), and the same for devices?  
> > 
> > You mean if we ever extend this to non-network devices?
> > This is a network specific feature bit so what makes
> > it a cross-device feature?
> 
> No, the network device location is alright. But it is in a section that
> deals with how the config space etc. is handled, and I'm not sure
> whether it fits well with the other statements in there. Not a major
> gripe, though.
> 
> > 
> > > > +
> > > > +If the driver negotiates VIRTIO_NET_F_STANDBY, it should act as a standby  
> > > for  
> > > > +another device with the same MAC address when available. The hypervisor can
> > > > +hot-plug a primary device with same MAC address if the feature is successfully
> > > > +negotiated with the driver.  
> > > 
> > > I don't think you should add implementation details like hotplugging
> > > into the spec.
> > > 
> > > What about:
> > > 
> > > "If the driver negotiates VIRTIO_NET_F_STANDBY, the device MAY act as a
> > > standby device for another device with the same MAC address, the
> > > 'failover device'.  
> > 
> > I find the name "failover device" confusing. Linux came up with
> > names primary and standby.
> 
> "...the device MAY act as a standby device for a primary device with
> the same MAC address."
> 
> ?

Right. But I agree we need a section that defines failover operation.

-- 
MST