Re: [RFC PATCH net-next v5 3/4] virtio_net: Extend virtio to use VF datapath when available

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

On Wed, Apr 11, 2018 at 10:03:32AM +0200, Jiri Pirko wrote:
> Wed, Apr 11, 2018 at 08:24:43AM CEST, sridhar.samudrala@intel.com wrote:
> >On 4/10/2018 11:03 PM, Jiri Pirko wrote:
> >> Tue, Apr 10, 2018 at 05:59:02PM CEST, sridhar.samudrala@intel.com wrote:
> >> > On 4/10/2018 8:43 AM, Jiri Pirko wrote:
> >> > > Tue, Apr 10, 2018 at 05:27:48PM CEST, sridhar.samudrala@intel.com wrote:
> >> > > > On 4/10/2018 8:22 AM, Jiri Pirko wrote:
> >> > > > > Tue, Apr 10, 2018 at 05:13:40PM CEST, sridhar.samudrala@intel.com wrote:
> >> > > > > > On 4/10/2018 3:55 AM, Jiri Pirko wrote:
> >> > > > > > > Mon, Apr 09, 2018 at 08:47:06PM CEST, sridhar.samudrala@intel.com wrote:
> >> > > > > > > > On 4/9/2018 1:07 AM, Jiri Pirko wrote:
> >> > > > > > > > > Sat, Apr 07, 2018 at 12:59:14AM CEST, sridhar.samudrala@intel.com wrote:
> >> > > > > > > > > > On 4/6/2018 5:48 AM, Jiri Pirko wrote:
> >> > > > > > > > > > > Thu, Apr 05, 2018 at 11:08:22PM CEST, sridhar.samudrala@intel.com wrote:
> >> > > > > > > > > [...]
> >> > > > > > > > > 
> >> > > > > > > > > > > > +static int virtnet_bypass_join_child(struct net_device *bypass_netdev,
> >> > > > > > > > > > > > +				     struct net_device *child_netdev)
> >> > > > > > > > > > > > +{
> >> > > > > > > > > > > > +	struct virtnet_bypass_info *vbi;
> >> > > > > > > > > > > > +	bool backup;
> >> > > > > > > > > > > > +
> >> > > > > > > > > > > > +	vbi = netdev_priv(bypass_netdev);
> >> > > > > > > > > > > > +	backup = (child_netdev->dev.parent == bypass_netdev->dev.parent);
> >> > > > > > > > > > > > +	if (backup ? rtnl_dereference(vbi->backup_netdev) :
> >> > > > > > > > > > > > +			rtnl_dereference(vbi->active_netdev)) {
> >> > > > > > > > > > > > +		netdev_info(bypass_netdev,
> >> > > > > > > > > > > > +			    "%s attempting to join bypass dev when %s already present\n",
> >> > > > > > > > > > > > +			    child_netdev->name, backup ? "backup" : "active");
> >> > > > > > > > > > > Bypass module should check if there is already some other netdev
> >> > > > > > > > > > > enslaved and refuse right there.
> >> > > > > > > > > > This will work for virtio-net with 3 netdev model, but this check has to be done by netvsc
> >> > > > > > > > > > as its bypass_netdev is same as the backup_netdev.
> >> > > > > > > > > > Will add a flag while registering with the bypass module to indicate if the driver is doing
> >> > > > > > > > > > a 2 netdev or 3 netdev model and based on that flag this check can be done in bypass module
> >> > > > > > > > > > for 3 netdev scenario.
> >> > > > > > > > > Just let me undestand it clearly. What I expect the difference would be
> >> > > > > > > > > between 2netdev and3 netdev model is this:
> >> > > > > > > > > 2netdev:
> >> > > > > > > > >        bypass_master
> >> > > > > > > > >           /
> >> > > > > > > > >          /
> >> > > > > > > > > VF_slave
> >> > > > > > > > > 
> >> > > > > > > > > 3netdev:
> >> > > > > > > > >        bypass_master
> >> > > > > > > > >           /     \
> >> > > > > > > > >          /       \
> >> > > > > > > > > VF_slave   backup_slave
> >> > > > > > > > > 
> >> > > > > > > > > Is that correct? If not, how does it look like?
> >> > > > > > > > > 
> >> > > > > > > > > 
> >> > > > > > > > Looks correct.
> >> > > > > > > > VF_slave and backup_slave are the original netdevs and are present in both the models.
> >> > > > > > > > In the 3 netdev model, bypass_master netdev is created and VF_slave and backup_slave are
> >> > > > > > > > marked as the 2 slaves of this new netdev.
> >> > > > > > > You say it looks correct and in another sentence you provide completely
> >> > > > > > > different description. Could you please look again?
> >> > > > > > > 
> >> > > > > > To be exact, 2 netdev model with netvsc looks like this.
> >> > > > > > 
> >> > > > > >       netvsc_netdev
> >> > > > > >         /
> >> > > > > >        /
> >> > > > > > VF_slave
> >> > > > > > 
> >> > > > > > With virtio_net, 3 netdev model
> >> > > > > > 
> >> > > > > >     bypass_netdev
> >> > > > > >         /     \
> >> > > > > >        /       \
> >> > > > > > VF_slave   virtio_net netdev
> >> > > > > Could you also mark the original netdev which is there now? is it
> >> > > > > bypass_netdev or virtio_net_netdev ?
> >> > > > bypass_netdev
> >> > > >       /     \
> >> > > >      /       \
> >> > > > VF_slave   virtio_net netdev (original)
> >> > > That does not make sense.
> >> > > 1) You diverge from the behaviour of the netvsc, where the original
> >> > >      netdev is a master of the VF
> >> > > 2) If the original netdev is a slave, you cannot have any IP address
> >> > >      configured on it (well you could, but the rx_handler would eat every
> >> > >      incoming packet). So you will break the user bacause he would have to
> >> > >      move the configuration to the new master device.
> >> > > This only makes sense that the original netdev becomes the master for both
> >> > > netvsc and virtio_net.
> >> > Forgot to mention that bypass_netdev takes over the name of the original
> >> > netdev and
> >> > virtio_net netdev will get the backup name.
> >> What do you mean by "name"?
> >
> >bypass_netdev also is associated with the same pci device as the original virtio_net
> >netdev via SET_NETDEV_DEV().  Also, we added ndo_get_phys_port_name() to virtio_net
> >that will return _bkup when BACKUP feature is enabled.
> 
> Okay.
> 
> >
> >So for ex: if virtio_net inteface was getting 'ens12' as the name assigned by udev
> >without BACKUP feature,  when BACKUP feature is enabled,  the  bypass_netdev will be
> >named 'ens12' and the original virtio_net will get named as ens12n_bkup.
> 
> Got it.
> 
> I don't like the bypass_master to look differently in netvsc and
> virtio_net :/ The best would be to convert netvsc to 3 netdev model and
> treat them the same. The more I think about it, the more the 2 netdev
> model feels wrong.

If you believe that, then this patchset is a step in the right
direction.

With something like this patchset applied, converting netvsc to a 3
device model will presumably be just a flag flip away. Afterwards
we'll be able to drop dead code handling the bypass_master flag.

> 
> >
> >
> >> 
> >> > So the userspace network configuration doesn't need to change.
> >> > 
> >> > 
> >