Re: [PATCH net] failover: eliminate callback hell

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

On Wed, Jun 06, 2018 at 03:24:08PM -0700, Stephen Hemminger wrote:
> On Thu, 7 Jun 2018 00:47:52 +0300
> "Michael S. Tsirkin" <mst@redhat.com> wrote:
> 
> > On Wed, Jun 06, 2018 at 02:24:47PM -0700, Stephen Hemminger wrote:
> > > On Wed, 6 Jun 2018 15:30:27 +0300
> > > "Michael S. Tsirkin" <mst@redhat.com> wrote:
> > >   
> > > > On Wed, Jun 06, 2018 at 09:25:12AM +0200, Jiri Pirko wrote:  
> > > > > Tue, Jun 05, 2018 at 05:42:31AM CEST, stephen@networkplumber.org wrote:    
> > > > > >The net failover should be a simple library, not a virtual
> > > > > >object with function callbacks (see callback hell).    
> > > > > 
> > > > > Why just a library? It should do a common things. I think it should be a
> > > > > virtual object. Looks like your patch again splits the common
> > > > > functionality into multiple drivers. That is kind of backwards attitude.
> > > > > I don't get it. We should rather focus on fixing the mess the
> > > > > introduction of netvsc-bonding caused and switch netvsc to 3-netdev
> > > > > model.    
> > > > 
> > > > So it seems that at least one benefit for netvsc would be better
> > > > handling of renames.
> > > > 
> > > > Question is how can this change to 3-netdev happen?  Stephen is
> > > > concerned about risk of breaking some userspace.
> > > > 
> > > > Stephen, this seems to be the usecase that IFF_HIDDEN was trying to
> > > > address, and you said then "why not use existing network namespaces
> > > > rather than inventing a new abstraction". So how about it then? Do you
> > > > want to find a way to use namespaces to hide the PV device for netvsc
> > > > compatibility?
> > > >   
> > > 
> > > Netvsc can't work with 3 dev model. MS has worked with enough distro's and
> > > startups that all demand eth0 always be present. And VF may come and go.  
> > 
> > Well failover seems to maintain this invariant with the 3 dev model.
> > 
> > > After this history, there is a strong motivation not to change how kernel
> > > behaves. Switching to 3 device model would be perceived as breaking
> > > existing userspace.  
> > 
> > I feel I'm misunderstood. I was asking whether a 3-rd device can be
> > hidden so that userspace does not know that you switched to a 3 device
> > model. It will think there are 2 devices and will keep working.
> > 
> > If you do that, then there won't be anything that
> > would be perceived as breaking existing userspace, will there?
> 
> DPDK now knows about the netvsc 2 device model and drivers in userspace
> depend on it.

Interesting but I'm not sure how this answers the question. How would
DPDK care that there's a hidden device? If you can point out the
code in question, maybe a way can be found to make changes while
keeping it working.

> > 
> > 
> > > With virtio you can  work it out with the distro's yourself.
> > > There is no pre-existing semantics to deal with.
> > > 
> > > For the virtio, I don't see the need for IFF_HIDDEN.
> > > With 3-dev model as long as you mark the PV and VF devices
> > > as slaves, then userspace knows to leave them alone. Assuming userspace
> > > is already able to deal with team and bond devices.  
> > 
> > That's clear enough.
> > 
> > > Any time you introduce new UAPI behavior something breaks.  
> > 
> > Not if we do it right.
> > 
> > > On the rename front, I really don't care if VF can be renamed.  
> > 
> > OK that's nice.
> > 
> > > And for
> > > netvsc want to allow the PV device to be renamed.  
> > 
> > That's because of the 2 device model, right?  So that explains why even
> > if the delayed hack is good for the goose it might not be good for the
> > gander :)
> 
> You are bringing up the VF right away. How does the 3-device initialization
> state machine work?  Do you give a window for udev to possibly rename the
> VF? Do you rely on that?
> 
> > 
> > > Udev developers want that
> > > but have not found a stable/persistent value to expose to userspace
> > > to allow it.