Re: [patch net-next RFC 0/4] introduce infrastructure for support of switch chip datapath

[Thomas Graf <tgraf@suug.ch>]

On 03/25/14 at 02:42pm, Florian Fainelli wrote:
> 2014-03-25 14:26 GMT-07:00 Thomas Graf <tgraf@suug.ch>:
> > On 03/25/14 at 01:11pm, Florian Fainelli wrote:
> >> 2014-03-25 12:35 GMT-07:00 Neil Horman <nhorman@tuxdriver.com>:
> >> > On Tue, Mar 25, 2014 at 06:00:09PM +0000, Thomas Graf wrote:
> >> >> On top of that, like for VFs, we provide extended nested attributes or
> >> >> alternate control paths such as via OVS that provide the additional
> >> >> flexibility and control required by the more advanced devices.
> >> > I'm sorry, I don't understand the relevance here.  Are you suggesting that to
> >> > make this modification, we would need to augment more than a single set of
> >> > netlink control paths?
> >>
> >> Not sure if I got this right, but there might be additional control
> >> knobs required for specific Ethernet switch features that do not map
> >> nicely, if at all with existing interfaces provided by ip/tc,
> >> ethtool... although I guess one would say, well, then go add these
> >> APIs instead of creating "extended" ones?
> >
> > Exactly. Some of the logic and configuration structure will not
> > fit the existing model and is too switch specific to justify
> > extending the generic link model. It also seems likely that some
> > knobs will be switch specific. Not an issue as long as they are
> > tunneled through the standard API and any effort is undertaken
> > to generalize where it makes sense.
> 
> The question is how you would imagine conveying these switch-specific
> features that do not (yet) map into a general feature, shall we go for
> a separate netlink family, just like what Felix did in OpenWrt with
> swconfig, without much stability from one kernel release to another,
> as we migrate what was once a switch specific feature into a general
> Ethernet switch feature?

I believe it is essential to transport them as part of the standard
Netlink API and have a single channel for all configuration. It also
eases message synchronization.

We also want to enforce strict ABI compatibility rules just like
for all other Netlink users. As we know, it's not difficult to design
the message format in a way to allow for extendability and backwards
compatibility.