Message-ID: <20140825221148.GC30140@casper.infradead.org>
Date:	Mon, 25 Aug 2014 23:11:48 +0100
From:	Thomas Graf <tgraf@...g.ch>
To:	Jamal Hadi Salim <jhs@...atatu.com>
Cc:	Scott Feldman <sfeldma@...ulusnetworks.com>,
	John Fastabend <john.fastabend@...il.com>,
	Jiri Pirko <jiri@...nulli.us>, netdev@...r.kernel.org,
	davem@...emloft.net, nhorman@...driver.com, andy@...yhouse.net,
	dborkman@...hat.com, ogerlitz@...lanox.com, jesse@...ira.com,
	pshelar@...ira.com, azhou@...ira.com, ben@...adent.org.uk,
	stephen@...workplumber.org, jeffrey.t.kirsher@...el.com,
	vyasevic@...hat.com, xiyou.wangcong@...il.com,
	john.r.fastabend@...el.com, edumazet@...gle.com,
	f.fainelli@...il.com, roopa@...ulusnetworks.com,
	linville@...driver.com, dev@...nvswitch.org, jasowang@...hat.com,
	ebiederm@...ssion.com, nicolas.dichtel@...nd.com,
	ryazanov.s.a@...il.com, buytenh@...tstofly.org,
	aviadr@...lanox.com, nbd@...nwrt.org, alexei.starovoitov@...il.com,
	Neil.Jerram@...aswitch.com, ronye@...lanox.com
Subject: Re: [patch net-next RFC 10/12] openvswitch: add support for datapath
 hardware offload

First of all, thanks for the animated discussion, wouldn't
want to miss our arguments ;-)

On 08/25/14 at 12:48pm, Jamal Hadi Salim wrote:
> On 08/25/14 10:54, Thomas Graf wrote:
> >On 08/24/14 at 11:15am, Jamal Hadi Salim wrote:
> 
> >Let's keep vendors out of this discussion.
> 
> The API is from a vendor. It is clearly labelled as an OF API.
> It covers well abstracting that vendors SDK to enable OF. That
> is relevant info.
> If it covers all other vendors (which is where
> the quark handling comes in), I will be fine with it.
> I dont believe it does.

If I understand you correctly you are referring to the rocker
patch here. That is not part of the API.
 
> >That is simply not the case. The fact that John is using this model
> >to replace the flow director ioctl API should prove this.
> 
> depends what NIC classifier John is mapping to. The Intels have
> about 4-5 different types of classifier on different hardware

Sorry for not addressing this but I think John should speak for
himself here, I don't want to misrepresent his plans.

> I gave a simple example.
> There are a hell more quarks than that.
> There are cases where there are multiple tables in terms of net masks
> etc.
> Yes, this should be handled in the driver. The input is the route
> message we already specify and not some XXX_Flow_XXx struct.

I would argue that swflow is a superset of a Netlink route. It
may infact be very useful to extend the API with something that
understands the Netlink representation of a route and have the
API translate that to a classifier that can be offloaded.

> I would be tagging along with you guys for flows if you:
> a) allow for different classifiers. This allows me to implement
> u32 and offload it.

Agreed. What you seem to disagree on is:

 - ndo_add_type1([...])
 - ndo_add_type2([...])
 - ndo_add_type3([...])

vs.

 - ndo_add_classifier(type, [...])

I honestly have little against the 2nd. It sounds a bit like an
ioctl interface though where a giant switch statement will cast
the data to a classifier specific struct which is why I slightly
dislike it.

It looks to me that a specific chip may either work in a flow/filter
mode, in a generic programmable mode or by providing a list of very
specific filters without a generic flow -> action relation. Having
multiple classifier types for all of them gives the impression that
an API user could use them in any combination which I would say will
typically not be the case.

> b) different actions (I think this part is not controversial, you
> seem to be having it already).

Agreed
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists