Re: [dpdk-dev] [PATCH] ethdev: add DBDF action to RTE Flow

From: Thomas Monjalon <thomas@monjalon.net>
To: Kiran Kumar Kokkilagadda <kirankumark@marvell.com>
Cc: Ori Kam <orika@mellanox.com>, Wenzhuo Lu <wenzhuo.lu@intel.com>,
	Jingjing Wu <jingjing.wu@intel.com>,
	Bernard Iremonger <bernard.iremonger@intel.com>,
	John McNamara <john.mcnamara@intel.com>,
	Marko Kovacevic <marko.kovacevic@intel.com>,
	Ferruh Yigit <ferruh.yigit@intel.com>,
	Andrew Rybchenko <arybchenko@solarflare.com>,
	dev@dpdk.org
Subject: Re: [dpdk-dev] [PATCH] ethdev: add DBDF action to RTE Flow
Date: Thu, 19 Mar 2020 10:26:25 +0100	[thread overview]
Message-ID: <16170589.5WZRyvrzyv@xps> (raw)
In-Reply-To: <CH2PR18MB3272EA2A7835B0A9FEFBF038ACF40@CH2PR18MB3272.namprd18.prod.outlook.com>

19/03/2020 10:17, Kiran Kumar Kokkilagadda:
> From: Ori Kam
> > From: Kiran Kumar Kokkilagadda <kirankumark@marvell.com>
> > > From: Ori Kam <orika@mellanox.com>
> > > > From: kirankumark@marvell.com <kirankumark@marvell.com>
> > > > > From: Kiran Kumar K <kirankumark@marvell.com>
> > > > >
> > > > > Adding suuport to DBDF action in the RTE Flow.
> > > > > Application can specify the dbdf value using rte_flow_action_dbdf.
> > > > > Matched traffic will be sent to specified PCI DBDF device.
> > > > >
> > > > I would like to see more detail use case, for example to which
> > > > device / device type will the traffic be routed to?
> > > >
> > >
> > > We have the following use case.
> > > We have 2 PF's pf0, pf1 and corresponding VF's pf0_vf0 , pf1_vf0. And
> > > we have
> > > 3 applications running.
> > > 1st application on pf0 and pf1
> > > 2nd application on pf0_vf0
> > > 3rd application on pf1_vf0.
> > > We want to direct the traffic matching condition1 from application 1
> > > (traffic from both pf0 & pf1) needs to send to  application 2
> > > (pf0_vf0) And matching condition2 from application 1 (traffic from
> > > both pf0 & pf1) needs to send to application 3 (pf1_vf0).
> > > To summarize, we need to send traffic from pf0 to pf1_vf0 and traffic
> > > from pf1 to pf0_vf0. In this case This DBDF action will be useful.
> > >
> > 
> > It seems that what you are describing it the port action with representors, or any
> > other way you wish to implement it.
> 
> Let's say we have a VF with kernel and we want to send the traffic to that VF, then we can't
> Use port action. This will be useful in those scenarios.

Sorry I don't understand.
You mean the VF is managed by a kernel driver while the PF is managed by DPDK?
So what prevents having a VF representor?