[dpdk-dev] [PATCH v3 1/2] ethdev: add tunnel and port RSS offload types

Thomas Monjalon thomas.monjalon at 6wind.com
Wed Jun 22 09:52:10 CEST 2016


2016-06-22 12:45, Jerin Jacob:
> On Wed, Jun 22, 2016 at 08:43:52AM +0200, Thomas Monjalon wrote:
> > 2016-06-22 09:00, Jerin Jacob:
> > > On Tue, Jun 21, 2016 at 11:02:59PM +0200, Thomas Monjalon wrote:
> > > > 2016-03-31 02:21, Jerin Jacob:
> > > > > +#define RTE_ETH_FLOW_PORT               18
> > > > > +#define RTE_ETH_FLOW_VXLAN              19
> > > > > +#define RTE_ETH_FLOW_GENEVE             20
> > > > > +#define RTE_ETH_FLOW_NVGRE              21
> > > > > +#define RTE_ETH_FLOW_MAX                22
> > > > 
> > > > Please could you explain more what is PORT flow?
> > > 
> > > For example, a NIC card with two physical port where application
> > > configures RTE_ETH_FLOW_IPV4 for both, In that case
> > > HW generate same RSS value for a similar IPV4 packet,  However, in-case if
> > > application want to generate a flow that account physical port also then
> > > it can configure with RTE_ETH_FLOW_IPV4 | RTE_ETH_FLOW_PORT.
> > > 
> > > RTE_ETH_FLOW_PORT useful for the case where one physical port assigned for
> > > INBOUND traffic and other-one for OUTBOUND traffic etc
> > 
> > OK
> > 
> > > > Does it need a comment in the code?
> > > Not sure, commit log has description.
> > 
> > How do you expect the user to understand this new value in the API?
> > Users do not check in the git history.
> > They use doxygen, headers comments and/or examples.
> 
> The reason why I said because none of flow type has comments in the
> list. If you think RTE_ETH_FLOW_PORT needs a doxygen comment then I can
> add it.
> 
> It would be nice some else could add the comments for following,
> RTE_ETH_FLOW_RAW,
> RTE_ETH_FLOW_L2_PAYLOAD

These values passed without a proper check.
That's why we must not accept any line in API without good comment.

Please go ahead with what you can do and we'll fix or remove
the remaining later.


More information about the dev mailing list