[dpdk-dev] [PATCH v3 2/8] ring: dynamic rss configuration

[Kulasek, TomaszX]

> -----Original Message-----
> From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> Sent: Monday, July 13, 2015 17:11
> To: Kulasek, TomaszX
> Cc: dev at dpdk.org
> Subject: Re: [dpdk-dev] [PATCH v3 2/8] ring: dynamic rss configuration
> 
> 2015-07-13 14:43, Kulasek, TomaszX:
> > From: Thomas Monjalon [mailto:thomas.monjalon at 6wind.com]
> > > 2015-06-29 16:50, Tomasz Kulasek:
> > > > This implementation allows to set and read RSS configuration for
> > > > ring device, and is used to validate right values propagation over
> > > > the slaves, in test units for dynamic RSS configuration for bonding.
> > > >
> > > > It have no impact on packet processing by ring device.
> > >
> > > Adding some fake RSS to the ring PMD (in order to test bonding) is
> weird.
> > > The ring PMD is not a driver for testing. Maybe that the null PMD
> > > would be more appropriate.
> > > By the way the current RSS implementation is really bound to Intel
> > > devices.
> > > Before applying it to more drivers, we have to make sure it is
> > > generic enough. Maybe the RETA needs more abstraction.
> >
> > This is not RSS implementation, but implementation of Dynamic RSS
> > Configuration, already existing, and well defined in official
> > documentation for single port.
> > I don't know where you see bounding to Intel device.
> > It's an implementation of official API.
> 
> What do you mean by "official"?
> My concern is that the RSS API in ethdev comes from a time where DPDK was
> Intel DPDK. I may be wrong but I think that other devices could need
> something more generic and better defined.
> 
> > Anyway I will appreciate for any comment and opinion on that and
> > clarification what means more RETA abstraction in context of official
> API.
> >
> > I will check the possibility of moving it to the null pmd driver and
> > will prepare new version for easier reviewing. Meantime I'm waiting
> > for more opinions.
> 
> Exact, we need more opinions on this topic.

Hi Thomas,
I have fixed the error with wrong copy/paste. Also I changed usage of ring pmd with null pmd like you suggested. You are right, it is better place. Thanks.

I sent a v4 patch-set.

This implementation is using an existing ethdev API for RSS and introduces generic enhancements for configuration of bonding slaves. 
If other devices need a more generic ethdev API then this should be treated separately and they can submit separate patches for this in a future release.