[dpdk-dev] buf->hash.rss always empty with i40e

Zhang, Helin helin.zhang at intel.com
Tue Feb 7 02:59:39 CET 2017


Hi Tom

Sorry, by default, i40e HW is a bit different from ixgbe HW. It will not treat UDP as an IP packet, when setting hash enable flags.
But, a feature of configuring 'input set' can help to change the default HW behavior. Please refer to testpmd input set part to understand how to use that. Thanks!

Regards,
Helin

-----Original Message-----
From: tom.barbette at ulg.ac.be [mailto:tom.barbette at ulg.ac.be] 
Sent: Monday, February 6, 2017 11:08 PM
To: Ananyev, Konstantin <konstantin.ananyev at intel.com>
Cc: Richardson, Bruce <bruce.richardson at intel.com>; dev at dpdk.org; Zhang, Helin <helin.zhang at intel.com>; Wu, Jingjing <jingjing.wu at intel.com>
Subject: Re: buf->hash.rss always empty with i40e

Hi Konstantin,

It seems a little overkill to play with the key... The XL710 seems to be able to hash on IP fields only. It seems only a configuration issue, I'm adding i40e maintainers in CC so they can confirm? I think the i40e driver should configure XL710 to hash on IP fields for TCP and UDP when only ETH_RSS_IP is given instead of hashing to 0. That would mimic ixgbe behaviour btw.

Tom


----- Mail original -----
De: "Konstantin Ananyev" <konstantin.ananyev at intel.com>
À: "tom barbette" <tom.barbette at ulg.ac.be>, "Bruce Richardson" <bruce.richardson at intel.com>
Cc: dev at dpdk.org
Envoyé: Lundi 6 Février 2017 13:25:59
Objet: RE: buf->hash.rss always empty with i40e

Hi Tom,

> 
> That also leave the question of how to HASH only on the IP tuple for 
> TCP and UDP packets? The use case is that I want all packets from the same IP src/dst pair to be hashed to the same queue. This cannot be enforced with a complete hash on TCP/UDP fields.
> 

That's for IPv4 only?
It was a while ago, when I looked at it, but as I remember you can achieve that by filling first 64 bits of RSS hash with some meaningful values, and keeping remaining RSS bits as zeroes.
You probably can give it a quick try with testpmd.

There is a nice article about similar subject:
http://www.ndsl.kaist.edu/~kyoungsoo/papers/TR-symRSS.pdf

Konstantin



More information about the dev mailing list