[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20241217101724.v6kbmfqpq3kgyrd4@skbuf>
Date: Tue, 17 Dec 2024 12:17:24 +0200
From: Vladimir Oltean <olteanv@...il.com>
To: Lorenzo Bianconi <lorenzo@...nel.org>
Cc: Lorenzo Bianconi <lorenzo.bianconi@...hat.com>,
Oleksij Rempel <linux@...pel-privat.de>, netdev@...r.kernel.org,
andrew@...n.ch, davem@...emloft.net, edumazet@...gle.com,
kuba@...nel.org, pabeni@...hat.com, horms@...nel.org, nbd@....name,
sean.wang@...iatek.com, Mark-MC.Lee@...iatek.com,
lorenzo.bianconi83@...il.com
Subject: Re: [RFC net-next 0/5] Add ETS and TBF Qdisc offload for Airoha
EN7581 SoC
On Tue, Dec 17, 2024 at 11:01:28AM +0100, Lorenzo Bianconi wrote:
> I agree it is unnecessary, but w/o it we must rely on limited QoS capabilities
> of the hw dsa switch. The goal of the series is just exploit enhanced QoS
> capabilities available on the EN7581 SoC.
(...)
> AFIR there is even the possibility to configure inter-channel QoS on the chip,
> like a Round Robin scheduler or Strict-Priority between channels.
So the conclusion of both these statements is that it would be good to understand
the QoS algorithm among channels of the MAC chip, see if there is any classful
root Qdisc which can describe that algorithm, then offload it with ndo_setup_tc(),
then assign traffic to user ports probably using a software classifier + flowid.
The ETS Qdiscs which you are trying to add here should not be attached to the root
egress Qdisc of DSA user ports, but to the classes of this unknown conduit root
Qdisc. Agree?
Powered by blists - more mailing lists