[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <aa6e590e59fb3e360676a5ada5f80d01e0011d14.camel@inf.elte.hu>
Date: Thu, 16 Feb 2023 15:41:23 +0100
From: Ferenc Fejes <fejes@....elte.hu>
To: Vladimir Oltean <vladimir.oltean@....com>
Cc: peti.antal99@...il.com, netdev@...r.kernel.org,
"David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>,
Claudiu Manoil <claudiu.manoil@....com>,
Vinicius Costa Gomes <vinicius.gomes@...el.com>,
Kurt Kanzenbach <kurt@...utronix.de>,
Jacob Keller <jacob.e.keller@...el.com>,
Jamal Hadi Salim <jhs@...atatu.com>,
Cong Wang <xiyou.wangcong@...il.com>,
Jiri Pirko <jiri@...nulli.us>,
Simon Horman <simon.horman@...igine.com>
Subject: Re: [PATCH v6 net-next 02/13] net/sched: mqprio: refactor
offloading and unoffloading to dedicated functions
Hi Vladimir!
On Thu, 2023-02-16 at 16:28 +0200, Vladimir Oltean wrote:
> Hi Ferenc,
>
> On Thu, Feb 16, 2023 at 02:05:22PM +0100, Ferenc Fejes wrote:
> > This patch just code refactoring or it modifies the default
> > behavior of
> > the offloading too? I'm asking it in regards of the veth interface.
> > When you configure mqprio, the "hw" parameter is mandatory. By
> > default,
> > it tries to configure it with "hw 1". However as a result, veth
> > spit
> > back "Invalid argument" error (before your patches). Same happens
> > after
> > this patch too, right?
>
> Yup. iproute2 has a default queue configuration built in, if nothing
> else is specified, and this has "hw 1":
> https://git.kernel.org/pub/scm/network/iproute2/iproute2.git/tree/tc/q_mqprio.c#n36
>
> > For veth hardware offloading makes no sense, but giving the "hw 0"
> > argument explicitly as mqprio parameter might counterintuitive.
>
> Agree that giving the right nlattrs to mqprio and trying to slalom
> through their validation is a frustrating minesweeper game. I have
> some
> patches which add some netlink EXT_ACK messages to make this a bit
> less
> sour. I'm regression-testing those, together with some other mqprio
> changes and I hope to send them soon.
Nice, good to hear!
>
> OTOH, "hw 1" is mandatory with the "mode", "shaper", "min_rate" and
> "max_rate" options. This is logical when you think about it (driver
> has
> to act upon them), but indeed it makes mqprio difficult to configure.
>
> With veth, you need to use multi-queue to make use of mqprio/taprio,
> have you done that?
>
> ip link add veth0 numtxqueues 8 numrxqueues 8 type veth peer name
> veth1
Yes, usually I done it with ethtool --set-channels veth0 tx 8 but I
guess that both resulting the same.
Best,
Ferenc
Powered by blists - more mailing lists