linux-kernel - Re: [PATCH net-next 00/12] Add tc-mqprio and tc-taprio support for preemptible traffic classes

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <be80cf7b04b60e212dc4733f47e5d1f687a9f551.camel@inf.elte.hu>
Date:   Mon, 20 Feb 2023 09:11:10 +0100
From:   Ferenc Fejes <fejes@....elte.hu>
To:     Vladimir Oltean <vladimir.oltean@....com>
Cc:     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>,
        Jamal Hadi Salim <jhs@...atatu.com>,
        Cong Wang <xiyou.wangcong@...il.com>,
        Jiri Pirko <jiri@...nulli.us>,
        Vinicius Costa Gomes <vinicius.gomes@...el.com>,
        Kurt Kanzenbach <kurt@...utronix.de>,
        Gerhard Engleder <gerhard@...leder-embedded.com>,
        Amritha Nambiar <amritha.nambiar@...el.com>,
        Claudiu Manoil <claudiu.manoil@....com>,
        Alexandre Belloni <alexandre.belloni@...tlin.com>,
        UNGLinuxDriver@...rochip.com, Andrew Lunn <andrew@...n.ch>,
        Florian Fainelli <f.fainelli@...il.com>,
        Xiaoliang Yang <xiaoliang.yang_1@....com>,
        linux-kernel@...r.kernel.org,
        Pranavi Somisetty <pranavi.somisetty@....com>,
        Harini Katakam <harini.katakam@....com>
Subject: Re: [PATCH net-next 00/12] Add tc-mqprio and tc-taprio support for
 preemptible traffic classes

Hi Vladimir!

Thank you for the update!

On Sun, 2023-02-19 at 14:58 +0200, Vladimir Oltean wrote:
> Hi Ferenc,
> 
> On Sun, Feb 19, 2023 at 10:47:31AM +0100, Ferenc Fejes wrote:
> > Do you have the iproute2 part? Sorry if I missed it, but it would
> > be
> > nice to see how is that UAPI exposed for the config tools. Is there
> > any
> > new parameter for mqprio/taprio?
> 
> I haven't posted the iproute2 part (yet). For those familiar with my
> recent development, FP is a per-traffic-class netlink attribute just
> like queueMaxSDU from tc-taprio. That was exposed in iproute2 as an
> array of values, one per tc.
> 
> What I have in my tree would allow something like this:
> 
> tc qdisc replace dev $swp1 root stab overhead 20 taprio \
>         num_tc 8 \
>         map 0 1 2 3 4 5 6 7 \
>         queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
>         base-time 0 \
>         sched-entry S 0x7e 900000 \
>         sched-entry S 0x82 100000 \
>         max-sdu 0 0 0 0 0 0 0 200 \
>         fp P E E E E E E E \   # this is new (one entry per tc)
>         flags 0x2
> 
> tc qdisc replace dev $swp1 root mqprio \
>         num_tc 8 \
>         map 0 1 2 3 4 5 6 7 \
>         queues 1@0 1@1 1@2 1@3 1@4 1@5 1@6 1@7 \
>         fp P E E E E E E E \   # this is new (one entry per tc)
>         hw 1
> 
> of course the exact syntax is a potential matter of debate on its
> own,
> and does not really matter for the purpose of defining the kernel
> UAPI,
> which is why I wanted to keep discussions separate.

Fair enough. What you have right here is pretty straightforward IMO, I
would definitely support something like this.

> 
> For hardware which understands preemptible queues rather than traffic
> classes, how many queues are preemptible, and what are their offsets,
> will be deduced by translating the "queues" argument.
> 
> For hardware which understands preemptible priorities rather than
> traffic classes, which priorities are preemptible will be deduced by
> translating the "map" argument.

Great, that cover both cases with the same UAPI. I love the fact that
this even lets open the possibility to use prio-s (map) instead of
queues for FP.

> 
> The traffic class is the kernel entity which has the preemptible
> priority in my proposed UAPI because this is what my analysis of the
> standard has deduced that the preemptible quality is fundamentally
> attached to.
> 
> Considering that the UAPI for FP is a topic that has been discussed
> to
> death at least since August without any really new input since then,
> I'm
> going to submit v2 later today, and the iproute2 patch set afterwards
> (still need to write man page entries for that).

Best,
Ferenc