[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CA+NMeC-AWyQnrQsR+i+mgzdb5KyZVe1_994_DhZ4Ne7QaOZPgA@mail.gmail.com>
Date: Wed, 26 Nov 2025 17:31:41 -0300
From: Victor Nogueira <victor@...atatu.com>
To: Jamal Hadi Salim <jhs@...atatu.com>
Cc: Stephen Hemminger <stephen@...workplumber.org>, davem@...emloft.net, kuba@...nel.org,
edumazet@...gle.com, pabeni@...hat.com, jiri@...nulli.us,
xiyou.wangcong@...il.com, horms@...nel.org, netdev@...r.kernel.org
Subject: Re: [RFC PATCH net-next v2] net/sched: Introduce qdisc quirk_chk op
On Wed, Nov 26, 2025 at 1:29 PM Jamal Hadi Salim <jhs@...atatu.com> wrote:
>
> On Tue, Nov 25, 2025 at 10:11 PM Stephen Hemminger
> <stephen@...workplumber.org> wrote:
> >
> > > [...]
> > > - root = qdisc_root_sleeping(sch);
> > > + qopt = nla_data(opt);
> > > + duplicates = qopt->duplicate;
> > > +
> > > + dev = sch->dev_queue->dev;
> > > + root = rtnl_dereference(dev->qdisc);
> > >
> > > if (sch != root && root->ops->cl_ops == &netem_class_ops) {
> > > if (duplicates ||
> > > @@ -992,19 +1006,25 @@ static int check_netem_in_tree(struct Qdisc *sch, bool duplicates,
> > > if (!qdisc_dev(root))
> > > return 0;
> > >
> > > + root_is_mq = root->flags & TCQ_F_MQROOT;
> > > +
> >
> > What about HTB or other inherently multi-q qdisc?
> > Using netem on HTB on some branches is common practice.
>
> Agreed - this should cover all classful qdiscs. I think the check
> comes from the earlier discussion which involves offloadable qdiscs
> that have multiple children (where HTB fits as well).
> @Victor Nogueira how about check for nested netems with duplicates?
Ok, will look at that approach for the next update
cheers,
Victor
Powered by blists - more mailing lists