[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250423172416.4ee6378d@kernel.org>
Date: Wed, 23 Apr 2025 17:24:16 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Cong Wang <xiyou.wangcong@...il.com>
Cc: Paolo Abeni <pabeni@...hat.com>, Victor Nogueira <victor@...atatu.com>,
netdev@...r.kernel.org, jhs@...atatu.com, jiri@...nulli.us,
davem@...emloft.net, edumazet@...gle.com, toke@...hat.com,
gerrard.tai@...rlabs.sg, pctammela@...atatu.com
Subject: Re: [PATCH net v2 0/5] net_sched: Adapt qdiscs for reentrant
enqueue cases
On Wed, 23 Apr 2025 16:29:38 -0700 Cong Wang wrote:
> > + /*
> > + * If doing duplication then re-insert at top of the
> > + * qdisc tree, since parent queuer expects that only one
> > + * skb will be queued.
> > + */
> > + if (skb2) {
> > + struct Qdisc *rootq = qdisc_root_bh(sch);
> > + u32 dupsave = q->duplicate; /* prevent duplicating a dup... */
> > +
> > + q->duplicate = 0;
> > + rootq->enqueue(skb2, rootq, to_free);
> > + q->duplicate = dupsave;
> > + skb2 = NULL;
> > + }
> > +
> > finish_segs:
> > if (skb2)
> > __qdisc_drop(skb2, to_free);
> >
>
> Just FYI: I tested this patch, netem duplication still worked, I didn't
> see any issue.
Does it still work if you have another layer of qdiscs in the middle?
It works if say DRR is looking at the netem directly as its child when
it does:
first = cl->qdisc->q.qlen
but there may be another layer, the cl->qdisc may be something that
hasn't incremented its qlen, and something that has netem as its child.
Powered by blists - more mailing lists