Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Tue, 23 Mar 2021 13:39:00 +0200
From:   Vladimir Oltean <olteanv@gmail.com>
To:     Yunsheng Lin <linyunsheng@huawei.com>
Cc:     davem@davemloft.net, kuba@kernel.org, ast@kernel.org,
        daniel@iogearbox.net, andriin@fb.com, edumazet@google.com,
        weiwan@google.com, cong.wang@bytedance.com, ap420073@gmail.com,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        linuxarm@openeuler.org, mkl@pengutronix.de,
        linux-can@vger.kernel.org, jhs@mojatatu.com,
        xiyou.wangcong@gmail.com, jiri@resnulli.us, andrii@kernel.org,
        kafai@fb.com, songliubraving@fb.com, yhs@fb.com,
        john.fastabend@gmail.com, kpsingh@kernel.org, bpf@vger.kernel.org,
        jonas.bonn@netrounds.com, pabeni@redhat.com, mzhivich@akamai.com,
        johunt@akamai.com, albcamus@gmail.com, kehuan.feng@gmail.com,
        a.fatoum@pengutronix.de
Subject: Re: [RFC v3] net: sched: implement TCQ_F_CAN_BYPASS for lockless
 qdisc
Message-ID: <20210323113900.g3c2mg5z4elamyz3@skbuf>
References: <1616050402-37023-1-git-send-email-linyunsheng@huawei.com>
 <1616404156-11772-1-git-send-email-linyunsheng@huawei.com>
 <20210322200033.uphemtsunfqsvjej@skbuf>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210322200033.uphemtsunfqsvjej@skbuf>
Precedence: bulk

On Mon, Mar 22, 2021 at 10:00:33PM +0200, Vladimir Oltean wrote:
> Hi Yunsheng,
> 
> On Mon, Mar 22, 2021 at 05:09:16PM +0800, Yunsheng Lin wrote:
> > Currently pfifo_fast has both TCQ_F_CAN_BYPASS and TCQ_F_NOLOCK
> > flag set, but queue discipline by-pass does not work for lockless
> > qdisc because skb is always enqueued to qdisc even when the qdisc
> > is empty, see __dev_xmit_skb().
> > 
> > This patch calls sch_direct_xmit() to transmit the skb directly
> > to the driver for empty lockless qdisc too, which aviod enqueuing
> > and dequeuing operation. qdisc->empty is set to false whenever a
> > skb is enqueued, see pfifo_fast_enqueue(), and is set to true when
> > skb dequeuing return NULL, see pfifo_fast_dequeue().
> > 
> > There is a data race between enqueue/dequeue and qdisc->empty
> > setting, qdisc->empty is only used as a hint, so we need to call
> > sch_may_need_requeuing() to see if the queue is really empty and if
> > there is requeued skb, which has higher priority than the current
> > skb.
> > 
> > The performance for ip_forward test increases about 10% with this
> > patch.
> > 
> > Signed-off-by: Yunsheng Lin <linyunsheng@huawei.com>
> > ---
> > Hi, Vladimir and Ahmad
> > 	Please give it a test to see if there is any out of order
> > packet for this patch, which has removed the priv->lock added in
> > RFC v2.
> > 
> > There is a data race as below:
> > 
> >       CPU1                                   CPU2
> > qdisc_run_begin(q)                            .
> >         .                                q->enqueue()
> > sch_may_need_requeuing()                      .
> >     return true                               .
> >         .                                     .
> >         .                                     .
> >     q->enqueue()                              .
> > 
> > When above happen, the skb enqueued by CPU1 is dequeued after the
> > skb enqueued by CPU2 because sch_may_need_requeuing() return true.
> > If there is not qdisc bypass, the CPU1 has better chance to queue
> > the skb quicker than CPU2.
> > 
> > This patch does not take care of the above data race, because I
> > view this as similar as below:
> > 
> > Even at the same time CPU1 and CPU2 write the skb to two socket
> > which both heading to the same qdisc, there is no guarantee that
> > which skb will hit the qdisc first, becuase there is a lot of
> > factor like interrupt/softirq/cache miss/scheduling afffecting
> > that.
> > 
> > So I hope the above data race will not cause problem for Vladimir
> > and Ahmad.
> > ---
> 
> Preliminary results on my test setup look fine, but please allow me to
> run the canfdtest overnight, since as you say, races are still
> theoretically possible.

I haven't found any issues during the overnight test and until now.

Tested-by: Vladimir Oltean <vladimir.oltean@nxp.com> # flexcan