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;
MIME-Version: 1.0
References: <20201023111352.GA289522@rdias-suse-pc.lan> <CANn89iJDt=XpUZA_uYK98cK8tctW6M=f4RFtGQpTxRaqwnnqSQ@mail.gmail.com>
 <20201023155145.GA316015@rdias-suse-pc.lan> <CANn89iL2VOH+Mg9-U7pkpMkKykDfhoX-GMRnF-oBmZmCGohDtA@mail.gmail.com>
 <20201023160628.GA316690@rdias-suse-pc.lan>
In-Reply-To: <20201023160628.GA316690@rdias-suse-pc.lan>
From:   Eric Dumazet <edumazet@google.com>
Date:   Fri, 23 Oct 2020 18:36:29 +0200
Message-ID: <CANn89i+OZF2HJQYT0FGtzyFeZMdof9RAfGXQRKUVY6Hg9ZPpcg@mail.gmail.com>
Subject: Re: [PATCH] tcp: fix race condition when creating child sockets from syncookies
To:     Ricardo Dias <rdias@memsql.com>
Cc:     David Miller <davem@davemloft.net>,
        Jakub Kicinski <kuba@kernel.org>,
        Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>,
        Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>,
        netdev <netdev@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Fri, Oct 23, 2020 at 6:06 PM Ricardo Dias <rdias@memsql.com> wrote:

> And what about the loopback interface? Why couldn't the loopback
> interface also use a single RX queue?
>

Loopback is using a per-cpu queue, with no crossing, for efficiency.

That means : whenever a packet is sent on lo interface from CPU X, it
is put on CPU X backlog queue.

If the connect() and sendmsg() are run from different cpus, then the
ACK (from last packet of 3WH) and the data packet might land on
different queues.