Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Kuniyuki Iwashima <kuniyu@amazon.com>
To:     <lmb@isovalent.com>
CC:     <andrii@kernel.org>, <ast@kernel.org>, <bpf@vger.kernel.org>,
        <daniel@iogearbox.net>, <davem@davemloft.net>,
        <dsahern@kernel.org>, <edumazet@google.com>, <haoluo@google.com>,
        <hemanthmalla@gmail.com>, <joe@wand.net.nz>,
        <john.fastabend@gmail.com>, <jolsa@kernel.org>,
        <kpsingh@kernel.org>, <kuba@kernel.org>, <kuniyu@amazon.com>,
        <linux-kernel@vger.kernel.org>, <linux-kselftest@vger.kernel.org>,
        <martin.lau@linux.dev>, <mykolal@fb.com>, <netdev@vger.kernel.org>,
        <pabeni@redhat.com>, <sdf@google.com>, <shuah@kernel.org>,
        <song@kernel.org>, <willemdebruijn.kernel@gmail.com>, <yhs@fb.com>
Subject: Re: [PATCH bpf-next v2 3/6] net: remove duplicate reuseport_lookup functions
Date:   Wed, 14 Jun 2023 09:52:44 -0700
Message-ID: <20230614165244.59782-1-kuniyu@amazon.com>
In-Reply-To: <CAN+4W8ijtoew8ouaN3i1NXtg0_G_HHmZyAtf5LsCBb6shCAx2Q@mail.gmail.com>
References: <CAN+4W8ijtoew8ouaN3i1NXtg0_G_HHmZyAtf5LsCBb6shCAx2Q@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8bit
Precedence: bulk

From: Lorenz Bauer <lmb@isovalent.com>
Date: Wed, 14 Jun 2023 16:25:05 +0100
> On Tue, Jun 13, 2023 at 7:57 PM Kuniyuki Iwashima <kuniyu@amazon.com> wrote:
> >
> >                         else {
> >                                 result = sk;
> >                         }
> >
> > The assignment to result below is buggy.  Let's say SO_REUSEPROT group
> > have TCP_CLOSE and TCP_ESTABLISHED sockets.
> 
> I'm not very familiar with SO_REUSEPORT, I assumed (incorrectly
> probably) that such a group would only ever have TCP_CLOSE in UDP case
> and TCP_LISTENING in TCP case. Can you explain how I could end up in
> this situation?

When we call conenct() for UDP socket in SO_REUSEPORT group, the state
is changed from TCP_CLOSE to TCP_ESTABLISHED in __ip4_datagram_connect(),
and the socket remains in the group.

That's why we check TCP_ESTABLISHED in reuseport_select_sock_by_hash()
that is always false for TCP but true for UDP in the case above.