Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
MIME-Version: 1.0
From:   "NeilBrown" <neilb@suse.de>
To:     "Trond Myklebust" <trond.myklebust@hammerspace.com>,
        "Anna Schumaker" <anna.schumaker@netapp.com>
Cc:     linux-nfs@vger.kernel.org
Subject: Re: [PATCH] SUNRPC: prevent port reuse on transports which don't request it.
In-reply-to: <162371991856.23575.16102887900102220450@noble.neil.brown.name>
References: <162371991856.23575.16102887900102220450@noble.neil.brown.name>
Date:   Thu, 17 Jun 2021 13:24:19 +1000
Message-id: <162390025921.29912.9620777152373832527@noble.neil.brown.name>
Precedence: bulk

On Tue, 15 Jun 2021, someone wrote:
> If an RPC client is created without RPC_CLNT_CREATE_REUSEPORT, it should
> not reuse the source port when a TCP connection is re-established.
> This is currently implemented by preventing the source port being
> recorded after a successful connection (the call to xs_set_srcport()).
>=20
> However the source port is also recorded after a successful bind in xs_bind=
().
> This may not be needed at all and certainly is not wanted when
> RPC_CLNT_CREATE_REUSEPORT wasn't requested.
>=20
> So avoid that assignment when xprt.reuseport is not set.
>=20
> With this change, NFSv4.1 and later mounts use a different port number on
> each connection.  This is helpful with some firewalls which don't cope
> well with port reuse.
>=20
> Signed-off-by: NeilBrown <neilb@suse.de>

I guess this should have

Fixes: e6237b6feb37 ("NFSv4.1: Don't rebind to the same source port when reco=
nnecting to the server")

NeilBrown

> ---
>  net/sunrpc/xprtsock.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>=20
> diff --git a/net/sunrpc/xprtsock.c b/net/sunrpc/xprtsock.c
> index 316d04945587..3228b7a1836a 100644
> --- a/net/sunrpc/xprtsock.c
> +++ b/net/sunrpc/xprtsock.c
> @@ -1689,7 +1689,8 @@ static int xs_bind(struct sock_xprt *transport, struc=
t socket *sock)
>  		err =3D kernel_bind(sock, (struct sockaddr *)&myaddr,
>  				transport->xprt.addrlen);
>  		if (err =3D=3D 0) {
> -			transport->srcport =3D port;
> +			if (transport->xprt.reuseport)
> +				transport->srcport =3D port;
>  			break;
>  		}
>  		last =3D port;
> --=20
> 2.31.1
>=20
>=20