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;
MIME-Version: 1.0
References: <20230727125125.1194376-1-imagedong@tencent.com> <20230727125125.1194376-2-imagedong@tencent.com>
In-Reply-To: <20230727125125.1194376-2-imagedong@tencent.com>
From:   Eric Dumazet <edumazet@google.com>
Date:   Thu, 27 Jul 2023 21:17:41 +0200
Message-ID: <CANn89iLwLxzSLKtobE9y+ZBU_eizfNo6FUfBa61KeUXsodA2FQ@mail.gmail.com>
Subject: Re: [PATCH net-next 1/3] net: tcp: send zero-window ACK when no memory
To:     menglong8.dong@gmail.com
Cc:     davem@davemloft.net, kuba@kernel.org, pabeni@redhat.com,
        dsahern@kernel.org, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org, Menglong Dong <imagedong@tencent.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Jul 27, 2023 at 2:51=E2=80=AFPM <menglong8.dong@gmail.com> wrote:
>
> From: Menglong Dong <imagedong@tencent.com>
>
> For now, skb will be dropped when no memory, which makes client keep
> retrans util timeout and it's not friendly to the users.
>
> In this patch, we reply an ACK with zero-window in this case to update
> the snd_wnd of the sender to 0. Therefore, the sender won't timeout the
> connection and will probe the zero-window with the retransmits.
>
> Signed-off-by: Menglong Dong <imagedong@tencent.com>
> ---
>  include/net/inet_connection_sock.h |  3 ++-
>  net/ipv4/tcp_input.c               |  4 ++--
>  net/ipv4/tcp_output.c              | 14 +++++++++++---
>  3 files changed, 15 insertions(+), 6 deletions(-)
>
> diff --git a/include/net/inet_connection_sock.h b/include/net/inet_connec=
tion_sock.h
> index c2b15f7e5516..be3c858a2ebb 100644
> --- a/include/net/inet_connection_sock.h
> +++ b/include/net/inet_connection_sock.h
> @@ -164,7 +164,8 @@ enum inet_csk_ack_state_t {
>         ICSK_ACK_TIMER  =3D 2,
>         ICSK_ACK_PUSHED =3D 4,
>         ICSK_ACK_PUSHED2 =3D 8,
> -       ICSK_ACK_NOW =3D 16       /* Send the next ACK immediately (once)=
 */
> +       ICSK_ACK_NOW =3D 16,      /* Send the next ACK immediately (once)=
 */
> +       ICSK_ACK_NOMEM =3D 32,
>  };
>
>  void inet_csk_init_xmit_timers(struct sock *sk,
> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> index 3cd92035e090..03111af6115d 100644
> --- a/net/ipv4/tcp_input.c
> +++ b/net/ipv4/tcp_input.c
> @@ -5061,7 +5061,8 @@ static void tcp_data_queue(struct sock *sk, struct =
sk_buff *skb)
>                         reason =3D SKB_DROP_REASON_PROTO_MEM;
>                         NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVQDROP=
);
>                         sk->sk_data_ready(sk);
> -                       goto drop;
> +                       inet_csk(sk)->icsk_ack.pending |=3D ICSK_ACK_NOME=
M;

Also set ICSK_ACK_NOW ?

We also need to make sure to send an immediate ACK WIN 0 in the case we que=
ued
the skb in an empty receive queue and we were under pressure.

We do not want to have a delayed ACK sending a normal RWIN,
then wait for another packet that we will probably drop.

Look at the code :

if (skb_queue_len(&sk->sk_receive_queue) =3D=3D 0)
     sk_forced_mem_schedule(sk, skb->truesize);
else if (tcp_try_rmem_schedule(sk, skb, skb->truesize)) {

and refactor it to make sure to set  ICSK_ACK_NOMEM even on the first packe=
t
that bypassed the rmem_schedule().


> +                       goto out_of_window;

Why forcing quickack mode ? Please leave the "goto drop;"

>                 }
>
>                 eaten =3D tcp_queue_rcv(sk, skb, &fragstolen);
> @@ -5102,7 +5103,6 @@ static void tcp_data_queue(struct sock *sk, struct =
sk_buff *skb)
>  out_of_window:
>                 tcp_enter_quickack_mode(sk, TCP_MAX_QUICKACKS);
>                 inet_csk_schedule_ack(sk);
> -drop:
>                 tcp_drop_reason(sk, skb, reason);
>                 return;
>         }
>