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: <20230807134547.2782227-1-imagedong@tencent.com>
 <20230807134547.2782227-2-imagedong@tencent.com> <CANn89iJ762Y3KvL26-3s9vkZdkWi9PoJXhzLHr3+5v9Ti47gTw@mail.gmail.com>
In-Reply-To: <CANn89iJ762Y3KvL26-3s9vkZdkWi9PoJXhzLHr3+5v9Ti47gTw@mail.gmail.com>
From:   Menglong Dong <menglong8.dong@gmail.com>
Date:   Tue, 8 Aug 2023 09:58:16 +0800
Message-ID: <CADxym3ZqELjAz8iYwkgqndG6Ggto0DCU6g-ffbSitNWO6tZ7ng@mail.gmail.com>
Subject: Re: [PATCH net-next v2 1/3] net: tcp: send zero-window ACK when no memory
To:     Eric Dumazet <edumazet@google.com>
Cc:     ncardwell@google.com, 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 Mon, Aug 7, 2023 at 10:05=E2=80=AFPM Eric Dumazet <edumazet@google.com> =
wrote:
>
> On Mon, Aug 7, 2023 at 3:47=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>
> > ---
> > v2:
> > - send 0 rwin ACK for the receive queue empty case when necessary
> > - send the ACK immediately by using the ICSK_ACK_NOW flag
> > ---
> >  include/net/inet_connection_sock.h |  3 ++-
> >  net/ipv4/tcp_input.c               | 14 +++++++++++---
> >  net/ipv4/tcp_output.c              | 14 +++++++++++---
> >  3 files changed, 24 insertions(+), 7 deletions(-)
> >
> > diff --git a/include/net/inet_connection_sock.h b/include/net/inet_conn=
ection_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 (onc=
e) */
> > +       ICSK_ACK_NOW =3D 16,      /* Send the next ACK immediately (onc=
e) */
> > +       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 8e96ebe373d7..aae485d0a3b6 100644
> > --- a/net/ipv4/tcp_input.c
> > +++ b/net/ipv4/tcp_input.c
> > @@ -5059,12 +5059,20 @@ static void tcp_data_queue(struct sock *sk, str=
uct sk_buff *skb)
> >
> >                 /* Ok. In sequence. In window. */
> >  queue_and_out:
> > -               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))=
 {
> > +               if (skb_queue_len(&sk->sk_receive_queue) =3D=3D 0) {
> > +                       if (tcp_try_rmem_schedule(sk, skb, skb->truesiz=
e)) {
> > +                               sk_forced_mem_schedule(sk, skb->truesiz=
e);
>
> I think we want sk->sk_data_ready() here, to let applications drain the q=
ueue,
> regardless of sk->sk_rcvlowat value.
>
>
> > +                               inet_csk(sk)->icsk_ack.pending |=3D
> > +                                       (ICSK_ACK_NOMEM | ICSK_ACK_NOW)=
;
> > +                               inet_csk_schedule_ack(sk);
> > +                       }
> > +               } else if (tcp_try_rmem_schedule(sk, skb, skb->truesize=
)) {
> >                         reason =3D SKB_DROP_REASON_PROTO_MEM;
> >                         NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVQDR=
OP);
> >                         sk->sk_data_ready(sk);
>
> We also want to keep this sk->sk_data_ready(sk) call.
>
> > +                       inet_csk(sk)->icsk_ack.pending |=3D
> > +                               (ICSK_ACK_NOMEM | ICSK_ACK_NOW);
> > +                       inet_csk_schedule_ack(sk);
> >                         goto drop;
> >                 }
>
> This would suggest a different code refactoring, to avoid code duplicatio=
n.
>
> diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
> index 57c8af1859c16eba5e952a23ea959b628006f9c1..dde6c44f2c1e33dcf60c23b49=
cd99f270874ca96
> 100644
> --- a/net/ipv4/tcp_input.c
> +++ b/net/ipv4/tcp_input.c
> @@ -5050,13 +5050,17 @@ static void tcp_data_queue(struct sock *sk,
> struct sk_buff *skb)
>
>                 /* Ok. In sequence. In window. */
>  queue_and_out:
> -               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)) {
> -                       reason =3D SKB_DROP_REASON_PROTO_MEM;
> -                       NET_INC_STATS(sock_net(sk), LINUX_MIB_TCPRCVQDROP=
);
> +               if (tcp_try_rmem_schedule(sk, skb, skb->truesize)) {
> +                       /* TODO: maybe ratelimit these WIN 0 ACK ? */
> +                       inet_csk(sk)->icsk_ack.pending |=3D
> CSK_ACK_NOMEM | ICSK_ACK_NOW;
> +                       inet_csk_schedule_ack(sk);
>                         sk->sk_data_ready(sk);
> -                       goto drop;
> +                       if (skb_queue_len(&sk->sk_receive_queue)) {
> +                               reason =3D SKB_DROP_REASON_PROTO_MEM;
> +                               NET_INC_STATS(sock_net(sk),
> LINUX_MIB_TCPRCVQDROP);
> +                               goto drop;
> +                       }
> +                       sk_forced_mem_schedule(sk, skb->truesize);
>                 }
>
>                 eaten =3D tcp_queue_rcv(sk, skb, &fragstolen);

Looks much better, thank you!