[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CADvbK_e+J2nut4Q5NE3oAdUqEDXAFZrecs4zY+CrLE9ob8AtZg@mail.gmail.com>
Date: Fri, 22 Dec 2023 11:08:03 -0500
From: Xin Long <lucien.xin@...il.com>
To: Eric Dumazet <edumazet@...gle.com>
Cc: "David S . Miller" <davem@...emloft.net>, Jakub Kicinski <kuba@...nel.org>,
Paolo Abeni <pabeni@...hat.com>, netdev@...r.kernel.org, eric.dumazet@...il.com,
Jacob Moroni <jmoroni@...gle.com>, Marcelo Ricardo Leitner <marcelo.leitner@...il.com>
Subject: Re: [PATCH net-next] sctp: fix busy polling
On Tue, Dec 19, 2023 at 12:00 PM Eric Dumazet <edumazet@...gle.com> wrote:
>
> Busy polling while holding the socket lock makes litle sense,
> because incoming packets wont reach our receive queue.
>
> Fixes: 8465a5fcd1ce ("sctp: add support for busy polling to sctp protocol")
> Reported-by: Jacob Moroni <jmoroni@...gle.com>
> Signed-off-by: Eric Dumazet <edumazet@...gle.com>
> Cc: Marcelo Ricardo Leitner <marcelo.leitner@...il.com>
> Cc: Xin Long <lucien.xin@...il.com>
> ---
> net/sctp/socket.c | 10 ++++------
> 1 file changed, 4 insertions(+), 6 deletions(-)
>
> diff --git a/net/sctp/socket.c b/net/sctp/socket.c
> index 5fb02bbb4b349ef9ab9c2790cccb30fb4c4e897c..6b9fcdb0952a0fe599ae5d1d1cc6fa9557a3a3bc 100644
> --- a/net/sctp/socket.c
> +++ b/net/sctp/socket.c
> @@ -2102,6 +2102,10 @@ static int sctp_recvmsg(struct sock *sk, struct msghdr *msg, size_t len,
> if (unlikely(flags & MSG_ERRQUEUE))
> return inet_recv_error(sk, msg, len, addr_len);
>
> + if (sk_can_busy_loop(sk) &&
> + skb_queue_empty_lockless(&sk->sk_receive_queue))
> + sk_busy_loop(sk, flags & MSG_DONTWAIT);
> +
Here is no any sk_state check, if the SCTP socket(TCP type) has been
already closed by peer, will sctp_recvmsg() block here?
Maybe here it needs a `!(sk->sk_shutdown & RCV_SHUTDOWN)` check,
which is set when it's closed by the peer.
Thanks
> lock_sock(sk);
>
> if (sctp_style(sk, TCP) && !sctp_sstate(sk, ESTABLISHED) &&
> @@ -9046,12 +9050,6 @@ struct sk_buff *sctp_skb_recv_datagram(struct sock *sk, int flags, int *err)
> if (sk->sk_shutdown & RCV_SHUTDOWN)
> break;
>
> - if (sk_can_busy_loop(sk)) {
> - sk_busy_loop(sk, flags & MSG_DONTWAIT);
> -
> - if (!skb_queue_empty_lockless(&sk->sk_receive_queue))
> - continue;
> - }
>
> /* User doesn't want to wait. */
> error = -EAGAIN;
> --
> 2.43.0.472.g3155946c3a-goog
>
Powered by blists - more mailing lists