[MPTCP] Re: [PATCH v2] mptcp: fix state tracking for fallback socket

[Mat Martineau <mathew.j.martineau at linux.intel.com>]

[-- Attachment #1: Type: text/plain, Size: 1967 bytes --]

On Tue, 17 Nov 2020, Paolo Abeni wrote:

> We need to cope with some more state transition for
> fallback sockets, or could still end-up moving to TCP_CLOSE
> too early and avoid spooling some pending data
>
> Fixes: ("mptcp: refactor shutdown and close")
> Signed-off-by: Paolo Abeni <pabeni(a)redhat.com>
> ---
> v1 -> v2:
> - we actually need also the state transition I tried to
>   skip in the previous iteration, otherwise we will
>   keep dangling sockets around
> ---
> net/mptcp/protocol.c | 18 +++++++++++++-----
> 1 file changed, 13 insertions(+), 5 deletions(-)
>
> diff --git a/net/mptcp/protocol.c b/net/mptcp/protocol.c
> index 5f7a43564922..3908a6445321 100644
> --- a/net/mptcp/protocol.c
> +++ b/net/mptcp/protocol.c
> @@ -788,7 +788,9 @@ static void mptcp_check_for_eof(struct mptcp_sock *msk)
> 		inet_sk_state_store(sk, TCP_CLOSE_WAIT);
> 		break;
> 	case TCP_FIN_WAIT1:
> -		/* fallback sockets skip TCP_CLOSING - TCP will take care */
> +		inet_sk_state_store(sk, TCP_CLOSING);
> +		break;
> +	case TCP_FIN_WAIT2:
> 		inet_sk_state_store(sk, TCP_CLOSE);
> 		break;
> 	default:
> @@ -2110,10 +2112,16 @@ static void __mptcp_check_send_data_fin(struct sock *sk)
>
> 	WRITE_ONCE(msk->snd_nxt, msk->write_seq);
>
> -	/* fallback socket will not get data_fin/ack, can move to close now */
> -	if (__mptcp_check_fallback(msk) && sk->sk_state == TCP_LAST_ACK) {
> -		inet_sk_state_store(sk, TCP_CLOSE);
> -		mptcp_close_wake_up(sk);
> +	/* fallback socket will not get data_fin/ack, can move to the next
> +	 * stats now

'state' ?

Other than that, looks good to me. Thanks Paolo.

> +	 */
> +	if (__mptcp_check_fallback(msk)) {
> +		if ((1 << sk->sk_state) & (TCPF_CLOSING | TCPF_LAST_ACK)) {
> +			inet_sk_state_store(sk, TCP_CLOSE);
> +			mptcp_close_wake_up(sk);
> +		} else if (sk->sk_state == TCP_FIN_WAIT1) {
> +			inet_sk_state_store(sk, TCP_FIN_WAIT2);
> +		}

--
Mat Martineau
Intel