Correctly Configuring MSS in SYN-ACK Handshake

[Orvaxis]

The MSS is not returned when the current handshake is returned, causing the downstream upload to use a default value as the upload speed, resulting in a significantly lower upload speed than the download test.

Reproduce using iperf3 -c 10.3.0.1
And you can see through wireshark that iperf3 has been sending data packets with a content length of 512 to tun.

pub(crate) fn create_raw_packet(
    src_addr: SocketAddr,
    dst_addr: SocketAddr,
    calculate_payload_max_len: impl Fn(usize, usize) -> usize,
    flags: u8,
    ttl: u8,
    seq: u32,
    ack: u32,
    win: u16,
    mut payload: Vec<u8>,
) -> std::io::Result<NetworkPacket> {
    let mut tcp_header = etherparse::TcpHeader::new(src_addr.port(), dst_addr.port(), seq, win);
    tcp_header.acknowledgment_number = ack;
    tcp_header.syn = flags & SYN != 0;
    tcp_header.ack = flags & ACK != 0;
    tcp_header.rst = flags & RST != 0;
    tcp_header.fin = flags & FIN != 0;
    tcp_header.psh = flags & PSH != 0;
    tcp_header
        .set_options(&[TcpOptionElement::MaximumSegmentSize(3960)])
        .map_err(|e| std::io::Error::new(InvalidInput, e))?;
...

A simple hard-coded result will be an improvement from 400Mbps to 2.xGbps

[  5]   0.00-10.00  sec  3.63 GBytes  3.12 Gbits/sec    0             sender
[  5]   0.00-10.00  sec  2.41 GBytes  2.07 Gbits/sec                  receiver

ps. It may be that the lack of flow control causes the sending speed to be higher than the receiving speed.

[SajjadPourali]

Hello,

Thanks for the point. I plan to add a struct for TcpConfig, which will include TcpOptions to apply the mentioned config.

[Orvaxis]

Hello,

Thanks for the point. I plan to add a struct for TcpConfig, which will include TcpOptions to apply the mentioned config.

That's great news! Thank you for working on this.

[SajjadPourali]

#65