Unable to implement tcp handshake via libpnet

[rikonaka]

Unable to implement tcp handshake via libpnet

[mkeeter]

I've run into similar issues, and found that opening an unused dummy socket will prevent the OS from automatically replying:

    // Open a bogus socket to listen on port 2000, which prevents the OS from
    // replying with ICMPv6 messages about the port being unreachable.
    let _socket = std::net::UdpSocket::bind(format!("[::]:{SOURCE_PORT}"))?;

This is for IPv6 + UDP, but you might be able to do something similar for IPv4 + TCP.

[rikonaka]

I've run into similar issues, and found that opening an unused dummy socket will prevent the OS from automatically replying:

    // Open a bogus socket to listen on port 2000, which prevents the OS from
    // replying with ICMPv6 messages about the port being unreachable.
    let _socket = std::net::UdpSocket::bind(format!("[::]:{SOURCE_PORT}"))?;

This is for IPv6 + UDP, but you might be able to do something similar for IPv4 + TCP.

Thank you very much for your reply 😜, I haven't tested it yet, but I have a stupid question, will this socket conflict with libpnet?

[mkeeter]

In my testing, it didn't interfere. I was using pnet::datalink::channel to build a (tx, rx) pair, which gives you every packet on the interface; then, you have to deconstruct + filter them manually. For my code, it looked like this:

        let start = Instant::now();
        while Instant::now() - start < timeout {
            let Ok(rx) = self.receiver.next() else { continue; };
            let packet = EthernetPacket::new(rx).unwrap();
            if EtherTypes::Ipv6 == packet.get_ethertype() {
                let header = Ipv6Packet::new(packet.payload()).unwrap();
                if IpNextHeaderProtocols::Udp == header.get_next_header() {
                    let udp = UdpPacket::new(header.payload()).unwrap();
                    if udp.get_destination() == SOURCE_PORT {
                        return Some(udp.payload().to_owned());
                    }
                }
            }
        }

[rikonaka]

In my testing, it didn't interfere. I was using pnet::datalink::channel to build a (tx, rx) pair, which gives you every packet on the interface; then, you have to deconstruct + filter them manually. For my code, it looked like this:

        let start = Instant::now();
        while Instant::now() - start < timeout {
            let Ok(rx) = self.receiver.next() else { continue; };
            let packet = EthernetPacket::new(rx).unwrap();
            if EtherTypes::Ipv6 == packet.get_ethertype() {
                let header = Ipv6Packet::new(packet.payload()).unwrap();
                if IpNextHeaderProtocols::Udp == header.get_next_header() {
                    let udp = UdpPacket::new(header.payload()).unwrap();
                    if udp.get_destination() == SOURCE_PORT {
                        return Some(udp.payload().to_owned());
                    }
                }
            }
        }

Thank you very much for your patience in replying, it seems that this is an unfixable bug that can only be remedied with a few tricks. 🤪