Need help setting udp hole punching for p2p data transfer. Can't get STUN to work

Question

I'm trying to send UDP packets from one client to another. Let C1 be the client I'm using to receive data and C2 the client who will send the data. What I'm doing:

1. Ask a STUN server for the public IP addresses and ports of both clients.

2. I inform each client of the others IP address and public port.

3. C1 sends a packet to C2 on it's public address and port and starts listening for packets. At the same time C2 starts sending packets with the data I want to be transmited.

C1 just hangs on receiving a packet that never arrives. am I doing the UDP hole puching wrong?

I am testing this with both clients on the same computer on the same network but the server is on a server on another network with a public IP. What I'm trying to say is that the public address for both clients is the same while I'm testing, is this a problem?

UPDATE:

I know now that my router allows for hairpinning. I fowarded the UDP ports on my router and using the public IP address I can send packets from one client to the other.

This is the code I'm using to test if I can do hole punching in C#:

static UdpClient udpClient = new UdpClient(30001);
static IPAddress remoteIp;
static int remotePort;

static void Main(string[] args)
{
        // Send packet to server so it can find the client's public address and port 
        byte[] queryServer = Encoding.UTF8.GetBytes("something");
        udpClient.Send(testMsg, testMsg.Length, new IPEndPoint(IPAddress.Parse("199.254.173.154"), 30000));

        // Wait for a response from the server with the other client's public address and port
        IPEndPoint sender = new IPEndPoint(IPAddress.Any, 0);           
        byte[] dataJsonBytes = udpClient.Receive(ref sender);
        JObject json = JObject.Parse(Encoding.UTF8.GetString(dataJsonBytes));
        remoteIp = IPAddress.Parse(json["IP"].ToString());
        remotePort = Int32.Parse(json["port"].ToString());                                                    

        // Start spamming the other client with packets and waiting for response
        Thread sendingThread = new Thread(sendPacket);
        Thread receivingThread = new Thread(receivePacket);

        sendingThread.Start();
        receivingThread.Start();

        Console.ReadLine();
    }

    private static void sendPacket()
    {
        byte[] udpMessage = Encoding.UTF8.GetBytes("Forcing udp hole punching!!\n");
        while (true)
        {               
            udpClient.Send(udpMessage, udpMessage.Length, new IPEndPoint(remoteIp, remotePort));
            Thread.Sleep(100);
        }
    }

    private static void receivePacket()
    {
        IPEndPoint senderEp = new IPEndPoint(IPAddress.Any, 0);
        udpClient.Client.ReceiveTimeout = 500;
        while (true)
        {
            try
            {
                byte[] receivedData = udpClient.Receive(ref senderEp);
                Console.Write(Encoding.UTF8.GetString(receivedData));
            }
            catch (Exception)
            {
                Console.Write("Timeout!!\n");
            }
        }
    }

UPDATE 2:

With the help of @Tahlil I was able to figure out that the algorithm I used to check the NAT type was flawed and I was trying to do hole punching behind a symmetric NAT. So the problem was not noticing a mistake on my code... I recommend asserting the NAT type with certainty to all that face the same problem.

Thank you all for your help.

Answer 1

Some NAT doesn't allow packets from unknown sources. That's why we punch hole in the NAT so that the NAT can recognize that unknown IP and creates a mapping for it.

If C1 and C2 are in different network. Then if C2 sent a packet to C1s public IP:port then C2's NAT will create a mapping for C1's public IP:Port. After that any data from C1 in C2's public IP:Port would have been forwarded to C2.

But I think in your scenario the packets are not leaving the NAT because they are both under same network. Thats why the NAT is not creating any mapping. So your hole punching is failing.

So when you send data to C2's public IP address from C1, you are assuming by that time C2 has punched a hole for C1s address and NAT has created a mapping. But C2's hole punching has been failed as C1 is in same network.

And why do you need to punch hole if you are in same network? This is a bizarre intention as C1 and C2 can directly send packets to each others private IP and establish a connection. Hole punching is needed when C1 and C2 is in different network.

EDIT:

Hole punching will not work if the NAT combination of two clients is like Symmetric vs Symmetric. So first check what is the NAT types of both end.

Do not spam the other end with packets in order to punch a hole. NAT can block your IP if you flood it with packets. Do hole punching from one side and with low TTL value. So that the packets intended for hole punching leave the NAT but doesn't reach the other NAT. This way hole punching will be done successfully and other side NAT won't be flooded. From the other side try sending packet a little late so that hole punch is completed from the other side. When hole punched side receives first packet then stop the hole punching and start sending packet normally.

Answer 2

Of course C2 will hang. Because the C2 NAT likely rejected the original packet sent by C1 before C2 sent his packet. Build some retry logic into all of this to overcome these issues.

Also, if C1 and C2 are on the same subnet, then it's possible your NAT just doesn't allow hairpinning. "Hairpinning" is the ability to send internally through your external address. Not all NATs support this. Be prepared to try to connect with both internal candidate addresses as well as STUN discovered external addresses.

In any case, start with the answer I've given out before.

https://stackoverflow.com/a/8524609/104458