In C#, how to check if a TCP port is available?

You're on the wrong end of the Intertube. It is the server that can have only one particular port open. Some code:

  IPAddress ipAddress = Dns.GetHostEntry("localhost").AddressList[0];
  try {
    TcpListener tcpListener = new TcpListener(ipAddress, 666);
    tcpListener.Start();
  }
  catch (SocketException ex) {
    MessageBox.Show(ex.Message, "kaboom");
  }

Fails with:

Only one usage of each socket address (protocol/network address/port) is normally permitted.

Since you're using a TcpClient, that means you're checking open TCP ports. There are lots of good objects available in the System.Net.NetworkInformation namespace.

Use the IPGlobalProperties object to get to an array of TcpConnectionInformation objects, which you can then interrogate about endpoint IP and port.

---

 int port = 456; //<--- This is your value
 bool isAvailable = true;

 // Evaluate current system tcp connections. This is the same information provided
 // by the netstat command line application, just in .Net strongly-typed object
 // form.  We will look through the list, and if our port we would like to use
 // in our TcpClient is occupied, we will set isAvailable to false.
 IPGlobalProperties ipGlobalProperties = IPGlobalProperties.GetIPGlobalProperties();
 TcpConnectionInformation[] tcpConnInfoArray = ipGlobalProperties.GetActiveTcpConnections();

 foreach (TcpConnectionInformation tcpi in tcpConnInfoArray)
 {
   if (tcpi.LocalEndPoint.Port==port)
   {
     isAvailable = false;
     break;
   }
 }

 // At this point, if isAvailable is true, we can proceed accordingly.

When you set up a TCP connection, the 4-tuple (source-ip, source-port, dest-ip, dest-port) has to be unique - this is to ensure packets are delivered to the right place.

There is a further restriction on the server side that only one server program can bind to an incoming port number (assuming one IP address; multi-NIC servers have other powers but we don't need to discuss them here).

So, at the server end, you:

• create a socket.
• bind that socket to a port.
• listen on that port.
• accept connections on that port. and there can be multiple connections coming in (one per client).

On the client end, it's usually a little simpler:

• create a socket.
• open the connection. When a client opens the connection, it specifies the ip address and port of the server. It can specify its source port but usually uses zero which results in the system assigning it a free port automatically.

There is no requirement that the destination IP/port be unique since that would result in only one person at a time being able to use Google, and that would pretty well destroy their business model.

This means you can even do such wondrous things as multi-session FTP since you set up multiple sessions where the only difference is your source port, allowing you to download chunks in parallel. Torrents are a little different in that the destination of each session is usually different.

And, after all that waffling (sorry), the answer to your specific question is that you don't need to specify a free port. If you're connecting to a server with a call that doesn't specify your source port, it'll almost certainly be using zero under the covers and the system will give you an unused one.