I've got a listener socket that accepts, receives and sends as a TCP server typically does. I've given my accept and receive code below, it's not that different from the example on Microsoft's documentation. The main difference is that my server doesn't kill a connection after it stops receiving data (I don't know if this is a bad design or not?).
private void on_accept(IAsyncResult xResult)
{
Socket listener = null;
Socket handler = null;
TStateObject state = null;
Task<int> consumer = null;
try
{
mxResetEvent.Set();
listener = (Socket)xResult.AsyncState;
handler = listener.EndAccept(xResult);
state = new TStateObject()
{
Socket = handler
};
consumer = async_input_consumer(state);
OnConnect?.Invoke(this, handler);
handler.BeginReceive(state.Buffer, 0, TStateObject.BufferSize, 0, new AsyncCallback(on_receive), state);
}
catch (SocketException se)
{
if (se.ErrorCode == 10054)
{
on_disconnect(state);
}
}
catch (ObjectDisposedException)
{
return;
}
catch (Exception ex)
{
System.Console.WriteLine("Exception in TCPServer::AcceptCallback, exception: " + ex.Message);
}
}
private void on_receive(IAsyncResult xResult)
{
Socket handler = null;
TStateObject state = null;
try
{
state = xResult.AsyncState as TStateObject;
handler = state.Socket;
int bytesRead = handler.EndReceive(xResult);
UInt16 id = TClientRegistry.GetIdBySocket(handler);
TContext context = TClientRegistry.GetContext(id);
if (bytesRead > 0)
{
var buffer_data = new byte[bytesRead];
Array.Copy(state.Buffer, buffer_data, bytesRead);
state.BufferBlock.Post(buffer_data);
}
Array.Clear(state.Buffer, 0, state.Buffer.Length);
handler.BeginReceive(state.Buffer, 0, TStateObject.BufferSize, 0, new AsyncCallback(on_receive), state);
}
catch (SocketException se)
{
if(se.ErrorCode == 10054)
{
on_disconnect(state);
}
}
catch (ObjectDisposedException)
{
return;
}
catch (Exception ex)
{
System.Console.WriteLine("Exception in TCPServer::ReadCallback, exception: " + ex.Message);
}
}
This code is used to connect to an embedded device and works (mostly) fine. I was investigating a memory leak and trying to speed up the process a bit by replicating exactly what the device does (our connection speeds are in the realm of about 70kbps to our device, and it took an entire weekend of stress testing to get the memory leak to double the memory footprint of the server).
So I wrote a C# program to replicate the data transactions, but I've run into an issue where when I disconnect the test program, the server gets caught in a loop where it endlessly has its on_receive
callback called. I was under the impression that BeginReceive
wouldn't be triggered until something was received, and it seems to call on_receive
, ends the receiving like an async callback should do, process the data, and then I want the connection to await more data so I call BeginReceive
again.
The part of my test program where the issue occurs is in here:
private static void read_write_test()
{
mxConnection = new Socket(AddressFamily.InterNetwork, SocketType.Stream, ProtocolType.Tcp);
mxConnection.Connect("12.12.12.18", 10);
if (mxConnection.Connected)
{
byte[] data = Encoding.ASCII.GetBytes("HANDSHAKESTRING"); //Connect string
int len = data.Length;
mxConnection.Send(data);
data = new byte[4];
len = mxConnection.Receive(data);
if (len == 0 || data[0] != '1')
{
mxConnection.Disconnect(false);
return;
}
}
//Meat of the test goes here but isn't relevant
mxConnection.Shutdown(SocketShutdown.Both);
mxConnection.Close();
}
Up until the Shutdown(SocketShutdown.Both)
call, everything works as expected. When I make that call however, it seems like the server never gets notification that the client has closed the socket and gets stuck in a loop of endlessly trying to receive. I've done my homework and I think I am closing my connection properly as per this discussion. I've messed around with the disconnect section to just do mxConnection.Disconnect(false)
as well, but the same thing occurs.
When the device disconnects from the server, my server catches a SocketException
with error code 10054, which documentation says:
Connection reset by peer.
An existing connection was forcibly closed by the remote host. This normally results if the peer application on the remote host is suddenly stopped, the host is rebooted, the host or remote network interface is disabled, or the remote host uses a hard close (see setsockopt for more information on the SO_LINGER option on the remote socket). This error may also result if a connection was broken due to keep-alive activity detecting a failure while one or more operations are in progress. Operations that were in progress fail with WSAENETRESET. Subsequent operations fail with WSAECONNRESET.
I've used this to handle the socket being closed and has worked well for the most part. However, with my C# test program, it doesn't seem like it works the same way.
Am I missing something here? I'd appreciate any input. Thanks.