Related
In an application I'm working on I want to disconnect clients that are trying to send me packets that are too large.
Just before disconnecting them I want to send them a message informing them about the reason for disconnecting them.
The issue I am running into is that the client cannot receive this server message, if the server does not read everything the client has send him first. I do not understand why this is happening.
I've managed to narrow it down to a very small test setup where the problem is demonstrated.
The StreamUtil class is a simple wrapper class that helps to get around the TCP message boundary problem, basically on the sender side it sends the size of each message first and then the message itself, and on the receiver side it receives the size of the message first and then the message.
The client uses a ReadKey command to simulate some time between sending and receiving, seeing in my real application these two actions are not immediately back to back either.
Here is a test case that works:
Run server as shown below
Run client as shown below, it will show a "Press key message", WAIT do not press key yet
Turn off server since everything is already in the clients receive buffer anyway (I validated this using packet sniffer)
Press key on the client -> client correctly shows the messages from the server.
This is what I was expecting, so great so far no problem yet.
Now in the server code, comment out the 2nd receive call and repeat the steps above.
Step 1 and 2 complete successfully, no errors sending from client to server.
On step 3 however the client crashes on the read from the server, EVEN though the server reply HAS arrived on the client (again validated with packet sniffer).
If I do a partial shutdown (eg socket.Shutdown (...send...)) without closing the socket on the server, everything works.
1: I just cannot get my head around WHY not processing the line of text from the client on the server causes the client to fail on receiving the text send back from the server.
2: If I send content from server to client but STOP the server before actually closing the socket, this content never arrives, but the bytes have already been transmitted to the server side... (see ReadKey in server to simulate, basically I block there and then just quit the server)
If anyone could shed light on these two issues, I'd deeply appreciate it.
Client:
class TcpClientDemo
{
public static void Main (string[] args)
{
Console.WriteLine ("Starting....");
TcpClient client = new TcpClient();
try
{
client.Connect("localhost", 56789);
NetworkStream stream = client.GetStream();
StreamUtil.SendString(stream, "Client teststring...");
Console.WriteLine("Press key to initiate receive...");
Console.ReadKey();
Console.WriteLine("server reply:" + StreamUtil.ReceiveString(stream));
}
catch (Exception e)
{
Console.WriteLine(e.Message);
}
finally
{
client.Close();
}
Console.WriteLine("Client ended");
Console.ReadKey(true);
}
}
Server:
class TcpServerDemo
{
public static void Main (string[] args)
{
TcpListener listener = new TcpListener (IPAddress.Any, 56789);
listener.Start ();
Console.WriteLine ("Waiting for clients to serve...");
while (true)
{
TcpClient client = null;
NetworkStream stream = null;
try
{
client = listener.AcceptTcpClient();
stream = client.GetStream();
//question 1: Why does commenting this line prevent the client from receiving the server reply??
Console.WriteLine("client string:" + StreamUtil.ReceiveString(stream));
StreamUtil.SendString(stream, "...Server reply goes here...");
//question 2: If I close the server program without actually calling client.Close (while on this line), the client program crashes as well, why?
//Console.ReadKey();
}
catch (Exception e)
{
Console.WriteLine(e.Message);
break;
}
finally
{
if (stream != null) stream.Close();
if (client != null) client.Close();
Console.WriteLine("Done serving this client, everything closed.");
}
}
listener.Stop();
Console.WriteLine("Server ended.");
Console.ReadKey(true);
}
}
StreamUtil:
public class StreamUtil
{
public static byte[] ReadBytes (NetworkStream pStream, int byteCount) {
byte[] bytes = new byte[byteCount];
int bytesRead = 0;
int totalBytesRead = 0;
try {
while (
totalBytesRead != byteCount &&
(bytesRead = pStream.Read (bytes, totalBytesRead, byteCount - totalBytesRead)) > 0
) {
totalBytesRead += bytesRead;
Console.WriteLine("Read/Total:" + bytesRead + "/" + totalBytesRead);
}
} catch (Exception e) {
Console.WriteLine(e.Message);
}
return (totalBytesRead == byteCount) ? bytes : null;
}
public static void SendString (NetworkStream pStream, string pMessage) {
byte[] sendPacket = Encoding.ASCII.GetBytes (pMessage);
pStream.Write (BitConverter.GetBytes (sendPacket.Length), 0, 4);
pStream.Write (sendPacket, 0, sendPacket.Length);
}
public static string ReceiveString (NetworkStream pStream) {
int byteCountToRead = BitConverter.ToInt32(ReadBytes (pStream, 4), 0);
Console.WriteLine("Byte count to read:"+byteCountToRead);
byte[] receivePacket = ReadBytes (pStream, byteCountToRead);
return Encoding.ASCII.GetString (receivePacket);
}
}
The client fails because it detects the socket was already closed.
If C# socket operations detect a closed connection during earlier operations, an exception is thrown on the next operation which can mask data which would otherwise have been received
The StreamUtil class does a couple of things when the connection is closed before/during a read:
Exceptions from the reads are swallowed
A read of zero bytes isn't treated
These obfuscate what's happening when an unexpected close hits the client.
Changing ReadBytes not to swallow exceptions and to throw a mock socket-closed exception (e.g. if (bytesRead == 0) throw new SocketException(10053);) when it reads zero bytes I think makes the outcome more clear.
Edit
I missed something subtle in your examples - your first example causes a TCP RST flag to be sent as soon as the server closes connection, due to the socket being closed with data waiting to be read.
The RST flag results in a closedown that doesn't preserve pending data.
This blog has some discussion based on a very similar scenario (web server sending a HTTP error).
So I don't think there's an easy fix, options are:
As you already tried, shutdown the socket on the server before closing to force a FIN to be sent before the RST
Read the data in question but never process it (taking up bandwidth for no reason)
How can I detect that a client has disconnected from my server?
I have the following code in my AcceptCallBack method
static Socket handler = null;
public static void AcceptCallback(IAsyncResult ar)
{
//Accept incoming connection
Socket listener = (Socket)ar.AsyncState;
handler = listener.EndAccept(ar);
}
I need to find a way to discover as soon as possible that the client has disconnected from the handler Socket.
I've tried:
handler.Available;
handler.Send(new byte[1], 0,
SocketFlags.None);
handler.Receive(new byte[1], 0,
SocketFlags.None);
The above approaches work when you are connecting to a server and want to detect when the server disconnects but they do not work when you are the server and want to detect client disconnection.
Any help will be appreciated.
Since there are no events available to signal when the socket is disconnected, you will have to poll it at a frequency that is acceptable to you.
Using this extension method, you can have a reliable method to detect if a socket is disconnected.
static class SocketExtensions
{
public static bool IsConnected(this Socket socket)
{
try
{
return !(socket.Poll(1, SelectMode.SelectRead) && socket.Available == 0);
}
catch (SocketException) { return false; }
}
}
Someone mentioned keepAlive capability of TCP Socket.
Here it is nicely described:
http://tldp.org/HOWTO/TCP-Keepalive-HOWTO/overview.html
I'm using it this way: after the socket is connected, I'm calling this function, which sets keepAlive on. The keepAliveTime parameter specifies the timeout, in milliseconds, with no activity until the first keep-alive packet is sent. The keepAliveInterval parameter specifies the interval, in milliseconds, between when successive keep-alive packets are sent if no acknowledgement is received.
void SetKeepAlive(bool on, uint keepAliveTime, uint keepAliveInterval)
{
int size = Marshal.SizeOf(new uint());
var inOptionValues = new byte[size * 3];
BitConverter.GetBytes((uint)(on ? 1 : 0)).CopyTo(inOptionValues, 0);
BitConverter.GetBytes((uint)keepAliveTime).CopyTo(inOptionValues, size);
BitConverter.GetBytes((uint)keepAliveInterval).CopyTo(inOptionValues, size * 2);
socket.IOControl(IOControlCode.KeepAliveValues, inOptionValues, null);
}
I'm also using asynchronous reading:
socket.BeginReceive(packet.dataBuffer, 0, 128,
SocketFlags.None, new AsyncCallback(OnDataReceived), packet);
And in callback, here is caught timeout SocketException, which raises when socket doesn't get ACK signal after keep-alive packet.
public void OnDataReceived(IAsyncResult asyn)
{
try
{
SocketPacket theSockId = (SocketPacket)asyn.AsyncState;
int iRx = socket.EndReceive(asyn);
}
catch (SocketException ex)
{
SocketExceptionCaught(ex);
}
}
This way, I'm able to safely detect disconnection between TCP client and server.
This is simply not possible. There is no physical connection between you and the server (except in the extremely rare case where you are connecting between two compuers with a loopback cable).
When the connection is closed gracefully, the other side is notified. But if the connection is disconnected some other way (say the users connection is dropped) then the server won't know until it times out (or tries to write to the connection and the ack times out). That's just the way TCP works and you have to live with it.
Therefore, "instantly" is unrealistic. The best you can do is within the timeout period, which depends on the platform the code is running on.
EDIT:
If you are only looking for graceful connections, then why not just send a "DISCONNECT" command to the server from your client?
"That's just the way TCP works and you have to live with it."
Yup, you're right. It's a fact of life I've come to realize. You will see the same behavior exhibited even in professional applications utilizing this protocol (and even others). I've even seen it occur in online games; you're buddy says "goodbye", and he appears to be online for another 1-2 minutes until the server "cleans house".
You can use the suggested methods here, or implement a "heartbeat", as also suggested. I choose the former. But if I did choose the latter, I'd simply have the server "ping" each client every so often with a single byte, and see if we have a timeout or no response. You could even use a background thread to achieve this with precise timing. Maybe even a combination could be implemented in some sort of options list (enum flags or something) if you're really worried about it. But it's no so big a deal to have a little delay in updating the server, as long as you DO update. It's the internet, and no one expects it to be magic! :)
Implementing heartbeat into your system might be a solution. This is only possible if both client and server are under your control. You can have a DateTime object keeping track of the time when the last bytes were received from the socket. And assume that the socket not responded over a certain interval are lost. This will only work if you have heartbeat/custom keep alive implemented.
I've found quite useful, another workaround for that!
If you use asynchronous methods for reading data from the network socket (I mean, use BeginReceive - EndReceive methods), whenever a connection is terminated; one of these situations appear: Either a message is sent with no data (you can see it with Socket.Available - even though BeginReceive is triggered, its value will be zero) or Socket.Connected value becomes false in this call (don't try to use EndReceive then).
I'm posting the function I used, I think you can see what I meant from it better:
private void OnRecieve(IAsyncResult parameter)
{
Socket sock = (Socket)parameter.AsyncState;
if(!sock.Connected || sock.Available == 0)
{
// Connection is terminated, either by force or willingly
return;
}
sock.EndReceive(parameter);
sock.BeginReceive(..., ... , ... , ..., new AsyncCallback(OnRecieve), sock);
// To handle further commands sent by client.
// "..." zones might change in your code.
}
This worked for me, the key is you need a separate thread to analyze the socket state with polling. doing it in the same thread as the socket fails detection.
//open or receive a server socket - TODO your code here
socket = new Socket(....);
//enable the keep alive so we can detect closure
socket.SetSocketOption(SocketOptionLevel.Socket, SocketOptionName.KeepAlive, true);
//create a thread that checks every 5 seconds if the socket is still connected. TODO add your thread starting code
void MonitorSocketsForClosureWorker() {
DateTime nextCheckTime = DateTime.Now.AddSeconds(5);
while (!exitSystem) {
if (nextCheckTime < DateTime.Now) {
try {
if (socket!=null) {
if(socket.Poll(5000, SelectMode.SelectRead) && socket.Available == 0) {
//socket not connected, close it if it's still running
socket.Close();
socket = null;
} else {
//socket still connected
}
}
} catch {
socket.Close();
} finally {
nextCheckTime = DateTime.Now.AddSeconds(5);
}
}
Thread.Sleep(1000);
}
}
The example code here
http://msdn.microsoft.com/en-us/library/system.net.sockets.socket.connected.aspx
shows how to determine whether the Socket is still connected without sending any data.
If you called Socket.BeginReceive() on the server program and then the client closed the connection "gracefully", your receive callback will be called and EndReceive() will return 0 bytes. These 0 bytes mean that the client "may" have disconnected. You can then use the technique shown in the MSDN example code to determine for sure whether the connection was closed.
Expanding on comments by mbargiel and mycelo on the accepted answer, the following can be used with a non-blocking socket on the server end to inform whether the client has shut down.
This approach does not suffer the race condition that affects the Poll method in the accepted answer.
// Determines whether the remote end has called Shutdown
public bool HasRemoteEndShutDown
{
get
{
try
{
int bytesRead = socket.Receive(new byte[1], SocketFlags.Peek);
if (bytesRead == 0)
return true;
}
catch
{
// For a non-blocking socket, a SocketException with
// code 10035 (WSAEWOULDBLOCK) indicates no data available.
}
return false;
}
}
The approach is based on the fact that the Socket.Receive method returns zero immediately after the remote end shuts down its socket and we've read all of the data from it. From Socket.Receive documentation:
If the remote host shuts down the Socket connection with the Shutdown method, and all available data has been received, the Receive method will complete immediately and return zero bytes.
If you are in non-blocking mode, and there is no data available in the protocol stack buffer, the Receive method will complete immediately and throw a SocketException.
The second point explains the need for the try-catch.
Use of the SocketFlags.Peek flag leaves any received data untouched for a separate receive mechanism to read.
The above will work with a blocking socket as well, but be aware that the code will block on the Receive call (until data is received or the receive timeout elapses, again resulting in a SocketException).
Above answers can be summarized as follow :
Socket.Connected properity determine socket state depend on last read or receive state so it can't detect current disconnection state until you manually close the connection or remote end gracefully close of socket (shutdown).
So we can use the function below to check connection state:
bool IsConnected(Socket socket)
{
try
{
if (socket == null) return false;
return !((socket.Poll(5000, SelectMode.SelectRead) && socket.Available == 0) || !socket.Connected);
}
catch (SocketException)
{
return false;
}
//the above code is short exp to :
/* try
{
bool state1 = socket.Poll(5000, SelectMode.SelectRead);
bool state2 = (socket.Available == 0);
if ((state1 && state2) || !socket.Connected)
return false;
else
return true;
}
catch (SocketException)
{
return false;
}
*/
}
Also the above check need to care about poll respone time(block time)
Also as said by Microsoft Documents : this poll method "can't detect proplems like a broken netwrok cable or that remote host was shut down ungracefuuly".
also as said above there is race condition between socket.poll and socket.avaiable which may give false disconnect.
The best way as said by Microsoft Documents is to attempt to send or recive data to detect these kinds of errors as MS docs said.
The below code is from Microsoft Documents :
// This is how you can determine whether a socket is still connected.
bool IsConnected(Socket client)
{
bool blockingState = client.Blocking; //save socket blocking state.
bool isConnected = true;
try
{
byte [] tmp = new byte[1];
client.Blocking = false;
client.Send(tmp, 0, 0); //make a nonblocking, zero-byte Send call (dummy)
//Console.WriteLine("Connected!");
}
catch (SocketException e)
{
// 10035 == WSAEWOULDBLOCK
if (e.NativeErrorCode.Equals(10035))
{
//Console.WriteLine("Still Connected, but the Send would block");
}
else
{
//Console.WriteLine("Disconnected: error code {0}!", e.NativeErrorCode);
isConnected = false;
}
}
finally
{
client.Blocking = blockingState;
}
//Console.WriteLine("Connected: {0}", client.Connected);
return isConnected ;
}
//and heres comments from microsoft docs*
The socket.Connected property gets the connection state of the Socket as of the last I/O operation. When it returns false, the Socket was either never connected, or is no longer connected.
Connected is not thread-safe; it may return true after an operation is aborted when the Socket is disconnected from another thread.
The value of the Connected property reflects the state of the connection as of the most recent operation.
If you need to determine the current state of the connection, make a nonblocking, zero-byte Send call. If the call returns successfully or throws a WAEWOULDBLOCK error code (10035), then the socket is still connected; //otherwise, the socket is no longer connected .
Can't you just use Select?
Use select on a connected socket. If the select returns with your socket as Ready but the subsequent Receive returns 0 bytes that means the client disconnected the connection. AFAIK, that is the fastest way to determine if the client disconnected.
I do not know C# so just ignore if my solution does not fit in C# (C# does provide select though) or if I had misunderstood the context.
Using the method SetSocketOption, you will be able to set KeepAlive that will let you know whenever a Socket gets disconnected
Socket _connectedSocket = this._sSocketEscucha.EndAccept(asyn);
_connectedSocket.SetSocketOption(SocketOptionLevel.Socket, SocketOptionName.KeepAlive, 1);
http://msdn.microsoft.com/en-us/library/1011kecd(v=VS.90).aspx
Hope it helps!
Ramiro Rinaldi
i had same problem , try this :
void client_handler(Socket client) // set 'KeepAlive' true
{
while (true)
{
try
{
if (client.Connected)
{
}
else
{ // client disconnected
break;
}
}
catch (Exception)
{
client.Poll(4000, SelectMode.SelectRead);// try to get state
}
}
}
This is in VB, but it seems to work well for me. It looks for a 0 byte return like the previous post.
Private Sub RecData(ByVal AR As IAsyncResult)
Dim Socket As Socket = AR.AsyncState
If Socket.Connected = False And Socket.Available = False Then
Debug.Print("Detected Disconnected Socket - " + Socket.RemoteEndPoint.ToString)
Exit Sub
End If
Dim BytesRead As Int32 = Socket.EndReceive(AR)
If BytesRead = 0 Then
Debug.Print("Detected Disconnected Socket - Bytes Read = 0 - " + Socket.RemoteEndPoint.ToString)
UpdateText("Client " + Socket.RemoteEndPoint.ToString + " has disconnected from Server.")
Socket.Close()
Exit Sub
End If
Dim msg As String = System.Text.ASCIIEncoding.ASCII.GetString(ByteData)
Erase ByteData
ReDim ByteData(1024)
ClientSocket.BeginReceive(ByteData, 0, ByteData.Length, SocketFlags.None, New AsyncCallback(AddressOf RecData), ClientSocket)
UpdateText(msg)
End Sub
You can also check the .IsConnected property of the socket if you were to poll.
I've been working with windows app store programming in c# recently, and I've come across a problem with sockets.
I need to be able to read data with an unknown length from a DataReader().
It sounds simple enough, but I've not been able to find a solution after a few days of searching.
Here's my current receiving code (A little sloppy, need to clean it up after I find a solution to this problem. And yes, a bit of this is from the Microsoft example)
DataReader reader = new DataReader(args.Socket.InputStream);
try
{
while (true)
{
// Read first 4 bytes (length of the subsequent string).
uint sizeFieldCount = await reader.LoadAsync(sizeof(uint));
if (sizeFieldCount != sizeof(uint))
{
// The underlying socket was closed before we were able to read the whole data.
return;
}
reader.InputStreamOptions
// Read the string.
uint stringLength = reader.ReadUInt32();
uint actualStringLength = await reader.LoadAsync(stringLength);
if (stringLength != actualStringLength)
{
// The underlying socket was closed before we were able to read the whole data.
return;
}
// Display the string on the screen. The event is invoked on a non-UI thread, so we need to marshal
// the text back to the UI thread.
//MessageBox.Show("Received data: " + reader.ReadString(actualStringLength));
MessageBox.updateList(reader.ReadString(actualStringLength));
}
}
catch (Exception exception)
{
// If this is an unknown status it means that the error is fatal and retry will likely fail.
if (SocketError.GetStatus(exception.HResult) == SocketErrorStatus.Unknown)
{
throw;
}
MessageBox.Show("Read stream failed with error: " + exception.Message);
}
You are going down the right lines - read the first INT to find out how many bytes are to be sent.
Franky Boyle is correct - without a signalling mechanism it is impossible to ever know the length of a stream. Thats why it is called a stream!
NO socket implementation (including the WinSock) will ever be clever enough to know when a client has finished sending data. The client could be having a cup of tea half way through sending the data!
Your server and its sockets will never know! What are they going to do? Wait forever? I suppose they could wait until the client had 'closed' the connection? But your client could have had a blue screen and the server will never get that TCP close packet, it will just be sitting there thinking it is getting more data one day?
I have never used a DataReader - i have never even heard of that class! Use NetworkStream instead.
From my memory I have written code like this in the past. I am just typing, no checking of syntax.
using(MemoryStream recievedData = new MemoryStream())
{
using(NetworkStream networkStream = new NetworkStream(connectedSocket))
{
int totalBytesToRead = networkStream.ReadByte();
// This is your mechanism to find out how many bytes
// the client wants to send.
byte[] readBuffer = new byte[1024]; // Up to you the length!
int totalBytesRead = 0;
int bytesReadInThisTcpWindow = 0;
// The length of the TCP window of the client is usually
// the number of bytes that will be pushed through
// to your server in one SOCKET.READ method call.
// For example, if the clients TCP window was 777 bytes, a:
// int bytesRead =
// networkStream.Read(readBuffer, 0, int.Max);
// bytesRead would be 777.
// If they were sending a large file, you would have to make
// it up from the many 777s.
// If it were a small file under 777 bytes, your bytesRead
// would be the total small length of say 500.
while
(
(
bytesReadInThisTcpWindow =
networkStream.Read(readBuffer, 0, readBuffer.Length)
) > 0
)
// If the bytesReadInThisTcpWindow = 0 then the client
// has disconnected or failed to send the promised number
// of bytes in your Windows server internals dictated timeout
// (important to kill it here to stop lots of waiting
// threads killing your server)
{
recievedData.Write(readBuffer, 0, bytesReadInThisTcpWindow);
totalBytesToRead = totalBytesToRead + bytesReadInThisTcpWindow;
}
if(totalBytesToRead == totalBytesToRead)
{
// We have our data!
}
}
}
How can I detect that a client has disconnected from my server?
I have the following code in my AcceptCallBack method
static Socket handler = null;
public static void AcceptCallback(IAsyncResult ar)
{
//Accept incoming connection
Socket listener = (Socket)ar.AsyncState;
handler = listener.EndAccept(ar);
}
I need to find a way to discover as soon as possible that the client has disconnected from the handler Socket.
I've tried:
handler.Available;
handler.Send(new byte[1], 0,
SocketFlags.None);
handler.Receive(new byte[1], 0,
SocketFlags.None);
The above approaches work when you are connecting to a server and want to detect when the server disconnects but they do not work when you are the server and want to detect client disconnection.
Any help will be appreciated.
Since there are no events available to signal when the socket is disconnected, you will have to poll it at a frequency that is acceptable to you.
Using this extension method, you can have a reliable method to detect if a socket is disconnected.
static class SocketExtensions
{
public static bool IsConnected(this Socket socket)
{
try
{
return !(socket.Poll(1, SelectMode.SelectRead) && socket.Available == 0);
}
catch (SocketException) { return false; }
}
}
Someone mentioned keepAlive capability of TCP Socket.
Here it is nicely described:
http://tldp.org/HOWTO/TCP-Keepalive-HOWTO/overview.html
I'm using it this way: after the socket is connected, I'm calling this function, which sets keepAlive on. The keepAliveTime parameter specifies the timeout, in milliseconds, with no activity until the first keep-alive packet is sent. The keepAliveInterval parameter specifies the interval, in milliseconds, between when successive keep-alive packets are sent if no acknowledgement is received.
void SetKeepAlive(bool on, uint keepAliveTime, uint keepAliveInterval)
{
int size = Marshal.SizeOf(new uint());
var inOptionValues = new byte[size * 3];
BitConverter.GetBytes((uint)(on ? 1 : 0)).CopyTo(inOptionValues, 0);
BitConverter.GetBytes((uint)keepAliveTime).CopyTo(inOptionValues, size);
BitConverter.GetBytes((uint)keepAliveInterval).CopyTo(inOptionValues, size * 2);
socket.IOControl(IOControlCode.KeepAliveValues, inOptionValues, null);
}
I'm also using asynchronous reading:
socket.BeginReceive(packet.dataBuffer, 0, 128,
SocketFlags.None, new AsyncCallback(OnDataReceived), packet);
And in callback, here is caught timeout SocketException, which raises when socket doesn't get ACK signal after keep-alive packet.
public void OnDataReceived(IAsyncResult asyn)
{
try
{
SocketPacket theSockId = (SocketPacket)asyn.AsyncState;
int iRx = socket.EndReceive(asyn);
}
catch (SocketException ex)
{
SocketExceptionCaught(ex);
}
}
This way, I'm able to safely detect disconnection between TCP client and server.
This is simply not possible. There is no physical connection between you and the server (except in the extremely rare case where you are connecting between two compuers with a loopback cable).
When the connection is closed gracefully, the other side is notified. But if the connection is disconnected some other way (say the users connection is dropped) then the server won't know until it times out (or tries to write to the connection and the ack times out). That's just the way TCP works and you have to live with it.
Therefore, "instantly" is unrealistic. The best you can do is within the timeout period, which depends on the platform the code is running on.
EDIT:
If you are only looking for graceful connections, then why not just send a "DISCONNECT" command to the server from your client?
"That's just the way TCP works and you have to live with it."
Yup, you're right. It's a fact of life I've come to realize. You will see the same behavior exhibited even in professional applications utilizing this protocol (and even others). I've even seen it occur in online games; you're buddy says "goodbye", and he appears to be online for another 1-2 minutes until the server "cleans house".
You can use the suggested methods here, or implement a "heartbeat", as also suggested. I choose the former. But if I did choose the latter, I'd simply have the server "ping" each client every so often with a single byte, and see if we have a timeout or no response. You could even use a background thread to achieve this with precise timing. Maybe even a combination could be implemented in some sort of options list (enum flags or something) if you're really worried about it. But it's no so big a deal to have a little delay in updating the server, as long as you DO update. It's the internet, and no one expects it to be magic! :)
Implementing heartbeat into your system might be a solution. This is only possible if both client and server are under your control. You can have a DateTime object keeping track of the time when the last bytes were received from the socket. And assume that the socket not responded over a certain interval are lost. This will only work if you have heartbeat/custom keep alive implemented.
I've found quite useful, another workaround for that!
If you use asynchronous methods for reading data from the network socket (I mean, use BeginReceive - EndReceive methods), whenever a connection is terminated; one of these situations appear: Either a message is sent with no data (you can see it with Socket.Available - even though BeginReceive is triggered, its value will be zero) or Socket.Connected value becomes false in this call (don't try to use EndReceive then).
I'm posting the function I used, I think you can see what I meant from it better:
private void OnRecieve(IAsyncResult parameter)
{
Socket sock = (Socket)parameter.AsyncState;
if(!sock.Connected || sock.Available == 0)
{
// Connection is terminated, either by force or willingly
return;
}
sock.EndReceive(parameter);
sock.BeginReceive(..., ... , ... , ..., new AsyncCallback(OnRecieve), sock);
// To handle further commands sent by client.
// "..." zones might change in your code.
}
This worked for me, the key is you need a separate thread to analyze the socket state with polling. doing it in the same thread as the socket fails detection.
//open or receive a server socket - TODO your code here
socket = new Socket(....);
//enable the keep alive so we can detect closure
socket.SetSocketOption(SocketOptionLevel.Socket, SocketOptionName.KeepAlive, true);
//create a thread that checks every 5 seconds if the socket is still connected. TODO add your thread starting code
void MonitorSocketsForClosureWorker() {
DateTime nextCheckTime = DateTime.Now.AddSeconds(5);
while (!exitSystem) {
if (nextCheckTime < DateTime.Now) {
try {
if (socket!=null) {
if(socket.Poll(5000, SelectMode.SelectRead) && socket.Available == 0) {
//socket not connected, close it if it's still running
socket.Close();
socket = null;
} else {
//socket still connected
}
}
} catch {
socket.Close();
} finally {
nextCheckTime = DateTime.Now.AddSeconds(5);
}
}
Thread.Sleep(1000);
}
}
The example code here
http://msdn.microsoft.com/en-us/library/system.net.sockets.socket.connected.aspx
shows how to determine whether the Socket is still connected without sending any data.
If you called Socket.BeginReceive() on the server program and then the client closed the connection "gracefully", your receive callback will be called and EndReceive() will return 0 bytes. These 0 bytes mean that the client "may" have disconnected. You can then use the technique shown in the MSDN example code to determine for sure whether the connection was closed.
Expanding on comments by mbargiel and mycelo on the accepted answer, the following can be used with a non-blocking socket on the server end to inform whether the client has shut down.
This approach does not suffer the race condition that affects the Poll method in the accepted answer.
// Determines whether the remote end has called Shutdown
public bool HasRemoteEndShutDown
{
get
{
try
{
int bytesRead = socket.Receive(new byte[1], SocketFlags.Peek);
if (bytesRead == 0)
return true;
}
catch
{
// For a non-blocking socket, a SocketException with
// code 10035 (WSAEWOULDBLOCK) indicates no data available.
}
return false;
}
}
The approach is based on the fact that the Socket.Receive method returns zero immediately after the remote end shuts down its socket and we've read all of the data from it. From Socket.Receive documentation:
If the remote host shuts down the Socket connection with the Shutdown method, and all available data has been received, the Receive method will complete immediately and return zero bytes.
If you are in non-blocking mode, and there is no data available in the protocol stack buffer, the Receive method will complete immediately and throw a SocketException.
The second point explains the need for the try-catch.
Use of the SocketFlags.Peek flag leaves any received data untouched for a separate receive mechanism to read.
The above will work with a blocking socket as well, but be aware that the code will block on the Receive call (until data is received or the receive timeout elapses, again resulting in a SocketException).
Above answers can be summarized as follow :
Socket.Connected properity determine socket state depend on last read or receive state so it can't detect current disconnection state until you manually close the connection or remote end gracefully close of socket (shutdown).
So we can use the function below to check connection state:
bool IsConnected(Socket socket)
{
try
{
if (socket == null) return false;
return !((socket.Poll(5000, SelectMode.SelectRead) && socket.Available == 0) || !socket.Connected);
}
catch (SocketException)
{
return false;
}
//the above code is short exp to :
/* try
{
bool state1 = socket.Poll(5000, SelectMode.SelectRead);
bool state2 = (socket.Available == 0);
if ((state1 && state2) || !socket.Connected)
return false;
else
return true;
}
catch (SocketException)
{
return false;
}
*/
}
Also the above check need to care about poll respone time(block time)
Also as said by Microsoft Documents : this poll method "can't detect proplems like a broken netwrok cable or that remote host was shut down ungracefuuly".
also as said above there is race condition between socket.poll and socket.avaiable which may give false disconnect.
The best way as said by Microsoft Documents is to attempt to send or recive data to detect these kinds of errors as MS docs said.
The below code is from Microsoft Documents :
// This is how you can determine whether a socket is still connected.
bool IsConnected(Socket client)
{
bool blockingState = client.Blocking; //save socket blocking state.
bool isConnected = true;
try
{
byte [] tmp = new byte[1];
client.Blocking = false;
client.Send(tmp, 0, 0); //make a nonblocking, zero-byte Send call (dummy)
//Console.WriteLine("Connected!");
}
catch (SocketException e)
{
// 10035 == WSAEWOULDBLOCK
if (e.NativeErrorCode.Equals(10035))
{
//Console.WriteLine("Still Connected, but the Send would block");
}
else
{
//Console.WriteLine("Disconnected: error code {0}!", e.NativeErrorCode);
isConnected = false;
}
}
finally
{
client.Blocking = blockingState;
}
//Console.WriteLine("Connected: {0}", client.Connected);
return isConnected ;
}
//and heres comments from microsoft docs*
The socket.Connected property gets the connection state of the Socket as of the last I/O operation. When it returns false, the Socket was either never connected, or is no longer connected.
Connected is not thread-safe; it may return true after an operation is aborted when the Socket is disconnected from another thread.
The value of the Connected property reflects the state of the connection as of the most recent operation.
If you need to determine the current state of the connection, make a nonblocking, zero-byte Send call. If the call returns successfully or throws a WAEWOULDBLOCK error code (10035), then the socket is still connected; //otherwise, the socket is no longer connected .
Can't you just use Select?
Use select on a connected socket. If the select returns with your socket as Ready but the subsequent Receive returns 0 bytes that means the client disconnected the connection. AFAIK, that is the fastest way to determine if the client disconnected.
I do not know C# so just ignore if my solution does not fit in C# (C# does provide select though) or if I had misunderstood the context.
Using the method SetSocketOption, you will be able to set KeepAlive that will let you know whenever a Socket gets disconnected
Socket _connectedSocket = this._sSocketEscucha.EndAccept(asyn);
_connectedSocket.SetSocketOption(SocketOptionLevel.Socket, SocketOptionName.KeepAlive, 1);
http://msdn.microsoft.com/en-us/library/1011kecd(v=VS.90).aspx
Hope it helps!
Ramiro Rinaldi
i had same problem , try this :
void client_handler(Socket client) // set 'KeepAlive' true
{
while (true)
{
try
{
if (client.Connected)
{
}
else
{ // client disconnected
break;
}
}
catch (Exception)
{
client.Poll(4000, SelectMode.SelectRead);// try to get state
}
}
}
This is in VB, but it seems to work well for me. It looks for a 0 byte return like the previous post.
Private Sub RecData(ByVal AR As IAsyncResult)
Dim Socket As Socket = AR.AsyncState
If Socket.Connected = False And Socket.Available = False Then
Debug.Print("Detected Disconnected Socket - " + Socket.RemoteEndPoint.ToString)
Exit Sub
End If
Dim BytesRead As Int32 = Socket.EndReceive(AR)
If BytesRead = 0 Then
Debug.Print("Detected Disconnected Socket - Bytes Read = 0 - " + Socket.RemoteEndPoint.ToString)
UpdateText("Client " + Socket.RemoteEndPoint.ToString + " has disconnected from Server.")
Socket.Close()
Exit Sub
End If
Dim msg As String = System.Text.ASCIIEncoding.ASCII.GetString(ByteData)
Erase ByteData
ReDim ByteData(1024)
ClientSocket.BeginReceive(ByteData, 0, ByteData.Length, SocketFlags.None, New AsyncCallback(AddressOf RecData), ClientSocket)
UpdateText(msg)
End Sub
You can also check the .IsConnected property of the socket if you were to poll.
I recently wrote a quick-and-dirty proof-of-concept proxy server in C# as part of an effort to get a Java web application to communicate with a legacy VB6 application residing on another server. It's ridiculously simple:
The proxy server and clients both use the same message format; in the code I use a ProxyMessage class to represent both requests from clients and responses generated by the server:
public class ProxyMessage
{
int Length; // message length (not including the length bytes themselves)
string Body; // an XML string containing a request/response
// writes this message instance in the proper network format to stream
// (helper for response messages)
WriteToStream(Stream stream) { ... }
}
The messages are as simple as could be: the length of the body + the message body.
I have a separate ProxyClient class that represents a connection to a client. It handles all the interaction between the proxy and a single client.
What I'm wondering is are they are design patterns or best practices for simplifying the boilerplate code associated with asynchronous socket programming? For example, you need to take some care to manage the read buffer so that you don't accidentally lose bytes, and you need to keep track of how far along you are in the processing of the current message. In my current code, I do all of this work in my callback function for TcpClient.BeginRead, and manage the state of the buffer and the current message processing state with the help of a few instance variables.
The code for my callback function that I'm passing to BeginRead is below, along with the relevant instance variables for context. The code seems to work fine "as-is", but I'm wondering if it can be refactored a bit to make it clearer (or maybe it already is?).
private enum BufferStates
{
GetMessageLength,
GetMessageBody
}
// The read buffer. Initially 4 bytes because we are initially
// waiting to receive the message length (a 32-bit int) from the client
// on first connecting. By constraining the buffer length to exactly 4 bytes,
// we make the buffer management a bit simpler, because
// we don't have to worry about cases where the buffer might contain
// the message length plus a few bytes of the message body.
// Additional bytes will simply be buffered by the OS until we request them.
byte[] _buffer = new byte[4];
// A count of how many bytes read so far in a particular BufferState.
int _totalBytesRead = 0;
// The state of the our buffer processing. Initially, we want
// to read in the message length, as it's the first thing
// a client will send
BufferStates _bufferState = BufferStates.GetMessageLength;
// ...ADDITIONAL CODE OMITTED FOR BREVITY...
// This is called every time we receive data from
// the client.
private void ReadCallback(IAsyncResult ar)
{
try
{
int bytesRead = _tcpClient.GetStream().EndRead(ar);
if (bytesRead == 0)
{
// No more data/socket was closed.
this.Dispose();
return;
}
// The state passed to BeginRead is used to hold a ProxyMessage
// instance that we use to build to up the message
// as it arrives.
ProxyMessage message = (ProxyMessage)ar.AsyncState;
if(message == null)
message = new ProxyMessage();
switch (_bufferState)
{
case BufferStates.GetMessageLength:
_totalBytesRead += bytesRead;
// if we have the message length (a 32-bit int)
// read it in from the buffer, grow the buffer
// to fit the incoming message, and change
// state so that the next read will start appending
// bytes to the message body
if (_totalBytesRead == 4)
{
int length = BitConverter.ToInt32(_buffer, 0);
message.Length = length;
_totalBytesRead = 0;
_buffer = new byte[message.Length];
_bufferState = BufferStates.GetMessageBody;
}
break;
case BufferStates.GetMessageBody:
string bodySegment = Encoding.ASCII.GetString(_buffer, _totalBytesRead, bytesRead);
_totalBytesRead += bytesRead;
message.Body += bodySegment;
if (_totalBytesRead >= message.Length)
{
// Got a complete message.
// Notify anyone interested.
// Pass a response ProxyMessage object to
// with the event so that receivers of OnReceiveMessage
// can send a response back to the client after processing
// the request.
ProxyMessage response = new ProxyMessage();
OnReceiveMessage(this, new ProxyMessageEventArgs(message, response));
// Send the response to the client
response.WriteToStream(_tcpClient.GetStream());
// Re-initialize our state so that we're
// ready to receive additional requests...
message = new ProxyMessage();
_totalBytesRead = 0;
_buffer = new byte[4]; //message length is 32-bit int (4 bytes)
_bufferState = BufferStates.GetMessageLength;
}
break;
}
// Wait for more data...
_tcpClient.GetStream().BeginRead(_buffer, 0, _buffer.Length, this.ReadCallback, message);
}
catch
{
// do nothing
}
}
So far, my only real thought is to extract the buffer-related stuff into a separate MessageBuffer class and simply have my read callback append new bytes to it as they arrive. The MessageBuffer would then worry about things like the current BufferState and fire an event when it received a complete message, which the ProxyClient could then propagate further up to the main proxy server code, where the request can be processed.
I've had to overcome similar problems. Here's my solution (modified to fit your own example).
We create a wrapper around Stream (a superclass of NetworkStream, which is a superclass of TcpClient or whatever). It monitors reads. When some data is read, it is buffered. When we receive a length indicator (4 bytes) we check if we have a full message (4 bytes + message body length). When we do, we raise a MessageReceived event with the message body, and remove the message from the buffer. This technique automatically handles fragmented messages and multiple-messages-per-packet situations.
public class MessageStream : IMessageStream, IDisposable
{
public MessageStream(Stream stream)
{
if(stream == null)
throw new ArgumentNullException("stream", "Stream must not be null");
if(!stream.CanWrite || !stream.CanRead)
throw new ArgumentException("Stream must be readable and writable", "stream");
this.Stream = stream;
this.readBuffer = new byte[512];
messageBuffer = new List<byte>();
stream.BeginRead(readBuffer, 0, readBuffer.Length, new AsyncCallback(ReadCallback), null);
}
// These belong to the ReadCallback thread only.
private byte[] readBuffer;
private List<byte> messageBuffer;
private void ReadCallback(IAsyncResult result)
{
int bytesRead = Stream.EndRead(result);
messageBuffer.AddRange(readBuffer.Take(bytesRead));
if(messageBuffer.Count >= 4)
{
int length = BitConverter.ToInt32(messageBuffer.Take(4).ToArray(), 0); // 4 bytes per int32
// Keep buffering until we get a full message.
if(messageBuffer.Count >= length + 4)
{
messageBuffer.Skip(4);
OnMessageReceived(new MessageEventArgs(messageBuffer.Take(length)));
messageBuffer.Skip(length);
}
}
// FIXME below is kinda hacky (I don't know the proper way of doing things...)
// Don't bother reading again. We don't have stream access.
if(disposed)
return;
try
{
Stream.BeginRead(readBuffer, 0, readBuffer.Length, new AsyncCallback(ReadCallback), null);
}
catch(ObjectDisposedException)
{
// DO NOTHING
// Ends read loop.
}
}
public Stream Stream
{
get;
private set;
}
public event EventHandler<MessageEventArgs> MessageReceived;
protected virtual void OnMessageReceived(MessageEventArgs e)
{
var messageReceived = MessageReceived;
if(messageReceived != null)
messageReceived(this, e);
}
public virtual void SendMessage(Message message)
{
// Have fun ...
}
// Dispose stuff here
}
I think the design you've used is fine that's roughly how I would and have done the same sort of thing. I don't think you'd gain much by refactoring into additional classes/structs and from what I've seen you'd actually make the solution more complex by doing so.
The only comment I'd have is as to whether the two reads where the first is always the messgae length and the second always being the body is robust enough. I'm always wary of approaches like that as if they somehow get out of sync due to an unforseen circumstance (such as the other end sending the wrong length) it's very difficult to recover. Instead I'd do a single read with a big buffer so that I always get all the available data from the network and then inspect the buffer to extract out complete messages. That way if things do go wrong the current buffer can just be thrown away to get things back to a clean state and only the current messages are lost rather than stopping the whole service.
Actually at the moment you would have a problem if you message body was big and arrived in two seperate receives and the next message in line sent it's length at the same time as the second half of the previous body. If that happened your message length would end up appended to the body of the previous message and you'd been in the situation as desecribed in the previous paragraph.
You can use yield return to automate the generation of a state machine for asynchronous callbacks. Jeffrey Richter promotes this technique through his AsyncEnumerator class, and I've played around with the idea here.
There's nothing wrong with the way you've done it. For me, though, I like to separate the receiving of the data from the processing of it, which is what you seem to be thinking with your proposed MessageBuffer class. I have discussed that in detail here.