TCP Socket write error

Question

What is wrong with this simple Java TCP server/client example?

First, start the server:

import java.io.OutputStream;
import java.net.ServerSocket;
import java.net.Socket;

public class Server {

    public static void main(String[] args) throws Throwable {
        ServerSocket ss = new ServerSocket(2345);
        Socket s = ss.accept();
        OutputStream os = s.getOutputStream();
        Thread.sleep(5000);
        for (int i = 0; i < 3; i++) {
            os.write("A".getBytes());
            os.flush();
            System.out.println("Written in cycle " + i);
        }
        os.close();
        s.close();

    }

}

Start the client and watch the server after that:

import java.net.Socket;

public class Client {

    public static void main(String[] args) throws Throwable{
        Socket s=new Socket("localhost",2345);
        s.close();
        System.out.println("Closed");
    }

}

The client socket is closed immediately. However, the socket write operation on the server fails in the second loop, i.e. the first write does not throw an exception.

This is the server's execution output:

Written in cycle 0
Exception in thread "main" java.net.SocketException: Software caused connection abort: socket write error
    at java.net.SocketOutputStream.socketWrite0(Native Method)
    at java.net.SocketOutputStream.socketWrite(SocketOutputStream.java:92)
    at java.net.SocketOutputStream.write(SocketOutputStream.java:124)
    at Server.main(Server.java:16)

Doesn't the client socket have to do something to receive the outputstream from the server? Maybe I'm just missing something so please correct me if I'm wrong, but why do you close the client socket immediately after creating it? — love_me_some_linux
– love_me_some_linux, Commented Aug 10, 2011 at 14:32

glglgl · Accepted Answer · 2011-08-10 14:38:43Z

2

Your client connects and immediately closes. This is detected by the server which refuses to write into the "nowhere".

answered Aug 10, 2011 at 14:38

glglgl

91.5k13 gold badges157 silver badges230 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

user416472 · Accepted Answer · 2011-08-11 05:41:48Z

2

The following link matches the question 100% and explains what the problem is and how to solve it:

http://download.oracle.com/javase/1.5.0/docs/guide/net/articles/connection_release.html

So, to be more specific related to the above code, setting LINGER to 0 will solve the problem:

import java.net.Socket;

public class Client {

    public static void main(String[] args) throws Throwable {
        Socket s = new Socket("localhost", 2345);
        s.setSoLinger(true, 0);
        s.close();
    }

 }

answered Aug 11, 2011 at 5:41

user416472

1041 silver badge9 bronze badges

Comments

Peter Lawrey · Accepted Answer · 2011-08-10 14:26:00Z

1

When you write to a socket it has to pass that data to the remote system. If it were to wait for the remote system to wait to confirm every piece of data it would be many times slower. In any case this is not how TCP works.

If you need the remote socket to confirm it has received your data you need to add a protocol of your own to ensure every piece of data you send has been received.

answered Aug 10, 2011 at 14:26

Peter Lawrey

535k83 gold badges770 silver badges1.2k bronze badges

4 Comments

user416472 Over a year ago

Why the server detects that the client socket is closed on the second write attempt? Is this Windows specific (I'm using Windows 7 64 bit)? Imagine a push server that must send a single control byte ("A") and detect dead sockets if write wasn't successful. Note that if we split the payload (at last 2 bytes) the closed socket is detected immediately:os.write("A".getBytes()); os.flush();os.write("B".getBytes()); os.flush();

user416472 Over a year ago

Ok, I agree. We can't detect that the socket was closed until TCP stack receives the last client [RST, ACK] packet. Application layer logic will detect these use cases. Thanks!

user416472 Over a year ago

` TCP 40 altav-tunnel > dbm [FIN, ACK] Seq=1 Ack=1 Win=8192 Len=0 5 0.000000 192.168.2.2 192.168.2.2 TCP 40 dbm > altav-tunnel [ACK] Seq=1 Ack=2 Win=8192 Len=0 6 5.007600 192.168.2.2 192.168.2.2 TCP 41 dbm > altav-tunnel [PSH, ACK] Seq=1 Ack=2 Win=8192 Len=1 7 5.007600 192.168.2.2 192.168.2.2 TCP 40 altav-tunnel > dbm [RST, ACK] Seq=2 Ack=2 Win=0 Len=0` @Peter Lawrey , why the server sends PSH (push) packet after FIN packet was received and acknowledged already?

Peter Lawrey Over a year ago

Its normal to buffer ACK packets to minimise overhead. There are a number of strategies used to improve efficiency, however these tend to delay detection of a lost connection.

love_me_some_linux · Accepted Answer · 2011-08-10 15:05:44Z

0

Copied from comments as it answers your question.

Doesn't the client socket have to do something to receive the outputstream from the server? Maybe I'm just missing something so please correct me if I'm wrong, but why do you close the client socket immediately after creating it? This is the problem.

answered Aug 10, 2011 at 15:05

love_me_some_linux

2,7711 gold badge16 silver badges7 bronze badges

Collectives™ on Stack Overflow

TCP Socket write error

4 Answers 4

Comments

Comments

4 Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

4 Answers 4

Comments

Comments

4 Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related