[jcifs] Filter Bug: Transport Port 0 Incorrectly Interpreted

[Michael B Allen]

On Mon, 26 Sep 2005 14:23:41 +0200
Oliver Schoett <os at sdm.de> wrote:

>  From reading the code (guided by the line numbers in the exception 
> stack above), it appears to me that the SmbTransport is created in 
> SmbSession.java, line 72 (just before the call to connect in line 74) 
> with destination port 0, and the local port is also 0 by default.  Hence 
> I think that the Socket constructor is called with both port numbers 0, 
> and I find it justified for the IBM JDK 1.4.1 to throw an exception in 
> this case, as these port numbers do not appear to be good for connecting 
> from or to.
> 
> How is this code supposed to work? Should not the destination port 
> number be 139?

Port 445 but then fallback to 139 but yes, there is a bug here. Apparently
the port is normally supplied but in the case of SmbSession.logon and one
other place 0 can be specified which is incorrectly interpreted by the new
Transport code. Note that I think the faulty code will still work if you
can connect to port 139 because the exception will cause the transport
to fallback to that.

In any case, I think the fix is to add two lines to SmbTransport.java:232:

  222   private void negotiate( int port, ServerMessageBlock resp ) throws IOException {
  223     /* We cannot use Transport.sendrecv() yet because
  224      * the Transport thread is not setup until doConnect()
  225      * returns and we want to supress all communication
  226      * until we have properly negotiated.
  227      */
  228     synchronized (sbuf) {
  229       if (port == 139) {
  230         ssn139();
  231       } else {
  232         if (port == 0)
  233           port = DEFAULT_PORT; // 445
  234         if (localAddr == null) {
  235           socket = new Socket( address.getHostAddress(), port );

Nice catch and diagnosis. This will be added to The List.

Thanks,
Mike