Mailing-List: contact cygwin-developers-help AT sourceware DOT cygnus DOT com; run by ezmlm List-Subscribe: List-Archive: List-Post: List-Help: , Sender: cygwin-developers-owner AT sources DOT redhat DOT com Delivered-To: mailing list cygwin-developers AT sources DOT redhat DOT com Date: 13 Nov 2001 12:45:25 -0500 Message-ID: <20011113174525.4108.qmail@lizard.curl.com> From: Jonathan Kamens To: cygwin-developers AT cygwin DOT com Subject: "Address already in use" error from ssh with current Cygwin installation I just upgraded Cygwin on one of our build servers so that it has all of the most recent Cygwin packages, including version 1.3.5 of the DLL, installed. Shortly afterwards, I got this error from ssh: ssh: connect to address XXX.XXX.XXX.XXX port 22: Address already in use I managed to duplicate it on an SMP machine running Windows NT with this command: while ssh HOST date; do true; done HOST is a host on our internal network, i.e., the network to it is very fast. I was unable to duplicate the problem on a uniprocessor machine. When I run "netstat" immediately after the failure, I see a whole bunch of connections to sshd on the remote host in TIME_WAIT status. When I peruse fhandler_socket.cc, it appears to me that the code is randomly assigning a local port number to a socket during connect. Is that correct? If so, then that presumably explains the problem -- sometimes the random port number conflicts with one that's already in use (but this doesn't explain why I couldn't duplicate the problem on a uniprocessor machine). If someone here can confirm that my understanding of the code in fhandler_socket.cc is correct and that my theory for what is causing the bug is thus also correct, then I will begin working on figuring out how to fix it. Although I don't have any ideas off the top of my head, so I'm open to suggestions :-). Thanks, jik