From mboxrd@z Thu Jan 1 00:00:00 1970 From: Martin Pool Subject: Re: FIN_WAIT1 / TCP_CORK / 2.2 -- reproducible bug and test case Date: Wed, 18 Sep 2002 12:30:14 +1000 Sender: netdev-bounce@oss.sgi.com Message-ID: <20020918023012.GC2285@samba.org> References: <20020918020346.GA2285@samba.org> <20020917.185801.76605813.davem@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: ak@muc.de, kuznet@ms2.inr.ac.ru, netdev@oss.sgi.com, Alan.Cox@linux.org Return-path: To: "David S. Miller" Content-Disposition: inline In-Reply-To: <20020917.185801.76605813.davem@redhat.com> Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org For what it's worth, I can also reproduce the bug on 2.2.22rc3. It's on a non-SMP machine with a very standard kernel config, no firewalling, Debian 2.2. I'm testing it under VMWare, but the same symptoms were observed by other people on real hardware with Red Hat 6.2. It was originally noticed in distcc, and this test case is basically just a stripped-down version of the same code. If the program uncorks the socket before closing it, then it cleans up properly. However, if distcc is killed, then it can't do that and the problem occurs. (The most recent version makes an effort to always uncork, but obviously for e.g. SIGKILL it can't do much other than avoid corks entirely.) Some people observed a couple of hundred (?) sockets getting into this state. -- Martin