From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 3035 invoked from network); 27 Sep 2001 11:31:35 -0000 Received: from sunsite.dk (130.225.51.30) by ns1.primenet.com.au with SMTP; 27 Sep 2001 11:31:35 -0000 Received: (qmail 11132 invoked by alias); 27 Sep 2001 11:31:28 -0000 Mailing-List: contact zsh-workers-help@sunsite.dk; run by ezmlm Precedence: bulk X-No-Archive: yes X-Seq: 15880 Received: (qmail 11115 invoked from network); 27 Sep 2001 11:31:27 -0000 To: zsh-workers@sunsite.dk (Zsh hackers list) Subject: More tcp problems Date: Thu, 27 Sep 2001 12:30:49 +0100 Message-ID: <10779.1001590249@csr.com> From: Peter Stephenson There are a few more problems with the tcp module and its interaction with zftp. I was hoping to fix the first one before releasing a new development version, but it seems it goes beyond that. - Closing a zftp connection doesn't set the session pointer to NULL. This results (in my case) in a segmentation violation when opening a second connection. This is because the tcp_close() frees the session, so it can't be re-used or even tested again. (Simply setting the pointer to NULL caused other problems I didn't understand, maybe related to the rest of this list.) - Don't know if this is related, but I get BUG: attempt to free storage at invalid address when opening zftp connections, in particular the first (since I don't get as far as a second). - In general, it seems a little bit difficult to tell whether tcp_close() has actually freed the session or not. And if it hasn't, because it encountered an error with close(), it's hard to see how the session should be freed. I think another call to tcp_close() would do it --- but it's hard to know when you need that. If you do it when the session has already been freed, you're in big trouble. - There's a similiar problem with tests for (sess->fd == -1) in zftp. If they're true, the session is never freed; opening a new one will simply assign a different TCP session to the same pointer, so that the memory leaks. - With a failed zfopen, I now get zfopen:42: connection close failed: bad file number (plus a segmentation violation which I guess is something to do with the previous stuff). I don't get that message with 4.0.1. The function tests at that point to see if $ZFTP_HOST is set, and if it is, attempts to close the file. I *think* that all that's changed is the zfclose was silent before and isn't now, because of tcp_close(). This may be a knock-on effect of the things above, though, i.e. it goes away if the session pointers are handled properly. I hope it's possible to make tcp only return a session pointer if the fd is valid, and to free the session unconditionally on an attempt to close it. -- Peter Stephenson Software Engineer CSR Ltd., Science Park, Milton Road, Cambridge, CB4 0WH, UK Tel: +44 (0)1223 392070 ********************************************************************** The information transmitted is intended only for the person or entity to which it is addressed and may contain confidential and/or privileged material. Any review, retransmission, dissemination or other use of, or taking of any action in reliance upon, this information by persons or entities other than the intended recipient is prohibited. If you received this in error, please contact the sender and delete the material from any computer. **********************************************************************