From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.1) Gecko/20031114 Description of problem: I use saslauthd for SMTP authentication on a mail server. On a regular basis, saslauthd dies and leaves a message in /var/log/messages: Jun 2 11:43:37 maas saslauthd: saslauthd startup succeeded Jun 2 12:42:58 maas saslauthd[19605]: ipc_loop : socket accept failure Jun 2 12:43:01 maas saslauthd[19605]: ipc_loop : accept: Interrupted system call googleing for this error message geve me this link from the cyrus-sasl mailing list: http://asg.web.cmu.edu/archive/message.php?mailbox=archive.cyrus-sasl&msg=5310 one of the replies states that it is a bug, fixes in version 2.1.17, but fedora core 1 comes with 2.1.15, and the changelog of the rpm has nothing about a backported patch Version-Release number of selected component (if applicable): cyrus-sasl-2.1.15-6 How reproducible: Sometimes Steps to Reproduce: 1. service saslauthd restart 2. wait for some time, usually not to long before this happens and the authentication becomes unavailable 3. Actual Results: smtp authentication no longer possible so users can not send e-mail unless authenticated in another way (or host listed in allowed list for relaying) Additional info:
I am getting the same failure with RHEL AS 3. cyrus-sasl-2.1.15-8. Using MECH=kerberos5 and running testsaslauthd always produces Jul 21 13:32:32 localhost saslauthd[11537]: ipc_loop : socket accept failure Jul 21 13:32:32 localhost saslauthd[11537]: ipc_loop : accept: Interrupted system call Eventually producing a Segmentation Fault
Similar symptoms on FC 2 (cyrus-sasl-2.1.18-2), except there are no ipc_loop log entries, which were eradicated I believe in the upstream 2.1.17 source. I can always reproduce the problem by using the Mail.app that ships with 10.3.5 (earlier versions, afaict, don't have the problem). In any case, it's a client-initiated problem and therefore a potential denial-of-service scenario. I'm using MECH=pam and we're using NIS for most of the password map. It looks like the problem client will take out one (and only one) saslauthd thread/process each time it tries to authenticate. If it takes out a child thread (parent pid > 1), then other users can still authenticate. If it takes out the master thread (parent pid == 1), then no one can authenticate.
You can disregard my comment dated 2004-08-11 18:23; the problems were due to the "more than 8 groups" bug outlined in ticket #125653: https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=125653 Once I edited /etc/nsswitch.conf to read "group: files nis" then all was well again. Sorry for the misdiagnosis!
Fedora Core 2 is now maintained by the Fedora Legacy project for security updates only. If this problem is a security issue, please reopen and reassign to the Fedora Legacy product. If it is not a security issue and hasn't been resolved in the current FC3 updates or in the FC4 test release, reopen and change the version to match.
Since switching to FC3 I haven't seen this problem again, I guess the bug is closed then (unless of course it is still present in RHEL as reported in comment #1
Marking resolved:currentrelease. If this is a problem in RHEL, reopen or file a new bug.