openssh bugs - Jun 2004 - [Bug 881] PAM: authentication thread exited unexpectedly

http://bugzilla.mindrot.org/show_bug.cgi?id=881

           Summary: PAM: authentication thread exited unexpectedly
           Product: Portable OpenSSH
           Version: 3.8p1
          Platform: UltraSparc
        OS/Version: Solaris
            Status: NEW
          Severity: normal
          Priority: P2
         Component: PAM support
        AssignedTo: openssh-bugs at mindrot.org
        ReportedBy: a at wynes.com
                CC: a at wynes.com


Relevant components of our environment:

	Solaris 8
	Sun Resource Manager 1.3 (SRM) with all published patches.
	OpenSSH 3.8.1p1

We're having an issue where SSH logins using PAM authentication are
failing for certain accounts (those without SRM lnodes).  The client 
receives a "Connection Closed" message and sshd logs the message
"fatal: PAM:
authentication thread exited unexpectedly".

With SRM, the kernel uses lnodes (Limit nodes) associated with uids to 
track usage and allocation for RAM, CPU and logins.  

An lnode can be created manually after the userid has been created or
automatically the first time a user logs in.  This is done by a PAM
module - pam_srm.

	http://docs.sun.com/db/doc/816-7750/6mdo1u7fr?q=pam_srm&a=view

If the account logging in does not have an lnode, the module runs a script
/etc/srm/nolnode to generate one. SSH is exiting as a result of calling this 
PAM module - if I comment out that PAM module the user can log in. lnodes are
only created once, unless deleted by root.

Observations:

o If the user has an existing lnode, they can log in fine.
o This functionality worked with OpenSSH 3.7.1p1.
o If UsePassword or RSAAuthentication are used instead of PAM the users can log
in fine.
o We use PrivilegeSeparation. It happens if that is turned off/on.
o The /etc/srm/nolnode script is never running - I've made changes to
verify.
o The logins succeed if pam_srm is commented out in pam.conf.  That's our
work-around but it means those logins are not controlled by SRM.

This looks like a bug with SRM but I don't understand why it wasn't an
issue with OpenSSH 3.7.1p1. If it's obviously a problem with pam_srm
can you tell me what it would be doing to cause sshd to log the "exited
unexpectedly" message.  A SEGFAULT?

Thanks.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=881





------- Additional Comments From a at wynes.com  2004-06-18 06:19 -------
Created an attachment (id=649)
 --> (http://bugzilla.mindrot.org/attachment.cgi?id=649&action=view)
truss showing pam_srm open and error

This truss is from an sshd -ddd.  It shows the point in execution where pam_srm
is called and continues until the process exits.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=881

a at wynes.com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|3.8p1                       |3.8.1p1





------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=881





------- Additional Comments From dtucker at zip.com.au  2004-06-18 11:54 -------
The trace looks like a segfault.  Did it leave a core file (eg in /)?

The message is because sshd attempts to trap the case where the PAM
"thread"
bombs out.  (We were seing this when, eg, a PAM module crashed, and it would
hang sshd).



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=881





------- Additional Comments From a at wynes.com  2004-06-18 13:36 -------
No, no core files.  



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=881





------- Additional Comments From dtucker at zip.com.au  2004-06-18 13:43 -------
You can use coreadm to put them some place handy (I use /var/core/).  I think
the command is "coreadm -g /var/core/core.%f.%p", but check the man
page and
adjust to local conditions.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

http://bugzilla.mindrot.org/show_bug.cgi?id=881





------- Additional Comments From a at wynes.com  2004-06-19 07:05 -------
Created an attachment (id=650)
 --> (http://bugzilla.mindrot.org/attachment.cgi?id=650&action=view)
SSHD Core File

The default has global setid core dumps disabled.  I enabled it, here's the
gzipped core.



------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.