[SGE-discuss] 6.2u5: PAM modules writing to stderr causes strange qrsh failure

William Hay w.hay at ucl.ac.uk
Fri Oct 14 06:29:47 BST 2011

On 13 October 2011 20:25, Brendan Moloney <moloney at ohsu.edu> wrote:
> From: wish.dumain at gmail.com [wish.dumain at gmail.com] On Behalf Of William Hay [w.hay at ucl.ac.uk]
>> Having a similar problem with the sshd on SL5.  I found that a later
>> version of openssh works but  pam_sge-qrsh-setup.so didn't work
>> properly with the later verions I built unless I disabled privilege
>> separation.
> I am on Debian Squeeze with OpenSSH 5.5. Are you sure that the issue wasn't a PAM module writing out
> to stderr? Compare your auth.log from a successful qrsh login and a failed qrsh login.
Quite possibly it is but switching to a later version of sshd makes it
work.  Also there is apparently  a bug in the version of OpenSSL
provided by Scientific Linux that is worked around  by later versions
of openssh that can cause bad stuff to happen.

> I initially triggered this by setting up public key auth for ssh which caused one of my PAM modules
> (pam_mount) to complain. I have since found that the same thing happens if other PAM modules
> produce error messages (for non fatal errors) in auth.log.  Even enabling debugging in pam_mount
> will cause it.

More information about the SGE-discuss mailing list