I came in to my office on Sunday morning, May 1, to find that my PC running
SLES 9 was at the login screen (gdm). This was unusual, as I leave it
logged in 24x7. I logged in, only to be again presented with the login screen
after the screen went blank for a few seconds. This happens no matter which
login I use (root, trb, or games), or which window manager.

I assumed the machine had rebooted during the night, but that wasn't indicated
in /var/log/messages. Nor is there any indication of a hardware problem.
I tested the video card in my Son's computer, and it works perfectly at
1600x1200 resolution. Additionally, I can run "startx" from run level 3, and
everything works perfectly. So, this isn't a huge problem, but I would
like to figure out what caused it.

I'm running SLES 9, which includes KDE 3.2 - which is my default window
manager. The login/logout problem occurs even if I select any other available
window manager at the login screen.

Here are the messages in /var/log/messages that I suspect may give a clue
as to what happened. All the rest of the messages before and after this
time are from mailman - similar to the 2 at 4:20 in the middle of the list
immediately below:


May 1 04:15:16 gomer su: (to cyrus) root on none
May 1 04:15:16 gomer su: pam_unix2: session started for user cyrus, service su
May 1 04:15:19 gomer ctl_mboxlist[2820]: DBERROR: reading /var/lib/imap/db/skipstamp, assuming the worst: No such file or directory
May 1 04:15:21 gomer ctl_mboxlist[2820]: skiplist: recovered /var/lib/imap/mailboxes.db (0 records, 144 bytes) in 1 second
May 1 04:15:21 gomer su: pam_unix2: session finished for user cyrus, service su
May 1 04:20:00 gomer /USR/SBIN/CRON[2854]: (mailman) CMD (/usr/bin/python -S /usr/lib/mailman/cron/gate_news)
May 1 04:20:00 gomer /USR/SBIN/CRON[2856]: (mailman) CMD (/usr/bin/python -S /usr/lib/mailman/cron/gate_news)
May 1 04:20:52 gomer kdm[3593]: Unknown session exit code 0 (sig 1) from manager process
May 1 04:20:52 gomer kdm[3593]: Rescanning all config files
May 1 04:20:53 gomer gconfd (trb-4021): Received signal 1, shutting down cleanly
May 1 04:20:54 gomer gconfd (trb-4021): Exiting
May 1 04:20:56 gomer su: pam_unix2: session finished for user root, service su
May 1 04:21:03 gomer su: (to nobody) root on none
May 1 04:21:03 gomer su: pam_unix2: session started for user nobody, service su
May 1 04:21:38 gomer su: pam_unix2: session finished for user nobody, service su


Starting at run level 3, I get my KDE environment with all my applications
started in the proper place (i.e. the place saved by selecting "Save
Session" on the main KDE menu) by running "startx". The only issue is that
the "KDE Panel" won't start until I click on the "OK" button of a window
entitled "Connect Host - KDE Panel". It wants to know the "Connection Type".
ssh is selected by default, so a click on "OK" selects that. After this, the
KDE Panel starts and another window pops up entitled "OpenSSH Authentication".
Below the title and in the window, the word "Password:" is shown above
8 dark blue boxes. After clicking on "Okay" in this window, it goes away,
to be swiftly followed by another identical window. This happens one more
time, and then I'm able to type in my xterm windows. Until I click on
"Okay" in the authentication windows, not text is accepted in the shell
windows.

Here are the messages in /var/log/messages when running startx:


May 8 14:39:14 gomer kernel: mtrr: 0x88000000,0x8000000 overlaps existing 0x88000000,0x1000000
May 8 14:40:33 gomer sshd[5121]: error: PAM: Authentication failure
May 8 14:40:35 gomer last message repeated 2 times


When I go to runlevel 5 from run level 3, here are the messages I get:


May 8 13:31:57 gomer init: Switching to runlevel: 5
May 8 13:32:02 gomer kernel: mtrr: 0x88000000,0x8000000 overlaps existing 0x88000000,0x1000000


Going to run level 5 will bring up the login screen (gdm). Attempting to login
generates these messages:


May 8 13:34:39 gomer kdm: :0[18030]: pam_unix2: session started for user trb, service xdm
May 8 13:34:42 gomer kdm: :0[18030]: pam_unix2: session finished for user trb, service xdm
May 8 13:34:47 gomer kernel: mtrr: 0x88000000,0x8000000 overlaps existing 0x88000000,0x1000000


The only Warning messages (there are no Error messages) in
/var/log/XFree86.0.log are:


(WW) RADEON: No matching Device section for instance (BusID PCI:1:0:1) found
(WW) RADEON(0): config file hsync range 28-72kHz not within DDC hsync ranges.
(WW) RADEON(0): config file vrefresh range 50-60Hz not within DDC vrefresh ranges.
(WW) RADEON(0): Failed to set up write-combining range (0x88000000,0x8000000)
(WW) RADEON(0): Direct rendering not yet supported on Radeon 9500/9700 and newer cards



I've upgraded to SLES 9 SP1, and that didn't help. I also removed
~trb/.ssh, ~trb/.Xauthority, and ~trb/.ICEauthority to see if that helped.
It didn't.


Why am I getting logged out 3 seconds after logging in?

My first thought, even before I read the rest of your post, was that it was an authentication problem. The rest of your post seemed to confirm it.

If you are using an authentication server, might I suggest you check the user account on it? If not, maybe something went wrong with the pam module and you could try to reinstall it.

I installed the latest pam rpms, and that didn't help. Same error. Here is what I installed:

pam-0.77-227.i586.rpm
pam-modules-9.2-2.1.i586.rpm

I agree that it looks like an authentication problem, but I'll be darned if I can figure it out.

Any other suggestions?