[53918] in Hotline Meeting

home help back first fref pref prev next nref lref last post

Re: Case 201903: This case has recently been updated

daemon@ATHENA.MIT.EDU (Tom Cavin)
Tue Feb 20 19:14:09 2001

From: Tom Cavin <tec@ai.mit.edu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-Id: <14995.2121.296741.3243@vivaldi.mit.edu>
Date: Tue, 20 Feb 2001 19:14:01 -0500 (EST)
To: Garry Zacheiss <zacheiss@MIT.EDU>
Cc: tec@ai.mit.edu, ops@MIT.EDU, hotline@MIT.EDU, larugsi@MIT.EDU
In-Reply-To: <200102202141.QAA10795@riff-raff.mit.edu>


Hi Garry,

I think I solved the problem.  For some unknown reason, the
/usr/openwin/lib/libXext.so file didn't have group read permissions.

0000-Admin(0000) writes:
 > weepecket.mit.edu# ls -l /usr/openwin/lib/lib{SM,ICE,X11,Xext}.so.*
 > -rwxr-xr-x   1 bin   bin  110012 Aug 15  1998 /usr/openwin/lib/libICE.so.6
 > -rwxr-xr-x   1 root  bin   48464 Aug 25  1998 /usr/openwin/lib/libSM.so.6
 > -rwxr-xr-x   1 root  bin  758352 Sep  4  1998 /usr/openwin/lib/libX11.so.4
 > -rwxrwx--x   1 root  bin  101652 Aug 25  1998 /usr/openwin/lib/libXext.so.0

I added group read permissions and the problem went away.

I haven't a clue as to how the permissions got changed, and I didn't
check inode dates so I don't know when the change happened.

The full story goes as follows:

    Symptoms:  the system seems to start up fine and waits for login.
    However, when someone does login, all sorts of things crash with
    the message of permission denied for libXext.so.  The console
    screen is left with a basic X server and no applications.  (You
    have a blank screen with the x-weave background, and a mouse
    controlled sprite, but no x-clients and no way to get them.)
    If you login remotely, and kill Xsun, the system resets to the
    login screen.

    The Cause: init starts /etc/athena/dm as root, and can therefore
    bring up the necessary programs.  However, the moment someone does
    a login, root permissions are dropped and subsequent X clients
    can't be loaded properly.  The Xsun process is still up, but with
    no clients, the user is logged out, and the screen is left
    unusable.

    Hints: the /var/athena/console.log file lists the failure of
    several programs including /etc/athena/console and one or two
    other X clients.  All failures are libXext.so permission denied.
    If I had caught this file sooner, it might have been more
    obvious.

    My Debug Process: I went to the lab and logged in on a nearby
    system then did an "ssh -l root weepecket".  After trying a few
    things such as killing the running Xsun process and attempting to
    login a few times on weepecket's console, I decided to try running
    the process by hand.  I did:

	$ mv /etc/inittab /etc/inittab.save
	$ cp -p /etc/inittab.save /etc/inittab

    and edited /etc/inittab and removed runlevels 3 and 4:

      co:2:respawn:/etc/athena/dm /etc/athena/login/config ttyp0 console 

    then did "telinit q".  This stopped the automatic restart of dm
    and let me look at the failure conditions.  I then entered the
    full dm command for weepecket on the remote window, and went to
    the console to try to login again.  Same behavior, but this time I
    was able to follow it and get the console messages.  The
    suggestion from the messages was to try to login without my user
    customizations.  I tried that and it didn't make any difference.
    Eventually, I read the rest of the console messages and checked
    the permissions on libXext.so.0.  I did a "chmod g+r
    libXext.so.0", then tried once more by hand and it worked.  To
    clean up, I restored the inittab file and did "telinit q".

I still don't know how the bit got flipped, but if it happens again,
I'll have a better shot at fixing the problem quickly, and I'll start
looking for a cause.

If you still want to look at the system, let me know and I'll set you
up with a login.  As far as the users are concerned, this case is
closed.

Thanks,

	--Tom

Garry Zacheiss writes:
 > 	Hi Tom,
 > 
 > 	We can try to take a look at this machine.  You say you can log
 > in remotely to it, but not on the console.  Is this true for both xlogin
 > and a console login (control-p from xlogin)?  Could you quote us the
 > exact error messages you're seeing?
 > 
 >         Would it be possible to get an account on the machine (either as
 > myself or in root's klogin) to look at it further?
 > 
 > Garry

-- 
Tom Cavin                                Phone:  (617) 258 - 7806
WCCF Computer Operations Manager         Email:  tec@ai.mit.edu

home help back first fref pref prev next nref lref last post