[3119] in Kerberos-V5-bugs

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

krb5-kdc/758: kadmind changepw bug: core dumps

daemon@ATHENA.MIT.EDU (cthallen@binghamton.edu)
Tue Sep 28 17:14:06 1999

Resent-From: gnats@rt-11.MIT.EDU (GNATS Management)
Resent-To: krb5-unassigned@RT-11.MIT.EDU
Resent-Reply-To: krb5-bugs@MIT.EDU, admins@mail.binghamton.edu
Message-Id: <199909282113.RAA13977@jake2.cc.binghamton.edu.binghamton.edu>
Date: Tue, 28 Sep 1999 17:13:50 -0400
From: cthallen@binghamton.edu
Reply-To: admins@mail.binghamton.edu
To: krb5-bugs@MIT.EDU
Cc: mcronk@binghamton.edu, cthallen@binghamton.edu


>Number:         758
>Category:       krb5-kdc
>Synopsis:       kadmind core dumps after several password changes by admin user
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    krb5-unassigned
>State:          open
>Class:          sw-bug
>Submitter-Id:   unknown
>Arrival-Date:   Tue Sep 28 17:14:01 EDT 1999
>Last-Modified:
>Originator:     Chris Hallenbeck
>Organization:
Binghamton University
	
>Release:        krb5-current-19990927
>Environment:
	
	
System: SunOS jake2.cc.binghamton.edu 5.6 Generic_105181-05 sun4m sparc SUNW,SPARCstation-5
Architecture: sun4

>Description:
	We run krb5kdc, kadmind, and krb524d on Solaris 2.6 machines.  The "production"
version is "krb5-current-19990712", and was compiled using gcc 2.8.1 on a Sun SPARC5.  THIS
release (09/27/99) was compiled on a similar machine running Solaris 2.6 using SunProC. 

        In BOTH versions, kadmind coredumps after several password changes.  I have only been
able to verify this problem in situations where the password changes are being done as a part
of a script run by an admin user.  
ex:
  kadmin -c $KRB5CCNAME -q "cpw -pw $password $principal"

        It only seems to take about a dozen such requests before the daemon dies.  I have not yet
had a chance to test to see how long it takes for a "regular" user running 'kpasswd' to cause this
type of coredump -- or if it happens at all!

>How-To-Repeat:
	In "helpdesk" account's '.profile' (under ksh):
-----
KRB5CCNAME=/tmp/krb5cc_helpdesk
export PATH KRB5CCNAME
trap kdestroy 0 1 2 3 5 15
kinit -S kadmin/admin -k -t acct.keytab helpdesk/admin
----------END .profile-----------

In 'kpass script' :
-----
#!/usr/local/bin/expect --

[snip]
spawn kadmin -c $KRB5CCNAME -q "cpw -pw $password $principal"
expect {
	timeout {
[snip]
------------END 'kpass' script--------


     Our Helpdesk often runs this program up to 100 times a day (especially at the beginning of
the semester).  Since usage has dropped considerably -- no more than 15 times/day -- we have
had FAR fewer coredumps of kadmind.


>Fix:
	No known work-around short of restarting the daemon when you detect that it has died. :-P

We'll be more than glad to send you a few of the cores if you actually want them. ;-)


Separate, but similar issue:  in <path to src>/src/lib/krb5/os/changepw.c  there was a problem with 
a #ifdef on or about line 86 of the 09/27/99 version (ifdef KRB5_DNS_LOOKUP).  If that is NOT
defined you get an "undefined symbol: i" error, because the declartion "int i" is inside that #ifdef. 

    Again, that COMPILATION error was received under Solaris 2.6 using SunProC (/opt/SUNWspro/bin/cc).


Best regards,

Chris Hallenbeck
>Audit-Trail:
>Unformatted:

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