*** Please type your report below this line ***

!!! Attention: A german writes in english !!! ;-)

We are using Debian as a NIS-Server since 3 years,
with up to 100 clients on serveral locations.
With Etch we had no problems all the years.

Now we have upgraded the system to Lenny and the problem is as follow:

If a IBM-unix-workstation with aix5.3 has a networkproblem during
boot time, the system waits for the nis-service (server).
When the network is ok again, the ibm tells that the nis-server is ok, but
in
the same second the yp-server on Lenny crashs with the message:

ypserv[2358]: segfault at 1c4ccca4 ip b7e88468 sp bf9c30e4 error 6 in
libc-2.7.so[b7e17000+155000]

If you now restart the yp-server, it crashes again at once.

To bring the yp-server in normal operation, you now must shutdown all
IBM-workstations, start the yp-server and after
that you can start the IBM's again and hope that you have no network
problems during they start.

For first aid we went back from Lenny to Etch to have a secure system.
But, how we know the Etch system  get out of security-support in the next
days.
We hope you can fix the problem.

We try to send a debug-logfile of /etc/init.d/nis start as an attachment.


Thank you for your help.
-------------------------------------------------------------------------------------
WITTE Automotive
Schluessige Konzepte fuer die Automobilwelt/
Key Concepts for the Automotive World

http://www.WITTE-Automotive.com

severity 568518 important
kthxbye

Please file bugs with realistic severities, inflating the severity is
unhelpful.  In this case the problem only affects people trying to
interoperate with AIX, which leaves a rather large set of users
unaffected.

Please try with the current package from unstable to see if this issue
has been fixed subsequently.
 - capture the network traffic that occurs during startup.
 - start ypserv via gdb and obtain a backtrace (with the 'bt' command)
   when it crashes.

Hello Mr. Brwon,

we have tested with the unstable package of nis in lenny.

The test was good. NIS didn't crash.

In the debug-log are some messages: xdr_string: out of memory

I attched the logfile.





You find the tcpdump in the crash-start.txt and gdb-result in the
gdb-ypserv.log




Thomas Lefringhausen



p. h. G.: E. Witte Beteiligungsgesellschaft mbH, Velbert,
HRB 18366, Amtsgericht Wuppertal
Geschäftsführer: Rainer Gölz, Rolf-W. Walter

notfound 568518 3.17-30
thanks

OK, I believe I know what the issue is here - I suspect it is a variant
of http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=264587 (the fix for
which didn't cover all things).  Probably the AIX clients are triggering
a code path which Linux clients do not, causing it to try to send a
message it would not normally transmit and triggering the issue.