Strange bind crash on PPC G5 PowerMac running Squeeze plus 2.6.39 backports kernel

2011-12-12 Thread Mark Cave-Ayland

Hi all,

I'm not sure whether this is the right place to mention this or not, 
however we're experiencing a problem on one of our PPC G5 Mac servers 
and was wondering if someone could point me in the right direction?


Basically we have an old PPC G5 Mac which we have re-purposed to become 
a wireless access point for our office. It's currently running Debian 
Squeeze with a 2.6.39 backports kernel. The problem is that every so 
often (perhaps once every 1-2 weeks) the bind daemon, which is 
configured to forward requests to another nameserver, locks hard and can 
only be removed by kill -9 followed by a restart. AFAICT no other 
daemons on the server seem to be affected.


In order to try and debug the issue, I've rebuilt the PPC bind .deb with 
debug/nostrip, but unfortunately it has happened again and I still can't 
see any symbols in the resulting process, e.g.



root@cheeseburger:~# /etc/init.d/bind9 stop
Stopping domain name service...: bind9^Crndc: recv failed: operation 
canceled

root@cheeseburger:~# ps -ef | grep named
bind 17451 1  0 Dec01 ?00:00:05 /usr/sbin/named -u bind
root 22050 21760  0 14:26 pts/000:00:00 grep named
root@cheeseburger:~# file /usr/sbin/named
/usr/sbin/named: ELF 32-bit MSB shared object, PowerPC or cisco 4500, 
version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 
2.6.18, with unknown capability 0x4100 = 0x13676e75, with unknown 
capability 0x1 = 0xb0401, not stripped

root@cheeseburger:~# gdb -p 17451
GNU gdb (GDB) 7.0.1-debian
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later 
http://gnu.org/licenses/gpl.html

This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type show copying
and show warranty for details.
This GDB was configured as powerpc-linux-gnu.
For bug reporting instructions, please see:
http://www.gnu.org/software/gdb/bugs/.
Attaching to process 17451
Reading symbols from /usr/sbin/named...done.
0x1f8c1d7c in ?? ()
(gdb) thread apply all bt full

Thread 1 (process 17451):
#0  0x1f8c1d7c in ?? ()
No symbol table info available.
#1  0x1f8c1d68 in ?? ()
No symbol table info available.
#2  0x1fe4d36c in ?? ()
No symbol table info available.
#3  0x1fe4d468 in ?? ()
No symbol table info available.
#4  0x2032d6b4 in ?? ()
No symbol table info available.
#5  0x1f8a963c in ?? ()
No symbol table info available.
#6  0x1f8a9800 in ?? ()
No symbol table info available.
#7  0x in ?? ()
No symbol table info available.
(gdb)

root@cheeseburger:~# ldd /usr/sbin/named
linux-vdso32.so.1 =  (0x0010)
liblwres.so.60 = /usr/lib/liblwres.so.60 (0x6ff04000)
libdns.so.69 = /usr/lib/libdns.so.69 (0x6fd58000)
libgssapi_krb5.so.2 = /usr/lib/libgssapi_krb5.so.2 (0x6fd02000)
libcrypto.so.0.9.8 = /usr/lib/libcrypto.so.0.9.8 (0x6fb58000)
libbind9.so.60 = /usr/lib/libbind9.so.60 (0x6fb2b000)
libisccfg.so.62 = /usr/lib/libisccfg.so.62 (0x6faf2000)
libisccc.so.60 = /usr/lib/libisccc.so.60 (0x6fac9000)
libisc.so.62 = /usr/lib/libisc.so.62 (0x6fa4e000)
libdb-4.6.so = /usr/lib/libdb-4.6.so (0x6f8cf000)
libldap_r-2.4.so.2 = /usr/lib/libldap_r-2.4.so.2 (0x6f85f000)
liblber-2.4.so.2 = /usr/lib/liblber-2.4.so.2 (0x6f831000)
libcap.so.2 = /lib/libcap.so.2 (0x6f80d000)
libpthread.so.0 = /lib/libpthread.so.0 (0x6f7d2000)
libxml2.so.2 = /usr/lib/libxml2.so.2 (0x6f653000)
libc.so.6 = /lib/libc.so.6 (0x6f4c2000)
libGeoIP.so.1 = /usr/lib/libGeoIP.so.1 (0x6f45e000)
libkrb5.so.3 = /usr/lib/libkrb5.so.3 (0x6f374000)
libk5crypto.so.3 = /usr/lib/libk5crypto.so.3 (0x6f32c000)
libcom_err.so.2 = /lib/libcom_err.so.2 (0x6f309000)
libkrb5support.so.0 = /usr/lib/libkrb5support.so.0 (0x6f2e1000)
libdl.so.2 = /lib/libdl.so.2 (0x6f2bd000)
libkeyutils.so.1 = /lib/libkeyutils.so.1 (0x6f29b000)
libresolv.so.2 = /lib/libresolv.so.2 (0x6f264000)
libz.so.1 = /usr/lib/libz.so.1 (0x6f22e000)
libsasl2.so.2 = /usr/lib/libsasl2.so.2 (0x6f1f2000)
libgnutls.so.26 = /usr/lib/libgnutls.so.26 (0x6f122000)
libattr.so.1 = /lib/libattr.so.1 (0x6f0fd000)
/lib/ld.so.1 (0x205bf000)
libm.so.6 = /lib/libm.so.6 (0x6f02f000)
libtasn1.so.3 = /usr/lib/libtasn1.so.3 (0x6f00e000)
libgcrypt.so.11 = /usr/lib/libgcrypt.so.11 (0x6ef6b000)
libgpg-error.so.0 = /usr/lib/libgpg-error.so.0 (0x6ef46000)


Looking at the above address mappings, I'm wondering if something is 
getting wedged in the kernel somewhere - if so, can anyone point me 
towards some debug symbols that I can load into the relevant backports 
kernel for more information?


root@cheeseburger:~# uname -a
Linux cheeseburger 2.6.39-bpo.2-powerpc64 #1 SMP Thu Aug 4 12:38:28 UTC 
2011 ppc64 GNU/Linux


Note that I can probably leave the process in this state for a 

Re: Strange bind crash on PPC G5 PowerMac running Squeeze plus 2.6.39 backports kernel

2011-12-12 Thread Frédéric Massot

Le 12/12/2011 15:35, Mark Cave-Ayland a écrit :

Hi all,

I'm not sure whether this is the right place to mention this or not,
however we're experiencing a problem on one of our PPC G5 Mac servers
and was wondering if someone could point me in the right direction?

Basically we have an old PPC G5 Mac which we have re-purposed to become
a wireless access point for our office. It's currently running Debian
Squeeze with a 2.6.39 backports kernel. The problem is that every so
often (perhaps once every 1-2 weeks) the bind daemon, which is
configured to forward requests to another nameserver, locks hard and can
only be removed by kill -9 followed by a restart. AFAICT no other
daemons on the server seem to be affected.


Hi,

There is currently a vulnerability in Bind9, which makes it crash. You 
may be in this case.


https://lwn.net/Articles/467779


Regards.
--
==
|  FRÉDÉRIC MASSOT   |
| http://www.juliana-multimedia.com  |
|   mailto:frede...@juliana-multimedia.com   |
===Debian=GNU/Linux===


--
To UNSUBSCRIBE, email to debian-powerpc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/4ee61a79.7000...@juliana-multimedia.com



Re: Strange bind crash on PPC G5 PowerMac running Squeeze plus 2.6.39 backports kernel

2011-12-12 Thread Mark Cave-Ayland

On 12/12/11 15:15, Frédéric Massot wrote:


Hi,

There is currently a vulnerability in Bind9, which makes it crash. You
may be in this case.

https://lwn.net/Articles/467779


Hi Frédéric,

That was one of my thoughts, but with an unstripped executable it should 
be possible to determine a function name or similar for a userspace crash.


After some more investigation, I did find the contents of /proc/modules 
interesting:


root@cheeseburger:~# cat /proc/modules
(cut)
snd_page_alloc 9929 1 snd_pcm, Live 0xd1fbf000
ath5k 162739 0 - Live 0xd1f83000
ath 20320 1 ath5k, Live 0xd1f3c00
(cut)

Now I don't know much about the PPC architecture, but comparing with 
frame 0 of the backtrace at 0x1f8c1d7c the addresses look similar enough 
to imply perhaps the issue lies within ath5k somewhere (which is the 
kernel module for the wireless network card).


I think my next move will be to figure out how to build an unstripped 
PPC kernel from backports and then raise a bug report with a suitable 
backtrace for the ath5k people.



Many thanks,

Mark.

--
Mark Cave-Ayland - Senior Technical Architect
PostgreSQL - PostGIS
Sirius Corporation plc - control through freedom
http://www.siriusit.co.uk
t: +44 870 608 0063

Sirius Labs: http://www.siriusit.co.uk/labs


--
To UNSUBSCRIBE, email to debian-powerpc-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/4ee62266.1090...@siriusit.co.uk