>>>>> On Tue, 23 Aug 2005 14:28:53 +0200, Volker Sauer <[EMAIL PROTECTED]> said:

  Volker> On Di, 23 Aug 2005, Martin Simmons <[EMAIL PROTECTED]> wrote:
  >> >>>>> On Tue, 23 Aug 2005 12:30:45 +0200, Kern Sibbald <[EMAIL PROTECTED]>=
  Volker>  said:
  >> =20
  Kern> I've now found the time to look over your debug output below.  My=
  Volker>  analysis=20
  Kern> leads me to believe that what is show is "impossible". That is th=
  Volker> e code flow=20
  Kern> as created in the source code cannot possibly do what is indicate=
  Volker> d in the=20
  Kern> dump.  What is shown in the dump is that the subroutine get_next_=
  Volker> jcr_ is=20
  Kern> recursively called with the same argument (not possible).  This w=
  Volker> ill almost=20
  Kern> surely lead to a blocked situation.
  >> =20
  Kern> How could this happen?  Bad compiler code, an interrupt that happ=
  Volker> ens and=20
  Kern> restarts the stack at the wrong point, memory error (I doubt), ...
  >> =20
  >> I doubt that is really happening -- much more likely is that gdb can't
  >> understand the stack.  Look at the other threads and you'll see that
  >> jobq_server appears to call jobq_server!
  >> =20
  >> In all these cases, the extra "call" happens where there is a real call to
  >> something like pthread_mutex_lock.  The pthread library is probably compi=
  Volker> led
  >> with too much optimization and/or insufficient debug info for gdb to
  >> understand the stack inside there.

  Volker> Some more information:
  Volker> For a test I got rid of /lib/tls. Obviously bacula now forks into
  Volker> several processes which (I assume) means, that another threading model
  Volker> is used (I don't know anything about threading). (I read the manual
  Volker> which says, that this only applies to Kernel 2.4 on Red Hat, but it's
  Volker> easy to do and I'll give it a try).

  Volker> This time bacula crashed with the following gdb output:

  Volker> dakar: ~ 1# gdb /usr/sbin/bacula-dir
  Volker> GNU gdb 6.3-debian
  Volker> Copyright 2004 Free Software Foundation, Inc.
  Volker> GDB is free software, covered by the GNU General Public License, and 
you
  Volker> are welcome to change it and/or distribute copies of it under certain
  Volker> conditions. Type "show copying" to see the conditions.
  Volker> There is absolutely no warranty for GDB.  Type "show warranty" for
  Volker> details.
  Volker> This GDB was configured as "i386-linux"...Using host libthread_db
  Volker> library "/lib/

  Volker> (gdb) run -s -f -c /etc/bacula/bacula-dir.conf
  Volker> Starting program: /usr/sbin/bacula-dir -s -f -c
  Volker> /etc/bacula/bacula-dir.conf
  Volker> [Thread debugging using libthread_db enabled]
  Volker> [New Thread 16384 (LWP 32251)]
  Volker> [New Thread 32769 (LWP 32255)]
  Volker> [New Thread 16386 (LWP 32256)]
  Volker> [New Thread 32771 (LWP 32257)]
  Volker> [New Thread 49156 (LWP 32264)]
  Volker> [Thread 49156 (LWP 32264) exited]
  Volker> [New Thread 65540 (LWP 938)]
  Volker> [Thread 65540 (LWP 938) exited]
  Volker> [New Thread 81924 (LWP 4421)]
  Volker> [Thread 81924 (LWP 4421) exited]
  Volker> [New Thread 98308 (LWP 4971)]
  Volker> [Thread 98308 (LWP 4971) exited]
  Volker> [New Thread 114692 (LWP 5488)]
  Volker> [New Thread 131077 (LWP 5489)]
  Volker> [New Thread 147462 (LWP 5491)]
  Volker> [New Thread 163847 (LWP 5492)]
  Volker> [New Thread 180232 (LWP 5496)]
  Volker> [New Thread 196617 (LWP 5497)]
  Volker> [New Thread 213002 (LWP 5499)]
  Volker> [New Thread 229387 (LWP 5500)]
  Volker> [Thread 147462 (LWP 5491) exited]
  Volker> Cannot find thread 147462: invalid thread handle
  Volker> (gdb)
  Volker> (gdb)

  Volker> ...
  Volker> Segmentation fault

  Volker> Again segfault?? It did not reload the config file or something like
  Volker> this...

This is gdb barfing on threads and then giving the Segmentation fault, which
will be either be caused by a kernel bug or a gdb bug.  Might be fixed in the
latest CVS of gdb, but you'll need to build that yourself from source (not
difficult on Linux).

__Martin


-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
_______________________________________________
Bacula-users mailing list
Bacula-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/bacula-users

Reply via email to