[Crash-utility] Problem with "xm save" x86-64 cores -crash.4.0-3.5

[Dave Anderson]

Looking at the vmlinux files with dwarfdump and gdb 6.3 alone, I can
see now that in the target 2.6.16 xen kernel (built with gcc 4.0.0,
if that matters), what is happening is that for static kernel data that is
declared without being initialized, its data type information is being
completely left out of the debug info section.

So, taking kernel/pid.c as an example, symbols such as "pid_hash",
"pidhash_shift" and "last_pid" are declared uninitialized.  Looking
at the 2.6.16 xen kernel built with gcc 4.0.0:

  # gdb vmlinux
  GNU gdb Red Hat Linux (6.3.0.0-1.63rh)
  Copyright 2004 Free Software Foundation, Inc.
  GDB is free software, covered by the GNU General Public License, and you are
  welcome to change it and/or distribute copies of it under certain conditions.
  Type "show copying" to see the conditions.
  There is absolutely no warranty for GDB.  Type "show warranty" for details.
  This GDB was configured as "x86_64-redhat-linux-gnu"...Using host libthread_db library "/lib64/tls/libthread_db.so.1".

  (gdb) whatis last_pid
  type = <data variable, no debug info>
  (gdb) whatis pid_hash
  type = <data variable, no debug info>
  (gdb) whatis pidhash_shift
  type = <data variable, no debug info>
  (gdb)

Then looking at a non-xen 2.6.16-era kernel built with gcc 4.1.0:

  # gdb /usr/dumps/kdump/vmlinux
  GNU gdb Red Hat Linux (6.3.0.0-1.63rh)
  Copyright 2004 Free Software Foundation, Inc.
  GDB is free software, covered by the GNU General Public License, and you are
  welcome to change it and/or distribute copies of it under certain conditions.
  Type "show copying" to see the conditions.
  There is absolutely no warranty for GDB.  Type "show warranty" for details.
  This GDB was configured as "x86_64-redhat-linux-gnu"...Using host libthread_db library "/lib64/tls/libthread_db.so.1".

  (gdb) whatis last_pid
  type = int
  (gdb) whatis pid_hash
  type = struct hlist_head *[4]
  (gdb) whatis pidhash_shift
  type = int
  (gdb)

I don't believe the "xen" angle makes a difference.  I only have
2.6.17-era xen kernels as a reference -- built with gcc 4.1.1 -- and
they do not exhibit this behavior.

I presume that this affects files other than kernel/pid.c, but haven't
checked any further.

So it would be interesting to know whether an upgrade in compilers
would make a difference.

Regardless of that, I'll still come up with a fix and release for this,
since there's really nothing I can do with kernels of this ilk except
to work around their shortcomings.  If the debug data's not there,
it's not there...

Thanks,
  Dave