[OpenAFS-devel] Testing of 1.3.79 on Mandrakelinux 10.1 : looking good

Jeffrey Hutzelman jhutz@cmu.edu
Wed, 16 Mar 2005 19:27:34 -0500


On Tuesday, March 08, 2005 05:48:15 PM +0100 Andr=E9 Derrick Balsa=20
<Andre.Balsa@inria.fr> wrote:

> Hi,
>
> We at INRIA are at the moment testing 1.3.79 on Mandrakelinux 10.1, which
> uses  a slightly patched 2.6.8.1 kernel.
>
> The client is undergoing heavy testing with a 1.2.13 server. I will
> report on  any observed problems, but right now things look good.
>
> I have however a minor problem to report during compilation : I had to
> use an  ugly hack to have the libafs module load correctly.
>
> If 1.3.79 is compiled out-of-the-box, I get an "unknown symbol
> audit_putname"  error when insmodding the libafs module. Checking the
> kernel, this is an  audited version of __putname. (the linux kernel
> people have added a syscall  auditing function recently).
> I simply hacked 1.3.79, replacing the single instance of putname by
> __putname;  this is in /src/afs/LINUX/osi-misc.c. Obviously I am using
> the non-audited  version of putname, which is no problem as far as we are
> concerned here at  INRIA. Still, I would prefer if no hack was needed.

This sounds like audit_putname() isn't correctly exported.  Look in your=20
kernel source at kernel/auditsc.c; just below the definition of=20
audit_putname() there should be a line like

EXPORT_SYMBOL(audit_putname)

This export is present in stock 2.6.9; if it is not present in your kernel=20
then it is either a Mandrake bug or 2.6.8 bug which was fixed in 2.6.9.  In =

either case it is not a bug in OpenAFS.


-- Jeffrey T. Hutzelman (N3NHS) <jhutz+@cmu.edu>
   Sr. Research Systems Programmer
   School of Computer Science - Research Computing Facility
   Carnegie Mellon University - Pittsburgh, PA