First attempt with arla-0.6

[LeBlanc@mcc.ac.uk]

On Monday 1 June 1998, I wrote:

>I got a copy of arla-0.6 and tried to compile it on my Linux
>box (a Debian system with the frozen (hamm) release.  The first
>problem I had (which I also had with the December KTH version
>of Kerberos) is that configure aborts when it tries to create the
>file include/version.h
...
>because the variable Date is set to `date`, which on my system
>produces, for example, 'Mon Jun  1 12:41:40 /etc/localtime 1998',
>where the slashes cause the problem.

Actually this seems to have been caused by the file /etc/localtime
being absent from my system, evidently an installation failure.

I wrote:
>Although there are no installation instructions, ...

My apologies for this; it should be clear from the rest of what I
said that I had in fact read the INSTALL file and done things
mentioned there; e.g., tried 'arlad -t', and so forth.  I meant
that the instructions seemed to be incomplete, but what I said
was not particularly helpful.  Anyway, I went on to say:

>The result is that
>the module xfs.o does not load:

>./xfs.o: unresolved symbol inode_setattr
...

and so on.

"A.J.Martin" <A.J.Martin at qmw.ac.uk> wrote:
>    I think this is the same as I got last week, try removing 
>-DMODVERSIONS  from the Makefile.

That is, in effect, the following:

diff -ru2N arla-0.6/xfs/linux/Makefile.in arla-MCC/xfs/linux/Makefile.in
--- arla-0.6/xfs/linux/Makefile.in      Mon May 25 02:41:24 1998
+++ arla-MCC/xfs/linux/Makefile.in      Tue Jun  2 08:54:26 1998
@@ -29,5 +29,5 @@
 XFS_BIN		= $(MODULE)

-DEFS		= @DEFS@ -DMODULE -DMODVERSIONS -D__KERNEL__ -DDEBUG
+DEFS		= @DEFS@ -DMODULE -D__KERNEL__ -DDEBUG
 INC		= $(srcdir)/../include \
		  ../../include \

This works, at least in the sense that the module created by this
loads successfully.

Magnus Ahltorp <map at stacken.kth.se> wrote:
>What modutils or modules version are you using ("insmod -V")?
>It could be that a newer version of modutils will solve the problem.

It is modutils 2.1.85, and the kernel is 2.0.32.

Assar Westerlund <assar at sics.se> wrote:
>Magnus and I have discussed this and it seems that right answer is the
>appended patch.  Please try it out and see if it works for you.

Index: xfs_locl.h
===================================================================
RCS file: /usr/local/cvsroot/arla/xfs/linux/xfs_locl.h,v
retrieving revision 1.12
retrieving revision 1.13
diff -u -w -u -w -r1.12 -r1.13
--- xfs_locl.h	1998/05/19 06:12:36	1.12
+++ xfs_locl.h	1998/06/02 17:16:12	1.13
@@ -36,7 +36,7 @@
  * SUCH DAMAGE.
  */
 
-/* $Id: xfs_locl.h,v 1.12 1998/05/19 06:12:36 map Exp $ */
+/* $Id: xfs_locl.h,v 1.13 1998/06/02 17:16:12 assar Exp $ */
 /* $Source: /usr/local/cvsroot/arla/xfs/linux/xfs_locl.h,v $ */
 
 #ifdef HAVE_CONFIG_H
@@ -44,6 +44,7 @@
 #endif
 
 #include <linux/version.h>
+#include <linux/modversions.h>
 #include <linux/errno.h>
 #include <linux/fs.h>
 #include <linux/malloc.h>

=====================================

This also works, at least in that the module loads without errors.
So I have two possible solutions to my immediate problem.  As I
said before, 'arlad -t' appears to work, in that I can cd, ls,
and cat files from my local AFS cell.  The utilities pts and vos
work.  There are still problems, however:

avl 13:40:59 > ./startarla
avl 13:41:17 > lsmod
Module         Pages    Used by
xfs                6            2
nfs               12            4 (autoclean)
serial             8            2 (autoclean)
3c509              2            1
avl 13:41:25 > mount
/dev/hda1 on / type ext2 (rw)
proc on /proc type proc (rw)
/dev/hda6 on /usr type ext2 (rw)
arla on /afs type xfs (rw)
avl 13:41:28 > ls /afs/
ls: /afs/: Not a directory
avl 13:41:31 > ls /afs
avl 13:41:40 > file /afs
/afs: directory
avl 13:41:44 >

I have been trying to get debug information.  There's not a lot
of documentation, but I've tried 'arlad --debug=all' without any
effect.  Can anyone suggest further things to look at.  The kthkrb4
stuff is installed, but it is the December version, and does not
recognise arla as AFS, so that, for example, pagsh doesn't actually
get a PAG.

     -- Owen
     LeBlanc at mcc.ac.uk