[OpenAFS] Serious problem upon rebuild

[Derek Atkins]

Madhusudan Singh <chhabra@eecs.umich.edu> writes:

> -rw-r--r--    1 root     root        42472 Sep 20 12:46 libafs.map
> -rw-r--r--    1 root     root       495999 Sep 20 22:21
> libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002-i386.o
> -rw-r--r--    1 root     root       495999 Sep 20 22:24
> libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002-i586.o
> -rw-r--r--    1 root     root       495999 Sep 20 22:27
> libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002-i686.o
> -rw-r--r--    1 root     root       495999 Sep 20 22:30
> libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002-athlon.o
> -rw-r--r--    1 root     root       630178 Sep 20 22:34
> libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002-i386.mp.o
> -rw-r--r--    1 root     root       630178 Sep 20 22:37
> libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002-i586.mp.o
> -rw-r--r--    1 root     root       630178 Sep 20 22:47
> libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002-i686.mp.o
> -rw-r--r--    1 root     root       630178 Sep 20 23:17
> libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002-athlon.mp.o
> -rw-r--r--    1 root     root         1493 Sep 20 23:19 SymTable
> -rw-r--r--    1 root     root       524288 Sep 21 23:45
> libafs-2.4.18-3-i686.mp.o;3d8d3c93
> 
> The first few libafs-2.4.18-3* are understandable (old versions) but
> the newest one is puzzling.

You cannot use the SPEC file to build against a non Red Hat kernel
source tree.  If you compile your own kernel, you _MUST_ use the
openafs-kernel-source module and build your module that way.  The
problem here is that you have different modules that are built
differently but look the same to the loader.  Bad Juju.

> However, I tried to insmod the libafs-2.4.19-LAPTOP-CUSTOM-17SEP2002*
> modules - in the order i686, i586 and i386. After each attempt I got a
> message that it would taint the kernel : licence not ...

Known issue.  Ignore it.

> When I rebooted, I got the same old hang at :
> (quoting from memory)
> 
> Cache size (100000) should be less than 95 %(which is 96035) ... Lower
> cache size

How big is your /usr/vice/cache partition?

> I am not entirely certain that I used to get "Lower cache size" type
> messages during normal operation with the older rpm's.
> 
> Are the 1.2.6 rpm's for Redhat 7.3 for kernel 2.4.19 ? If so, would
> the kernel $EXTRAVERSION string cause problems ?

No.  Red Hat does not distribute a 2.4.19 kernel.  The 1.2.6 RPMS are
only built for 2.4.18-3,4, and 5.  I expect the 1.2.7 RPM will also
include 2.4.18-10.  As RH does not ship a 2.4.19 kernel, there is no
RPM that is pre-built for 2.4.19.

> Just in case it helps to localise the problem, I am attaching some
> additional info :
[snip]
> # Symbol  Version       MinCPU  Module
> #=======  ============  ======  ====================
> iget4     smp_bca964ec  -       2.4.19-LAPTOP-CUSTOM-17SEP2002-athlon.mp
>  > sock_create=smp_dae70453
> iget4     smp_bca964ec  -       2.4.19-LAPTOP-CUSTOM-17SEP2002-athlon
>  > sock_create=smp_dae70453
> iget4     smp_bca964ec  -       2.4.19-LAPTOP-CUSTOM-17SEP2002-i386.mp
>  > sock_create=smp_dae70453
> iget4     smp_bca964ec  -       2.4.19-LAPTOP-CUSTOM-17SEP2002-i386
>  > sock_create=smp_dae70453
> iget4     smp_bca964ec  -       2.4.19-LAPTOP-CUSTOM-17SEP2002-i586.mp
>  > sock_create=smp_dae70453
> iget4     smp_bca964ec  -       2.4.19-LAPTOP-CUSTOM-17SEP2002-i586
>  > sock_create=smp_dae70453
> iget4     smp_bca964ec  -       2.4.19-LAPTOP-CUSTOM-17SEP2002-i686.mp
>  > sock_create=smp_dae70453
> iget4     smp_bca964ec  -       2.4.19-LAPTOP-CUSTOM-17SEP2002-i686
>  > sock_create=smp_dae70453

This means you compiled against a non-RH kernel source tree, which
you cannot do.

Might I suggest:

1) rpm -e All your openafs RPMs (remove everything)
2) rm -rf /usr/vice
3) rpm -ivh openafs, openafs-client, openafs-kernel, openafs-kernel-source
   (install the RPMS from openafs.org)
4) cd /usr/src/openafs-1.2.6
5) read the README and follow the instructions to build for your 2.4.19 kernel

If that still fails, then it's possible that something was added to
2.4.19 that OpenAFS doesn't know about.  But we would need to get more
of a crash dump (oops message) from your machine.

-derek

-- 
       Derek Atkins, SB '93 MIT EE, SM '95 MIT Media Laboratory
       Member, MIT Student Information Processing Board  (SIPB)
       URL: http://web.mit.edu/warlord/    PP-ASEL-IA     N1NWH
       warlord@MIT.EDU                        PGP key available