[OpenAFS-devel] Strange (non)bug in openafs-1.2.7...

Jody Pearson jody.pearson@cern.ch
Tue, 01 Apr 2003 15:22:50 +0200


Hi List,

I have been banging my head against a very subtle problem with 
openafs-1.2.7.

The machine in question was running openafs happily, until it crashed, 
and a crash dump was taken.

The machine;

2x Pentium III 1Ghz
1Gb of Ram
EEpro100
Serverworks ATA

Linux 2.4.18-18 Redhat Kernel.
The same binaries now no longer work.

I have compared (using cmp) the binaries against a working machine.

/usr/vice/etc/afsd - identical with working machine
/usr/vice/etc/modload/libafs.......mp.o - identical with working machine
kernel - identical with a working machine.
insmod - identical with a working machine...

And so on, basically, the files on the machine have not changed.

The problem ?

/afs does not mount, and fails with these messages;

afsd: No check server daemon in client.
afsd: All AFS daemons started.
afsd: Can't mount AFS on /afs(22)


I looked into this, and found that any afs_syscall() is returning 
-ENOSYS, but the funtion call is listed in /proc/ksyms.

Also the module initialization works (AFS registered in /proc/filesystems)

I am really lost - I can't see any difference between this and a working 
machine.

Other things tried - remaking /usr/vice/cache (new fs) that does not fix 
the problem either.

I saw from the archives that a guy in Stanford has had the same problem, 
he was told to recompile.

So I recompiled libafs on the machine and - same problem !

Any help on where to look/a solution would be greatly appreated.

I can send machine output on request.

Thanks

Jody Pearson