[OpenAFS-devel] Openafs 1.0.3 + S/390 + kas wierdness

[jacobi@de.ibm.com]

> * kas setkey afs -new_key "<KEY THAT BOS JUST GOT"> -kvno <KVNO>
> -admin_user admin  which returns the error about bad parity or weak.

We have the same problem, and I traced the problem till those two lines:

admin_tools.c:771 (in the "obj"-folder)

ubik_Call (KAM_SetPassword, conn, 0, name, instance, kvno, key);

Till this line kas (I tried it with this program) behaves identically for
Linux/S390 as for Linux/x86.

But after he reaches KAM_SetPassword in kautch.cs.c (whheeee, generated
code; but I understand we need xdr to drive different architectures)
their is a striking difference:
The pointer "password" seems to be "too large" by four bytes. So I see
one key as

0000abcd on x86, and as
abcdxxxx on S/390, whereas I am not for sure whether the "x" bytes are
valid.

It is not an endianess-issue (than you may see something like "dcba0000",
which is not the case here)!
Now can anybody tell me what happens between the ubik_Call call and the
time when SetPassword is really entered? And why do all the other xdr-calls
work flawlessly but SetPassword doesn't? Many questions ...

Carsten Jacobi