[pnfs] Connectathon update

[Rahul Iyer]

Hey guys,
This is a (relatively) short message to sum up what we did/learned wrt 
pNFS at Connectathon.
1. Resolved issues regarding the clientid on the DS. It was agreed that 
the clients would not call EXCHANGE_ID
    on the DSs too. For more detail kindly refer to thread "Create 
Session for pNFS DS" on nfsv4 at ietf.org.

2. The Linux pNFS client was able to successfully interoperate with the 
Sun pNFS server with a few changes:
    - The clientid and seqid in OPEN, CLOSE, LOCK etc. needed to be 0 
instead of the actual values. The clientid
       and sequencing would be provided by the sequence op and the 
sessions infrastructure. To "fix" this, I zeroed out
       these fields in the XDR layer. It's a hack, I agree, and for this 
reason, isn't checked in.
    - In order to implement correct sequence ids for the sequence op, I 
implemented a half baked single slot slot table.
      This enabled us to get the sequence ids right, but it wasn't fully 
implemented to handle I/O > 1 page (4K).

So, the client interoperability came with a few twiddles, but it's 
better than nothing at all.

3. The Sun client could interoperate with the Netapp pNFS server. 
However, the Netapp server had sequence id checking
    turned off.

4. Bugs found and fixed:
    - Benny found an incorrect implementation of layout return and fixed 
it. Changes checked into the tree.
    - There is a soft lockup bug in the client. It's yet to be fixed.

Please feel free to add anything I may have left out.
Thanks,
Regards
Rahul