RE: iSCSI: v15 R2T and DATA-OUT

["Julian Satran" <Julian_Satran@il.ibm.com>]

Paul,

It would also be a bad idea to add too many words to the protocol.
The protocol is clear about field consistency.

Julo


                                                                                                                                            
                      Paul Koning                                                                                                           
                      <ni1d@arrl.net>          To:       rod.harrison@windriver.com                                                         
                      Sent by:                 cc:       ips@ece.cmu.edu                                                                    
                      owner-ips@ece.cmu        Subject:  RE: iSCSI: v15 R2T and DATA-OUT                                                    
                      .edu                                                                                                                  
                                                                                                                                            
                                                                                                                                            
                      07/25/2002 05:35                                                                                                      
                      PM                                                                                                                    
                                                                                                                                            
                                                                                                                                            



Excerpt of message (sent 24 July 2002) by Rod Harrison:
>
>            OK, it's been a long day. Let me try this one more time.
>
>            I was write (almost) first time but I meant DATA-OUT instead
of
> DATA-IN. Here's what I meant to say ...
>
>            There is a potential inconsistency in the description of the
use of
> the LUN field in DATA-OUT and R2T in the working v15 draft.
>
>            9.7.3 Target Transfer Tag, for DATA-OUT last paragraph says
...
>
>            "If the Target Transfer Tag is provided, then the LUN field
MUST hold
> a valid value and be consistent with whatever was specified with the
> command;"
>
>            9.8.5 Target Transfer Tag, for R2T says ...
>
>            "The Target Transfer Tag and LUN are copied in the outgoing
data PDUs
> and are used by the target only."
>
>            Potentially a target could return a different LUN field in the
R2T,
> for perhaps some funky LUN mapping or other internal reason expecting
> it to be copied to the DATA-OUT as per the R2T text.

I think that's a very bad idea and should not be permitted.

If someone wants to do LUN mapping, that's fine, so long as it's
transparent to the initiator.  To map a LUN and then let the mapped
number leak out to the initiator is broken.

>            I think we need to indicate what is expected of the initiator
if the
> LUN field in the R2T does not match the LUN in the command PDU.

Given that the subject was brought up I tend to agree that the spec
should say something.  I would prefer it to say that this is a
protocol error.

              paul