RE: Associating initiator names with SCSI commands

["Amir Shalit" <amir@astutenetworks.com>]

Please ignore (2)

-----Original Message-----
From: owner-ips@ece.cmu.edu [mailto:owner-ips@ece.cmu.edu]On Behalf Of
Amir Shalit
Sent: Thursday, May 09, 2002 10:27 AM
To: Julian Satran
Cc: ips@ece.cmu.edu; owner-ips@ece.cmu.edu
Subject: RE: Associating initiator names with SCSI commands


1) Editorial comment:
   chapter 1 defines SCSI Port Name as:
   A name made up as UTF-8 characters and basically...
   +---+
   Remove basically
   +---+

2) The iSCSI namning model on page 49 shows two iSCSI target nodes
   sharing the same portal group and the same TCP/IP addresses.
   How can one steer PDU headers to the correct iSCSI node given
   that setup?  Does it mean that a TCP connection doesn't belong
   to a single iSCSI session?

Amir



-----Original Message-----
From: owner-ips@ece.cmu.edu [mailto:owner-ips@ece.cmu.edu]On Behalf Of
Julian Satran
Sent: Tuesday, May 07, 2002 12:12 PM
To: Ken Craig
Cc: ips@ece.cmu.edu; owner-ips@ece.cmu.edu
Subject: Re: Associating initiator names with SCSI commands


No - the session is associated with the initiator - if you record the
session on which you received the command you have the association.
How you represent internally the session is an implementation issue - but I
assume you will not uses the names for that!

Julo


|---------+---------------------------->
|         |           "Ken Craig"      |
|         |           <kcraig@istor.com|
|         |           >                |
|         |           Sent by:         |
|         |           owner-ips@ece.cmu|
|         |           .edu             |
|         |                            |
|         |                            |
|         |           05/07/2002 07:11 |
|         |           PM               |
|         |           Please respond to|
|         |           "Ken Craig"      |
|         |                            |
|---------+---------------------------->

>---------------------------------------------------------------------------
------------------------------------|
  |
|
  |       To:       <ips@ece.cmu.edu>
|
  |       cc:
|
  |       Subject:  Associating initiator names with SCSI commands
|
  |
|
  |
|

>---------------------------------------------------------------------------
------------------------------------|



I have a question concerning associating
incoming SCSI commands with an initiator.
I come from a parallel SCSI background and
now find myself implementing a SCSI Target
port in an iSCSI world.  I have searched
the mailing list archives for discussions
on this subject but have been unable to find
anything that succinctly answers my question
so please bear with me.

In the parallel SCSI world association of an
initiator with a new command is very
straight-forward as the initiator's ID is
encapsulated in the Identify message that
occurs with the SCSI Selection phase that
precedes receiving the new command.

When I read the latest version of the iSCSI
draft (rev. 12) the only statement I seem to
find that correlates to this association is
in Section 2.2.3 on page 34 in the 2nd
sentence of the 3rd paragraph.

"Any persistent state (e.g., persistent reservations)
on the target that is associated with a SCSI
initiator port is identified based on the
value pair (InitiatorName, ISID)."

When I searched the mailing list archives I
came across statements that said this
association was done using ISID and TSID (now
TSIH?) but I do not see these statements in the
latest draft so I'm assuming that there was some
reason this association method was dropped.

My question is:
In order to associate initiators with incoming
commands to a SCSI Target do I have to compare
the Initiator Name and ISID (up to ~268 bytes?)
for every command I receive against a list of
logged in initiators or is there another method
using a lot fewer number of bytes?

I had thought about using the IP address in the
IP header but the draft seems to say that is not
allowed because IP addresses can change.  It
seems like I must perform this potentially rather
long comparison if I support multiple initiators
because I can not be guaranteed that different
initiators would not use the same ISID during
their login.  Am I wrong?

Thanks in advance,
Kenneth Ray Craig, Jr.