RE: iSCSI: ITT field in draft 11

["THALER,PAT (A-Roseville,ex1)" <pat_thaler@agilent.com>]

Deva,

An implementation can only rely on the field to always hold an ITT if it is
specified to do so. Currently, the BHS shows the field holding ITT "or
Opcode specific fields". Therefore, an implementation must check the Opcode
before it can use the field as an ITT.

The change Robert suggests isn't quite sufficient to accomplish allowing an
implementation to use the ITT before it checks the Opcode. To do that, we
would need to also need to change the BHS definition adding ITT to the list
of fields that always appear in the BHS and change the figure to reflect
that. Otherwise, vendor specific Opcodes and Opcodes defined in the future
might use that space in the BHS for something else.

Regards,
Pat

-----Original Message-----
From: Ranganathan, Deva [mailto:Deva_Ranganathan@adaptec.com]
Sent: Wednesday, March 20, 2002 8:59 AM
To: 'Robert D. Russell'; Julian Satran
Cc: ips@ece.cmu.edu; owner-ips@ece.cmu.edu
Subject: RE: iSCSI: ITT field in draft 11


Robert,

Is this not implementation related? 

-Deva


-----Original Message-----
From: Robert D. Russell [mailto:rdr@io.iol.unh.edu]
Sent: Wednesday, March 20, 2002 5:26 AM
To: Julian Satran
Cc: ips@ece.cmu.edu; owner-ips@ece.cmu.edu
Subject: iSCSI: ITT field in draft 11


Julian:

One of the changes that came in with draft 11 is that in the
Asynchronous Message PDU and in the Reject PDU the 4-byte field
at offset 16 became:
   16| Reserved -0xffffffff
whereas previously it had just been Reserved.

I believe the reason for explicitly specifying the value 0xffffffff
is because this field in all other PDUs is the Initiator Task Tag,
and for all those other PDUs this value means there is no valid
Initiator Transfer Tag.  This change now makes it possible for an
initiator to find the task associated with each incoming PDU before
looking at the opcode of the PDU.  So far so good.

The problem is that this field cannot be called Reserved due to section
9 which says that receivers MUST ignore reserved fields, thus defeating
the whole purpose of defining a value the initiator can use.

To correct this, the specs for both Asynchronous Message and Reject
should be changed to explicitly include the Initiator Task Tag field
and specify in the descriptions of each of those PDUs that this field
MUST be set to 0xffffffff.

If this is done, then every PDU will contain the Initiator Task Tag field,
and the description of the BHS in section 9.2.1 can be changed from:
  16| Initiator Task Tag or Opcode-specific fields
to
  16| Initiator Task Tag
which would clearly show the consistent meaning of this field over all PDUs.


Also, there are two related minor clarifications that should be made:

1) The 4th paragraph of section 9.16.1 currently says:
      "For Status SNACK, the Initiator Task Tag is reserved."
   It would be better if this said explicitly:
      "For Status SNACK, the Initiator Task Tag MUST be set to the
      reserved value 0xffffffff."

2) The last sentence in section 9.5.3 says:
      "For all the other functions this field is reserved."
   It would be better if this said explicitly:
      "For all the other functions this field MUST be set to the
      reserved value 0xffffffff."
   (Note that the corresponding sentence in section 9.6.2 already
   says this.)


Thanks,

Bob Russell
InterOperability Lab
University of New Hampshire
rdr@iol.unh.edu
603-862-3774