re: draft-ieft-ips-iSCSI-04.txt

[julian_satran@il.ibm.com]

Martin,

Thanks for the careful reading.

Answers in text .

Regards,
Julo

"Martin, Nick" <Nick.Martin@compaq.com> on 01/03/2001 01:01:30

Please respond to "Martin, Nick" <Nick.Martin@compaq.com>

To:   "'Julian_Satran@vnet.ibm.com'" <satran@haifa.vnet.ibm.com>
cc:
Subject:  re: draft-ieft-ips-iSCSI-04.txt




Hello Julian,

I have been attempting a demonstration implementation of iSCSI in C on
Linux.  I started with draft 03 and have now moved to draft 04.  My program
is not yet very mature and has many required aspects not implemented, but
using draft 03 I was able to perform some SCSI operations.
+++ excellenr! +++ 05 will be here in a day or two +++
During my study of draft 04, I found what appear to be several minor errors
or inconsistencies.

I am happy to provide you with the following list, although it may not
contain anything you do not already know about.

On page 44 the field Status or Rsvd, and the S bit.  It is my understanding
that it is not valid to return SCSI status other than GOOD (zero) in a SCSI
read data PDU.  Otherwise sense data bust be sent using SCSI response PDU.
If this is the case, then the S bit is the only significant field.  If it
is
set the status must be zero, if it is not set then the status is reserved
(0).
+++ this is a bit murky but we (as FCP before) chose to consider
over/underuns
as not necessarily errors.  SCSI will have to decide if its is or not based
on the specific command context (e.g., it may break linked commands etc.)
but not iSCSI +++

On page 46, the description of b7 is P bit, but this has been changed to F
bit.
+++ fixed +++
On page 54, the Op_code is 0x83 but should be 0x43.  I am not clear whether
the X bit should have been defined.
+++ fixed +++
On page 60, the Op_code field is 0x80, but should be 0x40.  No X bit is
defined.

On page 64, the Op_code field is 0x86, but should be 0x46.  No X bit.

On page 67, Op_code 0x90 should be 0x50.  No X bit.

On page 69, Op_code 0x91 should be 0x51.  No X bit.

On page 72, Op_code 0xef should be 0x6f.  There is a 0 in the X bit
position.

+++ all fixed +++

On page 73, it states that DataPDULength and FirstBurstSize are in units of
512 bytes, however on page 110 and 111, these are stated to be in units of
4096 bytes.  On page 112 there are references to DataPDULength*512 which
may
be related to the units of DataPDULength.

+++ fixed - to 512 to be aligned with other T10 standards +++

On page 95, the polynomial for crc-31Q has some repeated terms (x**7 and
x**5).  For now I will presume this is intentional.  I further presume the
notation x**5 means x raised to the 5th power.

++++ Thanks it was a typo +++

I hope that I may send questions if I find portions of the draft which I do
not understand.  At the moment I am able to make sufficient progress.
I would be interested in corresponding with other persons working on (or
having interest in) demonstration or reference implementations of iSCSI.


+++ write to the list and we all hope to get soon to a interoperability
session +++


Thanks,
Nick
------------------------------------------------------------------
Nick Martin M150801 Rm158A52
Systems Engineer
Server Storage Products
Compaq Computer Corporation
P.O. Box 692000
20555 State Highway 249
Houston, TX 77269-2000
email: Nick.Martin@compaq.com
voice: (281)514-2793
pager: (713)762-7153
fax:   (281)514-5270