Re: iSCSI : NOP-OUT & NOP-IN issues.

[julian_satran@il.ibm.com]

My reply in the text - thanks, Julo

Santosh Rao <santoshr@cup.hp.com> on 07/01/2001 06:19:31

Please respond to Santosh Rao <santoshr@cup.hp.com>

To:   IPS Reflector <ips@ece.cmu.edu>
cc:
Subject:  iSCSI : NOP-OUT & NOP-IN issues.




Julian,

I have several concerns regarding the current semantics for the NOP
operation, as defined in the lastest iSCSI draft. (iSCSI-02.txt) :

Section 2.12 NOP-OUT PDU
=========================
1) Quoting from the draft :
"The NOP-Out can be sent by an initiator because of a NOP-In with the
 poll bit set, in which case the Target Tag will copy the NOP-In value."

There is no poll bit in the NOP-IN PDU. What is the intent of the
above ?


<js> Bit 7 of the second byte (it was a typo) should read P - it is fixed
in 03.txt
</js>

2) Quoting from the draft :
"The NOP-Out MUST have the Initiator Task Tag set only if the P bit is
 one or the DataRN field is set."

There is no DataRN field in the NOP-OUT PDU. Is the intent of this to
state
that :
"The NOP-OUT must have the Initiator Task Tag set if the NOP-OUT is
being
used to acknowledge DataRN[s] from received READ Data PDU[s]" ?
Some clarification would help here.

<js> read ExpDataRN.  Correction and an explanation for the field have been
added
</js>

3) Quoting from the draft :
"The NOP-Out MUST have the Target Tag set only if it issued in
 response to a NOP-In with the P bit one, in which case it copies the
 Target Tag from the NOP-In PDU."

Some comments on the above :
i) There is no P bit in a NOP-IN PDU.

ii) If a target wanted to originate a NOP for some reason, it should use

NOP-OUT. The NOP-OUT should be intended for the originator of the NOP
and the
responder of the NOP should reply with NOP-IN.
This keeps the NOP semantics simple and intuitive.

ii) Why is there a need for a target to originate a NOP-OUT ? In the
interests
of simplicity, targets should be prohibited from originating the NOP
operation.

iii) With the above semantics, how is an initator to know if a received
NOP-IN
is being originated by a target [to which it needs to respond with a
NOP-OUT]
or the NOP-IN is being received in response to a transmitted NOP-OUT ?

<js> In and Out are always relative to the initiator.
A receiver of a NOP knows if the NOP is originated by the target through
the target task tag.
The code differentiation is meant for a stateless protocol analyzer
<js>

The semantics of NOP-OUT used by the originator of the NOP operation and

NOP-IN used by the responder of the NOP operation seem more intuitive.

Section 2.13. NOP-IN PDU
=========================
1) References to a P bit in the NOP-IN PDU which does not exist.

2) This section specifies the use of 0 for Initiator Task Tag when the
target
originates a NOP operation. 0 is a valid value for Initiator Task Tag
and its
usage can confuse the initiator b/n a target originated NOP-IN and a
target
response to NOP-OUT sent on an initiator task tag of 0.
The reserved value for Initiator Task Tag should be 0xFFFFFFFF, as is
the case
with the Target Task Tag.

Thanks,
Santosh

<js> corections where made - this typo was reported by several people </js>