Message-ID: <005b01c32710$ec29d3b0$b76e030f@rose.hp.com>
From: "Mallikarjun C." <cbm@rose.hp.com>
To: "Lakshmi Ramasubramanian" <nramas@windows.microsoft.com>,
   "Julian Satran" <Julian_Satran@il.ibm.com>, <ips@ece.cmu.edu>
References: <DDE1793D7266AD488BB4F5E8D38EACB801193E98@WIN-MSG-10.wingroup.windeploy.ntdev.microsoft.com>
Subject: Re: iSCSI: Plugging holes in CmdSN
Date: Fri, 30 May 2003 18:06:00 -0700
MIME-Version: 1.0
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
Sender: owner-ips@ece.cmu.edu
Precedence: bulk

Lakshmi,

Comments below.
--
Mallikarjun

Mallikarjun Chadalapaka
Networked Storage Architecture
Network Storage Solutions
Hewlett-Packard MS 5668 
Roseville CA 95747
cbm@rose.hp.com

----- Original Message ----- 
From: "Lakshmi Ramasubramanian" <nramas@windows.microsoft.com>
To: "Julian Satran" <Julian_Satran@il.ibm.com>; <ips@ece.cmu.edu>
Sent: Friday, May 30, 2003 12:59 PM
Subject: iSCSI: Plugging holes in CmdSN


>If the target detects data digest error in a PDU,
>and sends Reject the initiator 

I assume you mean the SCSI command PDU above.

> 
> * MAY retry the command with the same CmdSN, TaskTag, etc. 
> 
> * Or, Send a Task Mgmt command to ABORT TASK that task.
> 
>Target will not advance ExpCmdSN until initiator does something
>to clear the rejected command.

Yes, one of the two must happen (or the session drop, of course).

> 
>But, can the initiator send a Task Mgmt command with function
>LU Reset (instead of ABORT TASK) as IMMEDIATE command? 

The initiator may.  Because the LU Reset is an immediate command, it will be
processed by the iSCSI layer and then the SCSI layer and an LU Reset
action will result.

>Should 
>the target then clear the rejected command and any command that 
>was received after that? And, advance the ExpCmdSN so that initiator
>can proceed with that session?

I am not sure what you mean by "clear the rejected command".  

As for advancing ExpCmdSN, I don't think that is the legal behavior.  Unless
the target has a RefCmdSN to plug a hole, it cannot plug a CmdSN hole.
And only the Abort Task TMF has the RefCmdSN.  The iSCSI layer on
target cannot afford to simply plug all the CmdSN holes with an LU Reset 
because there may be commands destined to other LUs already in the CmdSN 
reordering queue.  Also, the iSCSI layer on the target does not look at the 
LUN field of each task - as a design rule.