RE: iSCSI: task reassign of TMF

["Julian Satran" <Julian_Satran@il.ibm.com>]

You have at least to know that you got
back from the SCSI layer whatever you sent out there.
You might choose not too but that is
probably not the safest thing to do as you will want to impose some time-limits
to every activity at least lower than host SCSI limits.

Julo

"Eddy Quicksall" <Eddy@Quicksall.com>

31/12/02 16:26

Please respond to
<Eddy@Quicksall.com>

To	Julian Satran/Haifa/IBM@IBMIL
cc	<ips@ece.cmu.edu>, <owner-ips@ece.cmu.edu>
Subject	RE: iSCSI: task reassign of TMF

Tracking of a TMF is implementation
dependent. Generally, it does not need to be tracked in the iSCSI layer
after it has been passed to the SCSI layer. This because the SCSI layer
or beyond tracks the TMF and when it completes, the context is handed back
to the iSCSI layer.
 
Eddy
-----Original Message-----
From: Julian Satran [mailto:Julian_Satran@il.ibm.com]
Sent: Monday, December 30, 2002 2:18 AM
To: Eddy@Quicksall.com
Cc: ips@ece.cmu.edu; owner-ips@ece.cmu.edu
Subject: Re: iSCSI: task reassign of TMF


Eddy, 

Besides being way to late the text change you request is not getting you
any real advantage for the cases in which the action is in the SCSI layer.
You must track them anyhow as your answer to the TMF must be given after
the SCSI layer has done its job (or pretends it has) and the target is
required not to pass any additional reponses after the TMF response.

Julo

"Eddy Quicksall" <Eddy@Quicksall.com>
Sent by: owner-ips@ece.cmu.edu

27/12/02 21:06

Please respond to
<Eddy@Quicksall.com>

To	<ips@ece.cmu.edu>
cc
Subject	iSCSI: task reassign of TMF

I know it is a little late to bring this up but incase there is another
revision, I thought it would be worthwhile to discuss.
  
Section 10.5.1 says that the target MUST reject a task reassignment of
a TMF. I think it should be up to the initiator to follow the protocol
and up to the target to check that he does. To adhere to this MUST, the
target will have to track all tasks even though they have passed on to
the SCSI layer or require the SCSI layer to understand this rule. If we
want wire speed, we should not require such tracking in the spec ... that
should be an implementation issue. 
  
A well checked out initiator will not violate the protocol and to make
the target do more work for something that will not happen is a waste of
time. The target should just be responsible to protect itself against a
crash. 
  
I suggest the verbiage be changed to "such a request MAY result in
...". 
  
re:

At the target a TASK REASSIGN function request MUST NOT be executed to reassign the connection allegiance of a Task Management function request an active text negotiation task, or a Logout task; such a request MUST result in Task Management response of "Function rejected".

Eddy