RE: iSCSI: different question about terminated tasks

[Eddy Quicksall <eddy_quicksall@ivivity.com>]

Here is how I understand ACA:

- one or more commands are in the enabled state
- the commands are using NACA
- one of the commands gets an error and reports a check condition
- the ACA state is set
- the rest of the commands are (1) aborted or (2) blocked depending on mode
page settings
- if aborted, the commands with the NACA bit set each give ACA Active
status; the commands without the NACA bit give a check condition that they
were aborted???
- if a command without the ACA attribute is received while in ACA, it gets
ACA ACTIVE status
- commands with the ACA attribute enter the enabled state (only one at a
time allowed or it gets ACA Active)
- when finished with recovery, the initiator sends a Clear ACA TMF to clear
the ACA state
- if commands were blocked, they now go to the enabled state

Is this correct?

Eddy

-----Original Message-----
From: Mallikarjun C. [mailto:cbm@rose.hp.com]
Sent: Wednesday, January 29, 2003 8:53 PM
To: Bill Studenmund; ips@ece.cmu.edu
Subject: Re: iSCSI: different question about terminated tasks


>My question is what do we do if there are multiple tasks with the
> NACA bit set? Do we initiate ACA for each of them, or only once?

This is not an iSCSI issue (please see SAM2r23, 5.8.1), but I'll offer
just my interpretation of SAM-2.  Also note that there's no formal model
in SAM-x for a SCSI transport layer to notify the SCSI application layer 
of a task termination due to transport dynamics.  The model (if and when 
one is spec'ed) would have to deal with the question of how the notification
affects that task even when the ACA is active (which is the specific
scenario
you're describing).  I read SAM-2 (with the model caveat) as saying that 
there shall be one ACA for each terminated task today, if the task with the 
ACA attribute (the cleanup task) didn't cleanup those to-be-terminated tasks

by then.

> Also, in the case of Logout Request, do we wait for all the tasks to go
> through ACA before sending the Logout Response?

That's not the intent of the "cleanup" wording in 10.15.  Dealing with ACA
is a 
SCSI matter.  The "cleanup" refers to the requirement on the iSCSI layer to
cleanup all the iSCSI-level task and connection resources associated with
the 
iSCSI connection (along with the task termination notifications, whose
after-effects 
SCSI will deal with), before sending a successful Logout Response.

Hope that helps.
--
Mallikarjun

Mallikarjun Chadalapaka
Networked Storage Architecture
Network Storage Solutions
Hewlett-Packard MS 5668 
Roseville CA 95747
cbm@rose.hp.com

----- Original Message ----- 
From: "Bill Studenmund" <wrstuden@wasabisystems.com>
To: <ips@ece.cmu.edu>
Sent: Tuesday, January 28, 2003 4:21 PM
Subject: iSCSI: different question about terminated tasks


> I have a question related to the terminated task thread we had.
> 
> When we terminate SCSI tasks for the reasons mentioned in the thread
> (10.14.5 in draft 20), and they have the NACA bit set, we will establish
> ACA. My question is what do we do if there are multiple tasks with the
> NACA bit set? Do we initiate ACA for each of them, or only once?
> 
> Also, in the case of Logout Request, do we wait for all the tasks to go
> through ACA before sending the Logout Response? The text which encourages
> this question is the first paragraph of 10.15:
> 
>    The logout response is used by the target to indicate if the cleanup
>    operation for the connection(s) has completed.
> 
> 
> Thanks!
> 
> Take care,
> 
> Bill
> 
>