RE: iSCSI: Logical Unit Reset

[Eddy Quicksall <eddy_quicksall@ivivity.com>]

I agree ... so what I said is correct; right? 

i.e., the statement in the spec is slightly incorrect in that it says "Task
management requests must act on all the commands having a CmdSN lower than
the task management CmdSN". That is not true because other sessions' CmdSN's
have no relationship to the issuing sessions commands.

Eddy

-----Original Message-----
From: Tony Battersby [mailto:tonyb@cybernetics.com]
Sent: Tuesday, October 08, 2002 1:02 PM
To: 'Eddy Quicksall'; ips@ece.cmu.edu
Subject: RE: iSCSI: Logical Unit Reset


> -----Original Message-----
> From: owner-ips@ece.cmu.edu [mailto:owner-ips@ece.cmu.edu]On Behalf Of
> Eddy Quicksall
> Sent: Tuesday, October 08, 2002 10:08 AM
> To: ips@ece.cmu.edu
> Subject: iSCSI: Logical Unit Reset
>
>
> Section 9.5.1 says:
>
> Task management requests must act on all the commands having
> a CmdSN lower
> than the task management CmdSN.
>
>
> But a Logical Unit Reset, Target [WARM | COLD] Reset and
> Clear Task Set
> w/TST=0 applies to all initiators for the Target or Target/LUN.
>
> Since the initiator that issued the TMF knows nothing about
> the CmdSN's of
> other initiators, how is this handled for the other sessions?
>
> One thing that can be done is to consider that the session
> that issued the
> TMF is synchronized only with itself and therefore its TMF
> would apply to
> all CmdSN's on the other sessions. A statement to this effect should
> probably be added to 9.5.1.

The initiator that issued the TMF is the only initiator that knows about the
TMF.  When a TMF aborts/clears a set of commands, the issuing iSCSI
initiator knows not to expect any more replies for the aborted commands.
Other initiators do not know that a TMF was executed, and therefore still
expect iSCSI replies for all the commands they have outstanding.  Therefore,
the iSCSI layer should NOT apply a TMF to commands from other initiators.
If need be, the SCSI layer can take care of aborting commands from other
initiators.  For example, when processing a Logical Unit Reset, the SCSI
layer can abort any outstanding SCSI commands from other initiators with
CHECK CONDITION / UNIT ATTENTION.

In other words, it is the SCSI layer rather than the iSCSI layer that
applies the TMF to commands from other initiators.

The Target Cold Reset is an exception to this, since the target terminates
all TCP connections to all initiators, effectively informing other
initiators that all outstanding commands were aborted.

Anthony J. Battersby
Cybernetics