RE: Tasks Management Actions on Task Sets (10.6.2)

[Black_David@emc.com]

> I have a question about the descriptions in 10.6.2, "The target."
> 
> Does this special target handling (wait for all affected 
> iscsi data-phases to complete before passing the task mgmt 
> command to the SCSI layer, plus stat-sn acknowledgment) also 
> apply to the LOGICAL UNIT RESET task management command?

No - that special handling is necessary for things to carry
on cleanly after an ABORT or CLEAR TASK SET.  LOGICAL UNIT
RESET is allowed to be somewhat messier, as it generates a Unit
Attention condition that the Initiator can use to clean up on
its side, whereas ABORT and CLEAR TASK SET do not generate their
own Unit Attention conditions.  A target is not prohibited from
doing the things described in 10.6.2, but the RESET has to
happen regardless of the Initiator's cooperation.

> Can I assume that a lun reset acts like a bunch of abort 
> tasks?  (That's the implication I get from SAM.)

No - that's necessary, but not sufficient - see Section 5.9.7
of SAM-2.  RESET really means reset the LUN to its power-on
condition.  It may be used as part of recovery from an
unexpected situation that the Initiator doesn't understand
how it got into, and hence Target dependence on Initiator
actions to complete the RESET may not be a good idea.

A useful thing to keep in mind is that the design rationale
for any RESET operation defined by SCSI is generally "something's
wrong - put the LU/target back in a known state (power-on-like)
so the Initiator can do something useful".  That's not the
only use of RESETs, but it helps explain their intent.

Thanks,
--David
----------------------------------------------------
David L. Black, Senior Technologist
EMC Corporation, 176 South St., Hopkinton, MA  01748
+1 (508) 293-7953             FAX: +1 (508) 293-7786
black_david@emc.com        Mobile: +1 (978) 394-7754
----------------------------------------------------