Task Set Type = "Per Initiator" & QErr = "Abort Tasks From All Initiators"

Ralph Weber ralphoweber at compuserve.com
Fri Nov 23 10:26:50 PST 2001


* From the T10 Reflector (t10 at t10.org), posted by:
* Ralph Weber <ralphoweber at compuserve.com>
*
The following issue may affect only Seagate because
they are the ones with the customers that demanded
creation of TST=001b (Task Set Type = 1 Task Set
Per Initiator).

Suppose you have TST=001b and QErr=01b. Taken at
face value from the SPC-2 definitions for the Control
mode page this configuration means that you have
one task set for each initiator but when a CHECK
CONDITION occurs you abort all tasks from all
initiators.

To be sure, this is a disjoint approach to task
set definitions but I can find nothing to prohibit
it.

Having set the stage, I now turn your attention
to the definition of the transition from dormant
task to enabled task in SAM-2. For ordered tasks,
this definition contains what looks like a specific
mention of the TST=001b/QErr=01b combination that
I think was introduced by the Seagate proposal that 
defined TST=001b.

"Provided an ACA or a CA condition does not exist or
if the TST field contains 001b in the Control mode page
provided the task is not for the faulted initiator and
the QErr field is not 01b in the Control mode page, a
dormant task having the ORDERED attribute shall enter
the enabled task state when all older tasks have ended."

Note that similar wording (at least as regards QErr)
does not appear in the definition for the dormant to
enabled transition for simple tasks.

My interpretation of all this is that for TST=001b
/QErr=01b task sets simple tasks from initiators
other than the faulted initiator become enabled
during a CA or ACA condition for another initiator
while ordered tasks must remain dormant until the
CA or ACA is cleared.

Even my over-active imagination cannot make any
sense of this distinction. So, I am asking the
target manufacturers with customers who use
TST=001b to see what they think should be
happening when TST=001b and QErr=01b.

We will be reviewing the dormant to enabled state 
transition definition line-by-line in January.  
Please being thoughts on this topic to the CAP 
meeting in Houston.

Thanks.

Ralph...


*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list