Queuing and ACA

ROWEBER at acm.org ROWEBER at acm.org
Sun May 10 14:23:16 PDT 1998

* From the T10 (formerly SCSI) Reflector (t10 at symbios.com), posted by:
* ROWEBER at acm.org

Your description involving the NML pile of write commands getting
reordered to NLM as a result of another initiator getting a fault
and putting a queue in ACA condition is completely correct, it
can happen as you describe.  (Forgive me for now reproducing the
description here.)

I don't have a good excuse, except to note that ACA was designed
with single initiator environments in mind.

I can wave my hands wildly and say that setting the newly incorporated
TST field to 001b will get around the problem you describe.  But, I
personally don't believe TST=001b is right for a multi-initiator

A more pleasant way to respond to the problem you describe is to
note that implementing the autosense (data) feature is a better way 
for disk drives to respond to the challenges of a multi-initiator Fiber 
Channel system.  Fortunately, FCP provides for autosense.  Using the 
autosense mechanism, the disk simply puts the sense data in the same 
packet as the CHECK CONDITION status.  They both return to the initiator 
together, and no CA or ACA condition is created at the disk drive.  So, 
no commands ever get blocked or reordered.  It is my understanding that
most (maybe all) FC disks implement autosense.

It also is my understanding that most disks do NOT implement support
for the NACA=1 case in the CDB control byte and therefore do not
set the NormACA bit in the Inquiry data.  But, I do not work for
a disk manufacturer and they generally don't tell me about all the
work they are doing.

Lastly, I agree that setting NACA=1 in the CDB control byte is step
that should be taken only with great care and planning.  Also, my
personal preference would be to set NACA=1 in all CDBs if I set
NACA=1 in any CDBs.  Mixing NACA=1 and NACA=0 is allowed, but the
specific operational aspects of such an modus operandi are poorly
understood, and I think, open to surprises.

Hope this helps.

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

More information about the T10 mailing list