96-277r1 - Proposed Change in QErr for SPC-2

ROWEBER at acm.org ROWEBER at acm.org
Tue Feb 18 13:59:33 PST 1997

* From the SCSI Reflector (scsi at symbios.com), posted by:
* ROWEBER at acm.org
I was going to weight-in with a lengthy discussion of the several QErr
questions posed in reflector responses to the revised proposal.  However,
while I was working on the response, Ed Gardner and Gerry Houlder covered
the pros and cons more eloquently and thoroughly.

Several Symbios colleagues and I strongly support amending the Basic
Queuing definition to use QErr=11b.  We believe that QErr=00b and QErr=11b
are the only functional choices for a multi-initiator environment, which by
the way, is identical to a dual controller RAID environment, at least in so
far as the real disks are concerned.  We are convinced that the use of
QErr=01b in defining Basic Queuing would be the only thing that prevents
using Basic Queuing in a multi-initiator environment.  Our support for the
change to QErr=11b is based on these observations.

While I cannot speak to the difficulties of disk microcode with the Gerry's
expertise, there is a great temptation to believe that the benefits of
eliminating ordered queuing and write ordering management are at least an
order of magnitude more valuable than the benefits of choosing QErr=01b
over QErr=11b.  If viewed in the context of the classic Jim McGrath 'this
behavior is too hard to verify' argument against Full Queuing, one would
have to recognize that verifying QErr=11b behavior is much easier than
verifying ordered queuing behavior.

Beyond all this standardize stuff, I cannot avoid the belief that, if Basic
Queuing is as cost effective as is claimed, then some disk manufactures are
going to be asked to provide microcode that does Basic Queuing with the
QErr=11b behavior.  This kind of 1+ to the standard seems inevitable, if
the costs and the desired usages actually are as advertised in this debate.
This issue looks very much like a place where a standard will not legislate
industry behavior.

See you all in San Diego.



Reference key:

QErr=00b means abort no tasks
QErr=01b means abort all tasks
QErr=11b means abort all tasks belonging to the initiator to which
     the CHECK CONDITION was sent
* For SCSI Reflector information, send a message with
* 'info scsi' (no quotes) in the message body to majordomo at symbios.com

More information about the T10 mailing list