WORRIES ABOUT OPTIONAL READ(6) AND RESERVE(6)

GFRAZIER at AUSVM6.VNET.IBM.COM GFRAZIER at AUSVM6.VNET.IBM.COM
Fri Aug 16 05:31:39 PDT 1996


* From the SCSI Reflector, posted by:
* GFRAZIER at AUSVM6.VNET.IBM.COM
*
As you recall, Reserve(6), Release (6), and Read (6) were made optional at
the last SCSI meeting. Our system programmers have been investigating the
impact of this, and they think it is a disaster!
They are most concerned about Read (6) since existing systems contain
IPL ROM which depends on READ(6). Customers with these systems can no
longer be sure that they can upgrade their systems with SCSI-3 drives!
They are concerned about Reserve (6) for the same reason, but at least
in this case it is not an IPL ROM problem--only a device driver problem.

Here is a bit of background from one of the programmers:

Our systems contain IPL ROM which won't work without Read (6). Also,
remember that systems must boot from SCSI-1, SCSI-2, and SCSI-3 devices,
and many SCSI-1 devices did not support READ(10). Therefore, an IPL ROM
update that uses READ (10) exclusively can't be used. Even if it could be
used, upgrading IPL ROM in the field is a very expensive proposition.

The problem with removing Reserve (6) is not an IPL ROM problem but it is
still serious. Remember that a driver must support SCSI-1, SCSI-2, and
SCSI-3 devices. Since previous revisions of SCSI-3 did not make Reserve
(10) mandatory, and since some device vendors already claim SCSI-3 in the
Inquiry Data, the ANSI level in standard inquiry data can not be used to
reliably determine if the SCSI-3 device supports RESERVE(10).  Instead, the
device driver will have to make a guess based on ANSI level as to which
Reserve it should use. If it chooses RESERVE(10) (for the ANSI level of 3),
then it needs additional logic to issue RESERVE(6) if it gets an illegal
request from the RESERVE (10). If it chooses RESERVE(6), then it needs to
have logic to issue RESERVE(10) for illegal request. Thus these standard
changes do not lend themselves well to clean interfaces in software.


Aren't there other systems out there with the same problems and the same
concerns? If you aren't worried, we'd like to know why and maybe we won't
worry either.

Thank you.

Giles Frazier
Jim Allen
IBM Austin
gfrazier at austin.ibm.com
jpallen at austin.ibm.com




More information about the T10 mailing list