sat-r08c: MODE SENSE (10) -> MODE SELECT (10); REPORT LUNS

Sheffield, Robert L robert.l.sheffield at intel.com
Sun Jul 30 14:28:27 PDT 2006


* From the T10 Reflector (t10 at t10.org), posted by:
* "Sheffield, Robert L" <robert.l.sheffield at intel.com>
*
Doug,
Thank you for your comments.  
I noticed some inconsistency between the descriptions of these commands
in SAT-r08b, and the only changes I made to MODE SELECT and MODE SENSE
were to make the text consistent with what's described in SPC-3.
I understand the ambiguity in the statement, "The SATL shall implement
the MODE SENSE (10) command using the translation described...".
SAT contains a large number of statements along the lines of HOW the
SATL implements a given translation. However, normally there is a
distinction between saying the SATL "shall implement" a given
translation versus HOW a given translation "shall be implemented" if it
is implemented.
With few exceptions, it is generally the intent of SAT to define HOW
translations shall be implemented (if implemented), and not WHICH
translations shall be implemented. Other SCSI commands standards, SPC-3,
SBC-2, SAM-3, define WHICH commands shall be supported by various device
types. SAT specifies that a SATL represents itself as a SCSI
direct-access block device. Therefore, it is reasonable, except where
explicitly stated otherwise in SAT, that a SATL shall support the same
set of commands as defined for any SCSI direct-access block device, as
defined in SBC-2, SPC-3, and SAM-3. As it happens, there are very few
required commands, and MODE SENSE and MODE SELECT are not among the
required commands. However SPC-3 does require that if one is supported,
the other shall be supported as well.
I will raise the subject of the text for the MODE SELECT and MODE SENSE
commands at the SAT WG in August as to whether any clarification in the
text is required.
Regarding REPORT LUNS, again, the objective of SAT in general is to
define HOW commands are translated where appropriate, not WHICH commands
are supported. One can infer based on the fact that a SATL represents
itself as a direct-access block device, that the SPC-3 and SBC-2
requirements that it support REPORT LUNS applies. There is nothing
specific to define regarding a translation for REPORT LUNS, since there
isn't necessarily any interaction with the corresponding ATA device in
processing a REPORT LUNS command. Furthermore, it would be difficult to
define any mapping between the devices reported in REPORT LUNS and the
ATA devices represented without unduly constraining SAT implementations
which might want flexibility to represent various ATA devices as
independent target devices, or as multiple logical units within a common
target device.
So, the general rule is, if we can't find anything useful to define for
a given command translation without establishing undue constraints, then
leave it "unspecified". Accordingly, SAT does not specify a translation
for REPORT LUNS. SAT-2, however, could define something useful for
REPORT LUNS.
Thanks again for your comments.
Regards,
Bob Sheffield
-----Original Message-----
From: Douglas Gilbert [mailto:dougg at torque.net] 
Sent: Friday, July 28, 2006 7:18 PM
To: Sheffield, Robert L
Cc: t10 at t10.org
Subject: sat-r08c: MODE SENSE (10) -> MODE SELECT (10); REPORT LUNS
Robert,
The second paragraph of section 8.6.1 (sat-r08c.pdf) starts
with:
"The SATL shall implement the MODE SENSE (10) command using
the translation described in 8.5. Device servers that
implement the MODE SENSE (10) command shall also implement
the MODE SELECT (10) command."
Since a SATL includes a device server, that implies
both commands shall be implemented. If that is intended
then it could be stated more tersely.
The second quoted sentence from the draft (especially
in the context of the 6 byte variants of MODE SENSE/SELECT)
seems to be too strong. The reversed statement is weaker
and IMO is more sensible, namely: "Device servers that
implement the MODE/LOG SELECT (n) command shall also
implement the MODE/LOG SENSE (n) command".
Since the REPORT LUNS command is mandatory in SPC-3
perhaps table 8 in sat-r08c.pdf (section 7.1) should
convey that (in a note). There is no other reference
to the REPORT LUNS command in the SAT draft.
Doug Gilbert
*
* 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