X3T9.2/88-027 Rev 0

To: X3T9.2 Committee (SCSI)   

From: Gerry Houlder   MPI/CDC

Subject: Proposed wording for MODE SELECT

This proposal is in response to an action item assigned to me at 
the Jan. 6-8 Working Group Meeting.  The proposed wording is 
intended to resolve the "square bracket" notes on pages 7-30 and 
7-41 of the Rev. 3 SCSI-2 draft plus some concerns that came up 
during discussion of how to handle values sent to unsupported 
MODE SELECT parameters.

Insert, as the second paragraph after table 7-24 on page 7-30,
the following paragraph:

"If a target supports saved pages, it may save only one copy of 
the page for each LUN and have it apply to all initiators or it 
may save separate copies for each initiator for each LUN.  If 
separate copies are saved, the target must also maintain separate 
current values for each initiator/LUN combination.  The advantage 
to maintaining separate copies is that each initiator can 
simultaneously use different values for some parameters without 
conflicting with the operation of other initiators.  For some 
pages, such as those describing how the media is formatted, it 
may not make sense to have separate copies for each initiator."

In the next paragraph (currently the second paragraph after table 
7-24), change the first sentence to read as "If an initiator 
sends a MODE SELECT that changes any parameters that apply to 
other initiators, the target may generate a unit attention 
condition for all affected initiators except the one that issued 
the MODE SELECT command (see section 6.1.3)."  I chose to use the 
word "..target MAY generate .." because of input from some 
sequential device makers that object to being required to 
generate unit attention for MODE SELECT changes.  If the working 
group thinks this should be "SHALL", then so be it.

Replace the first paragraph on page 7-31 (the underlined one) 
with the following two paragraphs:

"If the initiator attempts to send an unsupported value to a 
field (including reserved fields) in the MODE SELECT header, 
block descriptor, or any page header, the target shall terminate 
the command with CHECK CONDITION status, set the sense key to 
ILLEGAL REQUEST, and set the additional error code to INVALID 
FIELD IN PARAMETER LIST.

If the initiator attempts to send a value to a mode parameter 
field that is not changeable (as reported by the target in a MODE 
SENSE command requesting changeable values) and that value is 
different from the value returned in a MODE SENSE requesting 
current values, the target may either:
  (1) ignore the new value and complete the command with GOOD 
status.
  (2) terminate the command with CHECK CONDITION status, set the 
sense key to ILLEGAL REQUEST, and set the additional error code 
to INVALID FIELD IN PARAMETER LIST."

Also, the fourth sentence (underlined in Rev. 3) of the second to 
last paragraph on page 7-31 should be changed slightly.  Change 
"A parameter list length that results in truncation of any 
descriptor or header .." to read as "A parameter list length that 
results in truncation of any descriptor, header, or page of 
parameters ..".  This is needed because, if the page length of a 
page is longer than the number of bytes transferred for that 
page, the target will set an error condition rather than accept 
a partial page.  The discussion  of page length (why was the name 
complicated to "page-specific parameter length"?) on page 7-41 
seems to require the initiator to use the length returned by the 
target.  Unfortunately, that discussion doesn't say that the 
number of bytes transferred for that page must match the page 
length (which it should).  This gets into resolving the "square 
bracket" note on page 7-41, as proposed in the next paragraph.

In the 4th paragraph after table 7-34 on page 7-41, the second 
sentence reads "The initiator should set this value ..".  Change 
this to read "The initiator must set this value ..".  The 4th 
sentence reads "The target is permitted to return a parameter 
length value that is less than the full page length,..".  Change 
it to read "The target is permitted to return a page length value 
that is less than the full page length defined in this 
standard,..".  Add a sentence after that 4th sentence that 
reads "The number of bytes transferred must agree with the page-
specific parameter lengths of all pages transferred."

Does anyone object to changing "page-specific parameter length" 
back to "page length"?  I think the shorter name adds clarity and 
the name differentiation may keep the reader from confusing this 
parameter length with the parameter length in the CDB or the 
sense data length or the block descriptor length.  Note that the 
term "page length" is consistent with the naming of "sense data 
length" and "block descriptor length".  Would we want to add the 
word "parameter" to the names of these lengths also?  Let's 
direct the editor to change it back to page length!

The last paragraph on page 7-41 can be deleted, because the new 
paragraphs added at the top of page 7-31 covers the same thing.  
Since this is a MODE SELECT issue, and doesn't apply to MODE 
SENSE, page 7-31 is a better spot for it.  Note that I have not 
carried over the sentence objected to by Larry and Gary in the 
square bracket note.  I think their objection is valid, and the 
target shouldn't be required to 'change back' parameters that 
were successfully changed before a problem was detected.