RBC delays op x12 Inquiry?

Pat LaVarre LAVARRE at iomega.com
Wed Mar 13 08:17:00 PST 2002


* From the T10 Reflector (t10 at t10.org), posted by:
* "Pat LaVarre" <LAVARRE at iomega.com>
*
> paul.vonbehren at sun.com 03/12/02 02:38PM >>>

> Look at SPC "Model common to all devices types" discussion
> of Inquiry and Test Unit Ready (1 paragraph on each).

I trust you mean to refer to the two paragraphs I quote from spcr305.pdf
below?

> seems to have enough info to suggest
> that Test Unit Ready response should be immediate
> and Inquiry may take longer if it relies on the underlying media.

Hmmm.

Are you saying that Scsi has always let Inquiry be delayed significantly, and
that Rbc merely helps make that fact plainer by striking the almost-empty
phrase "the standard INQUIRY data should be available without incurring any
media access delays" from the description of op x12 Inquiry?

If I'm permitted to remark on your choice of ISP ... may we conclude that Sun
shares with Apple the culture that says life should begin with op x00
TestUnitReady, visibly distinct from the Microsoft culture that says life
should begin with op x12 Inquiry?  I don't have good data to back up this
conjecture, it's just an impression to which this latest evidence contributes
weakly.

Curiously yours, thanks again for all the links, Pat LaVarre
<http://members.aol.com/plscsi/>



>>> spc3r05.pdf (i.e. Revision 05, 9 March 2002)

5 Model common to all device types
...
5.1 Introduction to the model common to all device types
...

5.2.2 Using the INQUIRY command
The INQUIRY command may be used by an application client to determine the
configuration of the logical unit.  Device servers respond with information
that includes their type and standard version and may include the vendor's
identification, model number and other information.  It is recommended that
device servers be capable of returning this information (or whatever part of
it that is available) upon completing power-on initialization. A device server
may take longer to get certain portions of this information, especially if it
retrieves the information from the medium.

...

5.2.4 Using the TEST UNIT READY command
The TEST UNIT READY command allows an application client to poll a logical
unit until it is ready without the need to allocate space for returned data. 
The TEST UNIT READY command may be used to check the media status of logical
units with removable media.  Device servers should respond promptly to
indicate the current status of the SCSI device, delays to achieve GOOD status
may adversely affect initiator performance.

...



>>> Paul von Behren 03/12/02 02:38PM >>>
* From the T10 Reflector (t10 at t10.org), posted by:
* Paul von Behren <paul.vonbehren at sun.com>
*
Look at SPC "Model common to all devices types" discussion
of Inquiry and Test Unit Ready (1 paragraph on each).  It
seems to have enough info to suggest that Test Unit Ready 
response should be immediate and Inquiry may take longer
if it relies on the underlying media.

Paul

Pat LaVarre wrote:
> 
> * From the T10 Reflector (t10 at t10.org), posted by:
> * "Pat LaVarre" <LAVARRE at iomega.com>
> *
> > > I see RBC strikes the phrase "the standard INQUIRY
> > > data should be available without incurring any media
> > > access delays" from the description of op x12.
> ...
> > all the force and majesty of a "should"
> 
> Offline, thank you again, I was kindly supplied with the anonymised
explanation quoted below of when we may see -x 12 0 0 0 FF 0 -i xFF (Inquiry
for up to xFF bytes) take significantly longer to complete than the Windows
standard -x 12 0 0 0 24 0 -i x24 (Inquiry for up to x24 bytes).
> 
> Me, in bus traces, I see the "client application program[s] ... at the
initiator" vary in whether they begin life with op x00 TestUnitReady or op x12
Inquiry.
> 
> I'd say devices that make op x12 Inquiry slow are in effect voting to begin
life with op x00 TestUnitReady, in a de facto standard noone's bothered to
write down.
> 
> Pat LaVarre
> 
> >>> [anonymised] 03/01/02 01:22PM >>>
> ... there is serial number data (for example) in some of the inquiry
payloads - in vendor-unique areas (...?).  Thus, to get this data, they had to
spin up hda's before responding to the inquiry. The spin up can sometimes take
a few seconds to tens of seconds - and some initiators timed out on this basic
command usually used at probe time.   So the recommendation is that if you
asked for only enough inquiry data  to cover the standard data, then it would
be a near-immediate response.
> ...
> 
> >>> Peter Johansson 03/01/02 11:07AM >>>
> * From the T10 Reflector (t10 at t10.org), posted by:
> * Peter Johansson <PJohansson at ACM.org>
> *
> At 10:36 AM 3/1/2002 -0700, Pat LaVarre wrote:
> 
> >I see RBC strikes the phrase "the standard INQUIRY data should be
> >available without incurring any media access delays" from the description
> >of op x12.
> 
> I go out on a bit of a limb to say this, Pat, since I don't have detailed
> notes from the RBC meetings. But I suspect it was eliminated in the spirit
> that it didn't have much to say in the first place. That is, it has all the
> force and majesty of a "should" ...
> 
> Another perspective is that it's not really a standards matter: if it's not
> sufficiently important to rise to the level of a "shall", market forces
> will determine whether the successful devices return as much data as they
> can as soon as they can---or not.
> 
> To put it another way, if you are designing a client application program to
> execute at the initiator, what possible difference would this
> RECOMMENDATION make to your design?
...

*
* 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