SAT: Comments on 05-177r0 "write same"

James.C.Hatfield at seagate.com James.C.Hatfield at seagate.com
Mon May 2 16:08:48 PDT 2005


* From the T10 Reflector (t10 at t10.org), posted by:
* James.C.Hatfield at seagate.com
*
I am unable to attend the SAT meeting this week, but would like to comment
on
SAT proposal 05-177r0 regarding 'write same' translation.

I would greatly appreciate a reply (separate from the minutes) from the SAT
group regarding
the outcome of the discussion of this proposal.

Thank You !!!
-----------------------------------------------------------------
Jim Hatfield
Seagate Technology LLC
   e-mail:  James.C.Hatfield at seagate.com
   s-mail:  389 Disc Drive;  Longmont, CO 80503 USA
   voice:  720-684-2120
   fax....:  720-684-2711
==========================================


field       comment
----------------------
------------------------------------------------------------------------
OPCODE      Is there a preferred hierarchy of choices for implementation ?
            If a device does not support SCT, is there a preference
                  for the other 2 choices ?

            Is it worth putting "(NCQ)" next to the term "WRITE FPDMA
QUEUED" ?

            Please clarify the meaning of "device implements SCT".
            a)  What if the ATA device (target) 'supports' SCT, but does
not
                  have it (or the LBA Segment Access subcommand) enabled ?

            Please change the SCT reference
                  from  Long Segment Access
                  to    LBA Segment Access

            Please add a reference to the SCT technical report to the
            'References under development' section of the main SAT
document.

WRPROTECT
            ATA devices typically do not support protection information.

            This field is correctly marked as U = Unimplemented

PBDATA and LBDATA
            ATA devices typically do not support protection information.

            Please mark these fields as U = Unimplemented

GROUP NUMBER
            This field is correctly marked as U = Unimplemented

NUMBER OF BLOCKS
            1)    Mark this field as E = Emulated

            2)    "...If the NUMBER OF BLOCKS is not zero, the SATL shall
                  write the number of sectors specified to the device."

                  Please change the wording to clarify:
                  a) The SATL will only write ONE block to the device,
                        as sent to it by the initiator
                  b)  This field indicates how many blocks starting from
                        LOGICAL BLOCK ADDRESS are to be written
                        with the single sector of data from the initiator.

                  This is consistent with the SCT LBA Segment Acccess
commmand,
                  and with the obsolete ATA 'write same' command.

            3)    ..."if the NUMBER OF BLOCKS field is greater ..."

                  a) This seems to be a very awkward way of saying that if
                        the LOGICAL BLOCK ADDRESS plus the NUMBER OF BLOCKS
                        is greater than the largest LBA supported by the
device
                        then the SATL may terminate the command

                        Can some better wording be chosen ?

                  b)    Watch for devices that support 48-bit LBAs, but do
not have
                        48-bit mode enabled, and for devices that do not
support 48-bit LBAs at all.

                  c)    "... the SATL may terminate..."

                        i)   How about something more deterministic like
'shall' ?

                        ii)  or is it the responsibility of the ATA device
to check itself and
                              terminate a bad request from the SATL ?


=======================================================================


*
* 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