[t13] 04-136r0.pdf SCSI to ATA Command Translations

Pat LaVarre p.lavarre at ieee.org
Fri Jun 4 10:34:35 PDT 2004


This message is from the T13 list server.


> 04-136r0.pdf
> SCSI to ATA Command Translations
> April 27, 2004

Hello, I'm catching up now.

Hopefully the later drafts already cover what I discuss below, but if
you're curious, you can check now.

Pat LaVarre

---- ---- ----

> 1. OVERVIEW
> ...
> Unless otherwise noted,
> the IMMED bit (immediate return) is ignored.
> For the format unit command, this bit will be supported.

Why not IMMED op x1B "Start Stop Unit"?

Do we mean to be requiring spin up/ down to complete in, maybe, three
seconds?

> 3.3 Test Unit Ready Command (00h)
> 3.3.4 Miscellaneous Notes.
> ... at standby ...
> logical unit not ready, initializing command required.

Ah, x 2 04 02, reportedly seen in Iomega jaz 2GB etc.

Less than all hosts liked that.  When spin up/ down is quick, tradition
tells me more hosts like reporting ready whenever media has been spun up
well enough to know where each LBA is physically.

> 2. SUMMARY OF COMMAND MAPPINGS
> ...
> Read (6,10,12)  08h, 28h, A8h      X  3.10
> Write (6,10,12)  0Ah, 2Ah, AAh  Multiple     X  3.11
> Write and Verify (10)  2Eh  Multiple     X  3.12
> Write and Verify (12,16)  AEh, 8Eh  N/A      Not Supported

Seeing us mention x8E "Write and Verify (16)" makes me wonder,

Why we did not mention op x88 Read(16) and op x8A Write(16)?

We mean for our barrier of death to be the 48 bit/LBA barrier designed
into ATA, yes?  Not 32 or 28?  To report a 48-bit max LBA, we'll need to
report a 64-bit max LBA, up past the 32 bits of 2 TiB / 0.5 KiB/LBA.

> Write and Verify (10)  2Eh  Multiple     X  3.12
> Write and Verify (12,16)  AEh, 8Eh  N/A      Not Supported

Why support only the (10) and not the trivial translation between that
and the (12) and (16) flavours?

> Write Long  3Fh  N/A      Not Supported
> Read Long (10)  3Eh  N/A      Not Supported

Why not?  How do we inject read errors?

> Prevent Allow Medium Removal  1Eh  N/A      Not Supported

Why not?  ATA has an RMB bit too?

> 1. OVERVIEW
> ...
> All ATA commands with the exception of
> SATA Native Command Queued Reads and Writes
> are single threaded per device.
> The translator will queue IOs as necessary
> to enforce this.

Hmmm.

Pat LaVarre





More information about the T10 mailing list