[t13] T10/04-262r0

Jeff Garzik jgarzik at pobox.com
Thu Aug 12 21:57:03 PDT 2004

* From the T10 Reflector (t10 at t10.org), posted by:
* Jeff Garzik <jgarzik at pobox.com>

(please read my reply timestamped 00:48am EST, before this one)

Curtis Stevens wrote:
> 	One thing to keep in mind is that we do not want the bridge to look
> at the command code or any other register.  We want SCSI to be a blind
> tunnel to the ATA device.  The bridge should not look at the command code,
> sector count, or any other part of the register set.  This is what makes it
> work for standard and vendor specific commands.  It also makes it so new
> commands can be implemented without changing the bridge.  This is why length
> is also important.

I do agree 100% with this.

I thought of an additional point, relating to this, while I was driving 
through the rain tonight (hello Hurricane Bonnie):

One of the key values of SAT to my user base is the leverage of 
_existing_ SCSI application software.  Particularly, I wish to use this 
ATA passthru CDB via pre-existing SCSI system call methods (ioctl).

As such, there is no opportunity for me to add an ATA-specific 
command-protocol field to Linux's submit-a-SCSI-command system call.  I 
have only the SCSI CDB, transfer length, and transfer direction to work 
with.  Any ATA specifics _must_ be encapsulated within the ATA passthru CDB.

I think my need fits nicely with your "blind SCSI tunnel to ATA device" 
bridge model.


* 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