SCSI to ATA pitfalls

Sheffield, Robert L robert.l.sheffield at intel.com
Fri May 14 11:41:23 PDT 2004


* From the T10 Reflector (t10 at t10.org), posted by:
* "Sheffield, Robert L" <robert.l.sheffield at intel.com>
*
If there weren't a problem to be solved, there wouldn't be much purpose
in having the working group.
I don't know how successful we'll be either; but perhaps, at least in
some measure, we can help future implementers spend less time figuring
out what SCSI/ATA Translation means, and more time getting it right. You
brought up some very good places to start.

Bob Sheffield

-----Original Message-----
From: owner-t10 at t10.org [mailto:owner-t10 at t10.org] On Behalf Of Pat
LaVarre
Sent: Friday, May 14, 2004 9:06 AM
To: t10 at t10.org
Subject: SCSI to ATA pitfalls


* From the T10 Reflector (t10 at t10.org), posted by:
* Pat LaVarre <p.lavarre at ieee.org>
*
> Re: June-4 SCSI/ATA Translation (SAT) Teleconference

I remain curious to know how well this effort to connect an ATA drive as
if it were SCSI will avoid the usual design pitfalls, which I once
nutshelled as follows, to some acclaim offline.

Pat LaVarre
http://plavarre.blog-city.com/read/575770.htm

-----Forwarded Message-----
From: Pat LaVarre <p.lavarre at ieee.org>
To: ...
Cc: ..., forum at t13.org, t10 at t10.org 
Subject: RE: [t13] FYI: T10 news: SCSI to ATA Translation Study
Groupmeetin g
Date: 16 Apr 2004 17:06:46 -0600
...

> As far as I can tell all of the translation is very modular and
> should happen transparently to other devices.

Yes people sell bridges to IDE by claiming the theory that the bridge is
naturally so trivial and transparent it needs no testing.

In practice, people actually get the thirteen cases wrong, misaligned
address and/or length wrong, additional lengths wrong, x00 meaning x100
wrong, etc. etc.  The so-called transparent connection often actually
chokes over much of anything beyond read/ write of a few blocks: e.g.
zero blocks, beyond x100 blocks, firmware update, rapid disk erase, disk
passwords, etc.  From time to time people even get more elemental basics
like op x12 Inquiry and op x25 Read Capacity wrong, just because some
hosts in fact do tolerate trouble there, especially in op x12 Inquiry
for zero bytes.

Last soft bridge to ATA [from SCSI that I checked]
behaved bizarrely in the basic basic test case of
letting the cdb[4] Allocation Length of op x12 Inquiry fall out of sync
with the expected length of data - the parameter that in USB would
become dCBWDataTransferLength.

The products that ship with those kinds of errors only work enough less
often than everything else to cause a rash of minor incompatibilities.

Whether writing more English will help I'm not sure.

Contemporaneous as-yet-ineffective efforts include the usb.org msc
committee bco (binary-code-only) Compliance suite and the usb.org msc
committee English-only redefinition of "bootable" SCSI.

Broader distribution of open source compliance test suites would
actually help, I think.

Binary-code-only can do some good of course.  Plugfests designed merely
to plug new devices into more than just the latest Windows would help. 
Just plain plugging into Linux, Mac, Win ME/ 98/ 95 etc. often demoes
how unreasonably a device designed for Win XP actually behaves.

Pat LaVarre


*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org
*
* 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