Preservation of Sense Data -Reply

Mon Sep 22 13:39:48 PDT 1997

* From the T10 (formerly SCSI) Reflector (t10 at, posted by:
In response to the T10 (formerly SCSI) Reflector (t10 at,
posted by tdelacru at (Tony DeLaCruz) ...

The most polite devices store the sense data corresponding to an 02h
Check Status separately for each distinct I_T_L nexus, no matter if the
device allows read/write of each I_T_L combination used or not.

Moderately polite devices store separate sense data for each supported
I_T_L combination and any other sense data in one other place.  This
means that a command addressed to an invalid I_T_L can muck up sense
data waiting to be collected via another I_T_L, but that such a command
at least preserves the data waiting for any valid I_T_L.

And indeed, just as you describe, truly rude devices mix up sense data
more often.

If you want to write software that interoperates most widely, you will do
your best to write software that undoes the SCSI design decision that
lets 02h Check status get separated from the associated sense data. 
Auto-sense at the lowest possible layer.

The why is that people who design drive firmware usually care about
how much storage they set aside for freaks of protocol.  This leads them
to regret the design decision that separated the one byte of an 02h
Check Status phase from the bytes of associated sense data.  This
feeling then limits how polite they make their designs.

Hope this helps.   Pat LaVarre x4402
p.lavarre at

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

More information about the T10 mailing list