Re: iSCSI draft 02: digests

[Stephen Bailey <steph@cs.uchicago.edu>]

> Like on a data failure on any bus.  Raise a check condition and end
> the command with an error but let it go up to the normal end.
> I will spec it.

A data failure on any bus is not necessarily signalled by a SCSI check
condition.

The host implementations I have seen indicate this through a different
mechanism which reports a (kill me if I get this wrong, so I'll
remember the next time) SAM `service delivery failure'.

Three components of status are traditionally returned: 1) service
delivery status 2) SCSI status byte 3) SCSI sense.  SCSI status is
only valid if the service delivery status implies that it is.

I'm hugely against the service delivery subsystem cooking up SCSI
status to report its own errors.  There should be a direct
correspondence between SCSI status from the target, and returned SCSI
status.  In any case where SCSI status from the target can not be
returned, the chosen service delivery status should imply this.

iSCSI only has to specify that appropriate service delivery status is
returned in this case.

Steph