ACA bit = 0 mandatory?

Jeff Williams jlw at
Thu Sep 22 06:54:19 PDT 1994

> A deferred error should should cause the device to stop executing any commands
> from the command queue until the "autosense" function occurs. This means the
> target must find a new command from the same initiator that still hasn't
> reported status (a pending command could satisfy this, if there is one) and
> terminate it with CHECK status and perform autosense. During this period, the
> device must report ACA ACTIVE or BUSY status to commands from other initiators
> This could be a much longer interval than portrayed at the meeting. If we now
> have a "SCSI-3" version of ACA=0 as well as a "SCSI-2" version of ACA=0, 
> I hope a description of deferred error handling will occur in the "SCSI-3" 
> description of it, at least.

The discussion of deferred errors should not be mentioned with the autoclearing
of ACA.  They are distinct items.  ACA does not change the manner in which the
deferred errors are reported or cleared.  An ACA does not exist until the 
CHECK is returned to the initiator.  At that time, the ACA is treated as Bob's
proposal states.  The deferred error handling of new I/O and old I/O is 
separate and should not be combined with ACA since the time domains are 

              Error (deferred occurs)  \
                                        |  Deferred handling of other I/O
                                        |  as per SCSI-2 (rev 10l)
              Error is reported        /   
              Some error handling       -  ACA handling as per Bob's proposal
                                           and SAM.

Jeffrey L. Williams                                HP TelNet: 396-5030
Disk Controllers Lab             	 Telephone: (1) (208) 396-5030
Disk Memory Division                     Facsimile: (1) (208) 396-6858
Hewlett-Packard Co.                    ARPANET: jlw at

More information about the T10 mailing list