Seagate Ultra 320 ユーザーズマニュアル

Parallel SCSI Interface Product Manual, Rev. A                                        

   145

The control mode page contains parameters affecting the use of asynchronous event reporting (see Seagate 
SCSI Command Reference Manual, Part number 100293068).

Asynchronous Event Reporting is used to signal a device that one of the four events listed below has occurred:

a. an error condition was encountered after command completion;
b. a newly initialized device is available;
c. some other type of unit attention condition has occurred; or
d. an asynchronous event has occurred.

An example of the first case above occurs in a device that implements a write cache. If the target is unable to 
write cache data to the medium, it may use an asynchronous event report to inform the initiator of the failure.

An example of the second case above is a logical unit that generates an asynchronous event report, following 
a power-on cycle, to notify other SCSI devices that it is ready to accept I/O commands.

Sense data accompanying the report identifies the condition (see Section 7.6.4).

An error condition or unit attention condition shall be reported to a specific initiator once per occurrence of the 
event causing it. The logical unit may choose to use an asynchronous event report or to return CHECK CON-
DITION status on a subsequent command, but not both. Notification of an error condition encountered after 
command completion shall be returned only to the initiator that sent the affected task or tasks.

Asynchronous event reports may be used to notify devices that a system resource has become available. If a 
logical unit uses this method of reporting, the sense key in the AER sense data shall be set to Unit Attention.

Autosense is the automatic return of sense data to the application client coincident with the completion of an 
SCSI command under the conditions described below. The return of sense data in this way is equivalent to an 
explicit command from the application client requesting sense data immediately after being notified that an 
ACA condition has occurred. Inclusion of autosense support in an SCSI protocol standard is optional.

An application client may request autosense service for any SCSI command. If supported by the protocol and 
logical unit and requested by the application client, the device server shall only return sense data in this man-
ner coincident with the completion of a command with a status of Check Condition. After autosense data is 
sent, the sense data and the CA (NACA=0), if any, shall then be cleared. Autosense shall not affect ACA 
(NACA=1), see Section 7.6.1.

Protocol standards that support autosense shall require an autosense implementation to:

a. Notify the logical unit when autosense data has been requested for a command; and
b. Inform the application client when autosense data has been returned upon command completion (see 

It is not an error for the application client to request the automatic return of sense data when autosense is not 
supported by the SCSI protocol or logical unit implementation. If the application client requested the return of 
sense data through the autosense facility and the protocol service layer does not support this feature, then the 
confirmation returned by the initiator's service delivery port should indicate that no sense data was returned. If 
the protocol service layer supports autosense but the logical unit does not, then the target should indicate that 
no sense data was returned. In either case, sense information shall be preserved and the application client 
may issue a command to retrieve it.

An unexpected RESELECTION phase occurs if a SCSI target port attempts to do a physical reconnect to a 
task for which a nexus does not exist. a SCSI initiator port should respond to an unexpected RESELECTION 
phase by sending an ABORT TASK message.