SCSI Document 87-190   Deferred Error Information Bytes

From: Gary R. Stephens, IBM          Date: October 28, 1987

The following proposal closes an outstanding work item I have related to
the report in information bytes when a deferred error occurs in sequential
access devices.

Proposal for addition to REQUEST SENSE, Chapter 7, for handling
Unrecovered Write Errors for Sequential Access devices when operating
in buffered modes greater then zero (0):

Add the following on page 7-28, SCSI-2, Revision 2, to the numbered
list on information bytes as item number four (4):


(4) For an Unrecovered Write Error, for Sequential Access devices,
    when operating in buffered modes greater than zero (0) and 
    unwritten data blocks and/or file marks are in the buffer:

    a) If no commands is active in the LUN at the time the error is
       detected, the residual count in the information bytes
       shall be the number of data blocks and file marks in the buffer.

    b) If a command is active in the LUN, two cases occur:

       1) If the active command is a WRITE command, the residual count
          in the information bytes follows the rules for reporting
          based on the fixed or variable mode of operation
          currently in effect.  See the WRITE command in Section 9.

       2) For all other commands, the residual count in the information
          bytes shall be the number of data blocks and file marks
          in the buffer.

    IMPLEMENTORS NOTE: The READ POSITION command in Section 9 reports 
    both the numbeR of blocks and file marks in the buffer as well 
    as the number of bytes in the buffer.

END OF PROPOSAL

Gary R. Stephens