Questions on XDWrite and XDRead commands

Jay Elrod Jay_Elrod at notes.seagate.com
Wed Jun 4 09:15:11 PDT 1997


* From the SCSI Reflector (scsi at symbios.com), posted by:
* Jay Elrod <Jay_Elrod at notes.seagate.com>
*
Hi Rick,

In answer to your first question, be creative.  Resources can be consumed
rapidly by irresponsible hosts.  Don't forget about the "System Buffer Full"
ASC/ASCQ pair.

Regarding the overlapping XDWRITEs, each XDWRITE must have an
associated XDREAD.  The XDREAD sent for the first XDWRITE would
see the data for that XDWRITE, and the XDREAD sent for the second
XDWRITE would see the data for that one.  The example of two XDWRITEs
followed by only one XDREAD is not a valid example.

>  In reviewing the XDWrite and XDRead commands in the SCSI-3 working draft,
>  I have come across the following problems.
>
>  If an initiator is using the XOR commands at queue depth, the drive may 
>  run out of resources fairly quickly if no XDRead commands are issued.  This 
>  is due to the fact that XDWrite commands result in locking of cache segments
>  for retention of XOR result data.
>
>  Is there an assumption that most (all?) initiator will reserve the unit
>  or used linked commands to make the XDWrite/XDRead pair atomic?
>
>  In addition, if two overlapping XDWrites are issues and then followed by
>  an XDRead then what XOR result data should be returned?  Is this 
>  undefined?
>
>      Thanks,
>      Rick
>      rroche at tdh.qntm.com
*
* For SCSI Reflector information, send a message with
* 'info scsi' (no quotes) in the message body to majordomo at symbios.com




More information about the T10 mailing list