FW: Proposal to add a new modifier to SCSI Write commands.

[Tom]

* From the SCSI Reflector (scsi at symbios.com), posted by:
* coughlan at star.zko.dec.com (Tom)
*
From:	US2RMC::"milton at arraytech.com" "

Milton Scritsmier wrote:

>I do think it's true that most RAID vendors won't let READ LONG and WRITE
>LONG commands from the host get through to the drives because it makes it
>difficult to guarantee data integrity without a lot of close monitoring of
>the host. It can even be impossible to maintain data integrity if the host
>is not careful enough. In the past we allowed such "pass-through" commands
>and it caused us nothing but grief.

I am not looking for a pass-through function.  

Consider a host-based mirror set comprised of two RAID boxes.  To create
this set we have to make one of the virtual volumes presented by a RAID box
identical to a virtual volume on another RAID box.  If, during this copy
process, we get an unrecovered read error on the "master" volume, then we
need to issue a Write Long to the corresponding block(s) on the "target"
volume.  This is done so that subsequent reads will get the same result (an
error), no matter which volume they happen to be directed to.  

Note that the Write Long is directed to the RAID set, not an individual
disk.  It's up to the RAID controller to manage the virtual to physical
mapping, and to thereby "maintain data integrity". 
*
* For SCSI Reflector information, send a message with
* 'info scsi' (no quotes) in the message body to majordomo at symbios.com