FW: SCAM and Parity and Targets w/o ID Strings

Steve Gaskill sgaskill at QNTM.COM
Fri Dec 15 16:11:10 PST 1995


* From the SCSI Reflector, posted by:
* <sgaskill at qntm.com (Steve Gaskill)>

Jeff Williams writes:

> 2)  What happens when there are two or more targets on the bus which have 
not spun up when the SCAM isolations have begun and their unique IDs are on 
the drive?  In this case, all of the ID strings for same model targets will 
have identical isolation phases and will all be assigned the same ID by the 
two host bus adapters  I have tested.  Shouldn't the protocol take drives 
without IDs and ignore them if the SNA bit is clear?  The annex states
that the target shall stall the isolation phase if it cannot get an ID 
string.  What if the target is not configured to auto spin up?  Don't we 
have a chicken and egg problem here?  I can't spin up until I get a Start 
Unit command and I can't get a command until I spin up...
>


We've run into the same issue: The "Wait Spin" feature is mutually 
exclusive of SCAM iff the target device has its serial number (or SCAM 
code) stored only on the media.   We'll probably get around this by 
either
 (a) continuing to use onboard non-volatile memory to store the serial 
number, or
 (b) Not responding to the SCAM protocol if our "Wait Spin" feature is 
enabled.

Also, I don't think that SCAM host adapters are really robust yet, 
with respect to handling "abnormal" conditions such as SNA=0...


Steve Gaskill
Quantum Corp
(sgaskill at qntm.com)
 




More information about the T10 mailing list