Obsoleting QAS in SPI-4

Elliott, Robert (Hou) Robert.Elliott at COMPAQ.com
Tue May 2 17:21:20 PDT 2000

* From the T10 Reflector (t10 at t10.org), posted by:
* "Elliott, Robert (Hou)" <Robert.Elliott at compaq.com>
> * From the T10 Reflector (t10 at t10.org), posted by:
> * Mike Kosco <Mike at MVBuilders.com>
> *
> If we obsolete QAS in SPI-4, SPI-4 devices will suffer the 
> same fairness fate as the pre-SPI-3 devices.  All SPI-1,
> SPI-2 and SPI-4 devices will have to live on the "dirty"
> end of the fairness stick.  I don't believe
> there is a rule that says all devices on the bus must be QAS 
> in order to enable it.

That's effectively what this text in SPI-3 section
10.5.3 requires:
	In an environment where some SCSI devices have QAS
	enabled and other SCSI devices do not, it is possible
	for the SCSI devices that have QAS enabled to prevent
	SCSI devices that do not have QAS enabled from
	arbitrating for the bus. This occurs when SCSI devices
	that have QAS enabled never go to a BUS FREE phase.

	A QAS initiator may interrupt a sequence of QAS cycles
	to force a normal arbitration with the following procedure:

If you want fair arbitration, you can't mix QAS and non-QAS
devices.  The escape sequence was provided to let initiators
occasionally poll for newly hot-plugged devices that powered
up with QAS disabled.  T10 rejected more complicated algorithms
to try to provide fairness across the arbitration domains.

The fairness algorithm will continue to be in SPI-4.  SPI-3
devices were the first to start implementing it en mass.

> I do agree that obsoleting QAS simplifies design and testing. 
>  Has anyone shipped devices with QAS support?

I'm not aware of any.  Since QAS could not be used without
packetized, it was harder to deploy.  This makes now 
an opportune time to obsolete it.

PC: Robert.Elliott at compaq.com
UNIX: relliott at hobbit.eng.hou.compaq.com

* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org

More information about the T10 mailing list