multi scsi questions

Justin T. Gibbs gibbs at caspian.plutotech.com
Wed Jul 7 13:31:40 PDT 1999


> The priority, from highest to lowest, is
>> as follows: 7->0, 15->8.  When a device with an ID lower than 8
>> arbitrates along with a device with an ID greater than 7, the device
>> with the high ID knows that it has lost the arbitration and will get
>> off the bus. 
>But from this seems to follow: If there is a device ID 9 and a narrow
>device on ID 1 (ID 2 was a typo anyway) the wide device ID 9 recognises
>it will loose arbitration and free the bus for an access to device ID 9.
>So it won't work.
>Or does the high device recognise the call for ID 9.

Think of the arbitration like this:
        Low Byte High Byte
	XXXXXXXX XXXXXXXX
	-----------------
	01000000 10000000

This is what an arbitration for the bus might look like. The device
at id 1 cannot see that id 8 is also arbitrating.  So if they both
pop up during the same arbitration cycle, id 8 will bow out.  Luckily,
SCSI devices don't want to have the bus all the time, so ID 8 will win
some other time and then take the bus.  The BSY and SEL lines are seen
by all devices, so ID 1 can tell if the bus is busy or that it is too
late to participate in an arbitration cycle, so there is no conflict.

>Especially since I do have a setup with 2940UW and both narrow and wide
>devices -- and the Low ID's start to get crowded :-)
>Could I indeed switch both wide harddrives from their ID's 1 and 6 to
>ones above 8 and still use the (ID-8) by a narrow device?

Certainly.

>There were at least strong recommendations never to try that. Some
>people had it working but others said it's just because some UW devices
>are consistently faster and actually win the race.

It may be that some wide devices are buggy, but if they follow the spec
and your bus isn't too long (meaning the bus skew delays for the signals
violate the spec) you should be okay.

--
Justin



To Unsubscribe: send mail to majordomo at FreeBSD.org
with "unsubscribe aic7xxx" in the body of the message




More information about the aic7xxx mailing list