[Coco] MPI and NitrOS-9

Gene Heskett gheskett at shentel.net
Fri Oct 21 10:51:58 EDT 2016


On Friday 21 October 2016 08:15:02 Bill wrote:

> Was there some information lost in this section? Seems disjointed to
> me:

Yes, it is a bit disjointed, but because of all the interactions, I can't 
suggest a better rewrite. Thats why I wrote this below.  Its all there 
and perhaps too concise for easy understanding.

> "Much of this is better explained by a detailed read of the
> superdesc.asm file in the nitros9 build directory. Because bits in the
> descriptor are hard to find spares for adding these new functions, 1
> or more bits may control the meaning of 3 or 4 other bits, so read,
> and re-read until the interactions become clear.
>
> dmode, which can adjust a descriptor in memory on the fly, or a
> descriptor on the disk to build the next boot disk, will be your best
> friend while learning all this."

dmode can adjust everything mentioned in superdesc.asm was my point...

To adjust something in memory is probably the wrong way to do it because 
you lose it with a  reboot, so when experimenting, the descriptor s/b 
removed from the bootfile and loaded from disk after dmode-ing it, then 
unloading it if not right, wash rinse & repeat till it is, then put that 
one back into the os9gen-erated boot.

One of your posts indicated 4 drives but that won't work because I do not 
know how to keep the drive 0 from responding to the side select, so the 
limit is 3 drives per controller. I don't believe you need a second 
driver, but the descriptor will need to have the slot the 2nd controller 
is in, put into it by dmode. Doing that would allow 3 more drives, 3, 4, 
& 5 but they would be physically addressed as 0, 1, & 2. The diff is the 
slot number in the descriptor.

At some point, you'll likely run out of system ram because of the space 
use by the disk buffers.

This is dirt I haven't walked on in 15 or more years, Bill, so I'd not be 
amazed if something goes aglay in this.

> -----Original Message-----
> From: Coco [mailto:coco-bounces at maltedmedia.com] On Behalf Of Gene
> Heskett Sent: Thursday, October 20, 2016 11:07 PM
> To: coco at maltedmedia.com
> Subject: Re: [Coco] MPI and NitrOS-9
>
> On Thursday 20 October 2016 22:14:12 Bill wrote:
> > So, if I were able to build a ribbon cable to connect all four
> > drives, I could connect it to one controller? I might try that
> > tomorrow. I have the connectors and PLENTY of ribbon cable, but what
> > about the twists on the cable? Am I building a cable basically
> > identical to the two existing, but with four connectors?
>
> Bill; There are 4 drive select wires in the cable, but one is commonly
> used for side select, so that you can connect only 3 physical drives
> at any one time without useing another controller. Two would allow 6
> drives for instance. IIRC that would need some tricky stuff not
> usually played with in the descriptors to isolate the two controllers
> shareing common access addresses in the $FF40-$FF48 range.
>
> This means also that there are no missing teeth in any of the
> connectors, nor is a cable twist required.  What is required is that
> you go thru the drives and by solder bridges, flea clips or dip
> switches, each drive has its own unique address from 0 to 2. But the
> 3.5" drives have only 2 possible addresses, and as shipped they are
> all set to address 1, and the twisted cable makes the drive on the end
> of the cable into drive 0 by swapping those 2 select wires.  Works
> fine on a doggoned pc, but not for os9.
>
> So common practice here at the coyote.den is to have a pair of 3.5"
> drives, with one having its addressing solder blobs reset to make it
> drive 0. The next, is already drive 1, and a 5.25" of your drive
> choice at the drive 2 address. Since this drive can be a 35 track ss,
> a 40 track ds, or an 80 track ds, Robert and I worked out a patch for
> the nitros9 rb1773.dr floppy driver such that if a 40 track disk is
> inserted in the 80 track drive, that drive will be double stepped so
> the heads stay in the center of the wider 40 track pattern for reads,
> and if an attempt is made to write to that disk, it will be found to
> be write protected since the narrower 80 track heads will not properly
> overwrite the wider track.  So without the write protection, the disk,
> which may be the only one you have with that data on it, will refuse
> the write, which if allowed, normally destroys the disk and it must be
> reformatted to use it again.
>
> Much of this is better explained by a detailed read of the
> superdesc.asm file in the nitros9 build directory. Because bits in the
> descriptor are hard to find spares for adding these new functions, 1
> or more bits may control the meaning of 3 or 4 other bits, so read,
> and re-read until the interactions become clear.
>
> dmode, which can adjust a descriptor in memory on the fly, or a
> descriptor on the disk to build the next boot disk, will be your best
> friend while learning all this.


Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
Genes Web page <http://geneslinuxbox.net:6309/gene>


More information about the Coco mailing list