Re: Keyboard lockup with NetBSD5

["Michael L. Hitch" <mhitch%lightning.msu.montana.edu@localhost>]

On Thu, 12 Nov 2009, Frank Wille wrote:

> I would have assumed that delay() works, because the rest of the system
> seems to function correctly. But now as you mention it, I had a closer look

  Probably in most cases, a small error in the actual delay isn't going to
cause problems.

> on delay(). And after a few minutes I came to the conclusion that either I
> don't understand the routine, or there is something wrong with it:
...
> Seems that the author of the delay() function had a decreasing tick counter
> in mind...
>
> When I am right that this here is wrong, it would explain the keyboard
> timing problems partly. ;)

  Since the timecounter code (mostly) worked when I was testing it, I 
didn't look at the delay() code other than to note that it had been 
significantly redone.  Everything else seemed to work, so there wasn't
a need to verify it.

> >    Re that 2000 usec delay used apparently came from me when the Amiga
> >    1200 support was added.  Presumably that was needed to make the 1200
> > I had access to at the time,
>
> Did you ever find out why it was necessary? In the AmigaOS source a
> handshaking delay of 85us is used, as recommended in the Hardware Reference
> Manual, and it works fine for all A1200 models.

  I don't remember much at all about what the problem was back then (that 
was 15 years ago).  From trying to look at the history of DELAY(), it 
might be that it was inaccurate enough at the time that the 2000 was 
needed, but DELAY() was later recalibrated and might have been more 
accurate, but the 2000 didn't cause problems.  When the delay routine was 
rewritten for timecounters (obviously wrong), then it became an issue for 
the 1200 keyboard.

  I would suggest fixing delay() to work properly, and see if the 1200 
keyboard works, and what delay() value it will work properly at.  From the 
CVS history, even the delay(85) wasn't sufficient for some keyboards and 
was changed to delay(200).  [And it appears that the comment about that 
wasn't changed when the A1200 support was added :)]

--
Michael L. Hitch                        mhitch%montana.edu@localhost
Computer Consultant
Information Technology Center
Montana State University        Bozeman, MT     USA