Subject: Re: pmax autoconfig changes: please try out!
To: Jonathan Stone <jonathan@DSG.Stanford.EDU>
From: Simon Burge <simonb@telstra.com.au>
List: port-pmax
Date: 01/31/1996 08:56:47
> Today I checked in some code I've been bashing on part-time for about
> a month.  The Decstation TurboChannel drivers now use sys/dev/tc/tc.c
> for autoconfiguration.  The old, ad-hoc TC autoconfiguration tables
> have been changed to use  the TC bus-attach structure defined in
> sys/dev/tc/tcvar.h.
> 
> [ stuff deleted ]
> 
> I've also  committed these changes to the -current source tree,
> so anyone wants will be able to sup them tomorrow.  By default,
> the autoconfiguration code prints lots of debugging, so I can
> learn something informative if the kernel fails to boot on a
> particular machine :).

One thing that seems broken with these changes is the RTC interupt
calculations in pmax/clock.c.  Before these changes, the default RTC
interrupt rate was defined, then the SELECTED_RATE was figured out.
This has now gone ass-about.

A simple fix for now was "options HZ=1024" in my config file.

While on this topic, /sys/conf/param.c has something like:

	#ifndef HZ
	#define HZ 100
	#endif

in it.  It seems this is only used to set the "hz" variable, and
cpu_initclocks() resets "hz" to the `correct' value soon after.  Would
it be easier to make HZ a mandatory config file option, and do away
with defaults?

Simon.

ps:  The kernel is still compiling, more news later...

pps:  I noticed a few comments about 5100's in the new code.  We've got
one around here.  It's normally a production machine, but will be going
quiet for a little while.  I should be able to `borrow' it every now
and then to test on...