sps for Ultrix 4.1?

[Jeffrey Mogul]

In article <EMV.91Feb4012458 at crane.aa.ox.com> emv at ox.com (Ed Vielmetti) writes:
>In article <1991Feb2.184154.14827 at mintaka.lcs.mit.edu> metcalf at masala.lcs.mit.edu (Chris Metcalf) writes:
>
>   Has anyone ported sps to 4.1 yet?  I have a version for 4.0, but it
>   doesn't quite work.
>
>The latest version that I've seen posted is from Robert Ward back in
>November to alt.sources; in the posting he says the DEC port was
>incomplete because
>
>	... In particular, the code
>	that mimics the virtual to physical address translation is incomplete.
>	Upage information is accessed correctly but not the command line
>	arguments.  If anyone knows how to make this code work, please contact
>	me.  I have no idea if this works on 2100's or 5100's.
>
>If I had to pick based on past experience I'd bet that Jeff Mogul is
>up to the task; he's ported several otherwise Sun-only things to
>Ultrix...

While I might be "up to the task" I've gotten quite bored with reworking
such programs, and I have too much else to do at the moment.  I'd suggest
that anyone who wants to take a stab at it should look at the sources
for top and pmon currently residing on gatekeeper.dec.com:pub/DEC/ .

There seem to be 4 things that need to be conditionalized when doing
this kind of port:
	(1) MIPS machines (like SPARCs) don't do floating point in
	kernel mode, so some values are representing in a fixed-point
	format.  Mostly, use the right macros and you'll be fine.
	(2) Some fields in the proc structure have been renamed or
	split up into a number of new fields.
	(3) Multiprocessor support means that some kernel globals
	now come in arrays rather than as scalars.
	(4) The code to read the "u page" is different.
Examples of Ultrixificated ways of doing all these are available in
the top and pmon sources.

-Jeff

P.S.: if someone succeeds at porting SPS, I'd be happy to put the
sources on gatekeeper for public access.