🚀 go-pugleaf

>>>>> On 6 Oct 92 07:48:25 GMT, ogicse!news.u.washington.edu!stein.u.washington.edu!mpark@decwrl.dec.com  (Michael Park) said:

	Michael> 2. Given (1) above, two questions spring to mind:
	Michael> a. since the USP is never saved, does that mean that
	Michael>       the 68000 is ALWAYS in supervisor mode?  Or,

This is correct.  Non-VM versions of Xinu all run in "priveleged"
state, whatever that is for a particular machine.  Xinu was designed
to be a very small, fast system; protection was not a high priority.
Switching from user to protected mode is very expensive, particularly
so on modern RISC machines, because it generally involves a system
trap, saving all the registers, and etc, and it needs to happen on
every system call.  Even on a big CISC chip like the 68020, it's still
quite expensive.


	Michael> 3. I was leafing through the Mac edition of the xinu
	Michael>    book, and was forced to conclude that Mac xinu
	Michael>    runs completely in supervisor mode (this was never
	Michael>    stated explicitly as far as I could tell; there
	Michael>    wasn't even an index entry for "supervisor").  Is
	Michael>    that deduction correct, and is it true for Mac
	Michael>    programs in general that everything is done in
	Michael>    supervisor state?

That's probably correct, but I don't recall for sure.

	Michael> 4. Isn't running solely in supervisor state a bit of
	Michael>    a lose?

Why do you say that?  It's a typical tradeoff of speed and efficiency
for protection.  For our purposes, we want it fast and small.  For
other projects, protection was more important, so multiple modes were
added.  It all depends on your requirements.

Shawn
-------------------------------------------------------------------------
Shawn Ostermann  -  PhD Student  -  Purdue Internetworking Research Group
  Internet:  sdo@cs.purdue.edu                  AT&T: (317) 494-7836
  UUCP:  {backbone}!purdue!sdo                  Fax:  (317) 494-0739
-------------------------------------------------------------------------