From owner-freebsd-smp  Mon May  6 13:50:14 1996
Return-Path: owner-smp
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id NAA08110
          for smp-outgoing; Mon, 6 May 1996 13:50:14 -0700 (PDT)
Received: from spinner.DIALix.COM (spinner.DIALix.COM [192.203.228.67])
          by freefall.freebsd.org (8.7.3/8.7.3) with ESMTP id NAA08099
          for <smp@freebsd.org>; Mon, 6 May 1996 13:50:07 -0700 (PDT)
Received: from spinner.DIALix.COM (localhost.DIALix.oz.au [127.0.0.1]) by spinner.DIALix.COM (8.7.5/8.7.3) with ESMTP id EAA04998; Tue, 7 May 1996 04:49:55 +0800 (WST)
Message-Id: <199605062049.EAA04998@spinner.DIALix.COM>
X-Mailer: exmh version 1.6.6 3/24/96
To: Poul-Henning Kamp <phk@critter.tfs.com>
cc: smp@freebsd.org
Subject: Re: CVS update: sys/i386/i386 
In-reply-to: Your message of "Mon, 06 May 1996 20:18:28 GMT."
             <10204.831413908@critter.tfs.com> 
Date: Tue, 07 May 1996 04:49:54 +0800
From: Peter Wemm <peter@spinner.DIALix.COM>
Sender: owner-smp@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

>>  - the APIC is no longer mapped Virtual == Physical, because this 
required
>>    tweaking of the PTD's and other nightmares.  It now is dereferenced
>>    via a pointer (unsigned int *apic_base;)
>should be (unsigned int *lapic_base) ?

That's the least of our worries at the moment.. ;-) ;-)

>>  - the APIC register defines redone to cope with this.  Now are offsets
>>    instead of absolute addresses.  This means that the boot code needs 
to
>>    manually compensate when it's making a physical reference before 
paging
>>    is set up.  It also means that _get_mplock cannot be used before 
paging
>>    is enabled.
>get_mplock isn't called until after paging is enabled, no worries.

I was pretty sure this was the case, I didn't see any escapes from 
locore.s beforehand, and I'm not aware of any cases where paging is 
disabled after booting.. :-)

>>  - pmap.c sets up the apic mapping in rather than locore.  A lot of the 
othe
>r
>>    apic initialisation could be done in C too rather than 
mpcore.s/locore.s
>yes, yes, YES!

Also:
  One stack/PTD/PCB/UPAGES can be created later in the boot after the 
memory allocator is running so that one idle pcb (for the recursive mplock 
nesting count) and stack is available per 'n' cpus, rather than forking an 
entire process and keeping them running outside of the idle loop.

Hmm, machdep.c can call something to parse the MP config block to see how 
many cpus it can expect to find.  No wait..  We have an option for this, 
NCPU.. :-)

I'm tired.. It's 5AM again.. :-(

>--
>Poul-Henning Kamp           | phk@FreeBSD.ORG       FreeBSD Core-team.
>http://www.freebsd.org/~phk | phk@login.dknet.dk    Private mailbox.
>whois: [PHK]                | phk@ref.tfs.com       TRW Financial 
Systems, Inc
>.
>Future will arrive by its own means, progress not so.

Cheers,
-Peter