From owner-freebsd-smp  Wed Dec  3 17:44:26 1997
Return-Path: <owner-freebsd-smp>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id RAA03592
          for smp-outgoing; Wed, 3 Dec 1997 17:44:26 -0800 (PST)
          (envelope-from owner-freebsd-smp)
Received: from mailhost.Ipsilon.COM (mailhost.ipsilon.com [205.226.5.12])
          by hub.freebsd.org (8.8.7/8.8.7) with SMTP id RAA03587
          for <smp@freebsd.org>; Wed, 3 Dec 1997 17:44:23 -0800 (PST)
          (envelope-from jre@ipsilon.com)
Received: from radio.ipsilon.com (radio.Ipsilon.COM [205.226.28.3]) by mailhost.Ipsilon.COM (8.6.11/8.6.10) with SMTP id RAA05088; Wed, 3 Dec 1997 17:27:33 -0800
Message-ID: <34860715.31DFF4F5@ipsilon.com>
Date: Wed, 03 Dec 1997 17:27:49 -0800
From: Joe Eykholt <jre@Ipsilon.COM>
Organization: Ipsilon Software Engineering
X-Mailer: Mozilla 3.01Gold (X11; I; FreeBSD 2.1.0-RELEASE i386)
MIME-Version: 1.0
To: Steve Passe <smp@csn.net>
CC: smp@freebsd.org
Subject: Re: SMP
References: <199712040101.SAA10349@Ilsa.StevesCafe.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: owner-freebsd-smp@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

Steve,

I'm glad you see the problem.  I wasn't pressing for an immediate
solution so much as pointing out the problem.  I think the long term
solution is to do the EOI after the interrupt is handled, even if it
was deferred until doreti, but that probably requires assigning
priorities to interrupts, and everything that goes along with that,
maybe even making cpl a local-APIC priority instead of an IRQ mask.

I'm not sure UNMASK_IRQ() should do the eoi either, since that can
happen from doreti as well, right?  Maybe the Xintrxx and Xresumexx
should both call the handler, Xintrxx should do the EOI, and then
join up for the doreti.  ... but the problem with that is that lower
priority interrupts are blocked for the duration of the handler ...
so that seems to force the determination of interrupt priorities.

It's funny that the APIC design seemingly forces the old kernel
spl-priority model after BSD adapted to the PIC model.  It seems 
to me the h/w and s/w designers both tried to adapt to the other 
and both ended up changing to be compatible and thereby ended 
up incompatible!

It might take awhile before I'd be able to test any change 
in this area, I can't promise very quick turnaround.

	Joe