From owner-freebsd-smp  Thu Dec  5 15:39:12 1996
Return-Path: <owner-smp>
Received: (from root@localhost)
          by freefall.freebsd.org (8.8.4/8.8.4) id PAA12743
          for smp-outgoing; Thu, 5 Dec 1996 15:39:12 -0800 (PST)
Received: from clem.systemsix.com (clem.systemsix.com [198.99.86.131])
          by freefall.freebsd.org (8.8.4/8.8.4) with SMTP id PAA12738
          for <smp@freebsd.org>; Thu, 5 Dec 1996 15:39:08 -0800 (PST)
Received: from localhost (localhost [127.0.0.1]) by clem.systemsix.com (8.6.12/8.6.12) with SMTP id QAA15728 for <smp@freebsd.org>; Thu, 5 Dec 1996 16:39:04 -0700
Message-Id: <199612052339.QAA15728@clem.systemsix.com>
X-Authentication-Warning: clem.systemsix.com: Host localhost didn't use HELO protocol
X-Mailer: exmh version 1.6.5 12/11/95
From: Steve Passe <smp@csn.net>
To: smp@freebsd.org
Subject: Re: Crashing on activating other CPUs 
In-reply-to: Your message of "Wed, 04 Dec 1996 10:33:24 MST."
             <199612041733.KAA06644@clem.systemsix.com> 
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Date: Thu, 05 Dec 1996 16:39:04 -0700
Sender: owner-smp@freebsd.org
X-Loop: FreeBSD.org
Precedence: bulk

Hi,

> > Our 4 CPU Netserver works without SMP_INVLTLB and APIC_IO (the latter
> > has never worked). With SMP_INVLTLB, I get something similar the
>  ...
> I think the problem is that damn re-direction thingy for PCI INTs to ISA bus.
>  ...
> your system says the PCI disk controler uses:
> --
> I/O Ints:       Type    Polarity    Trigger     Bus ID   IRQ    APIC ID INT#
>     INT      conforms       level       19     7         14    7
>              ^^^^^^^^       ^^^^^
> 
> Which is different from the ISA standard of active-hi/edge.  So far our 
> experiments suggested that for cards (PCI & EISA) redirected to ISA INTs
> we should ignore the entries and stick with ISA levels.  This works
> on at least 1 EISA machine and 1 PCI machine that I have encountered.
> However the printout you sent earlier shows that your PCI card is
> completely loosing its INTs ubder APIC_IO, symptomatic of this pin
> being incorrectly programmed.

with Terje's success today running 4 CPUs with APIC_IO it appears I jumped
to the wrong conclusion about this issue.

So it appears to now be fact that:

  whenever PCI or EISA cards are redirected to ISA IRQs the levels get set
  to ISA values (active-hi/edge) reguardless of the "settings" in the
  mptable.

--
Steve Passe	| powered by
smp@csn.net	|            FreeBSD