From nobody  Mon Dec 14 15:04:17 1998
Received: (from nobody@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id PAA27210;
          Mon, 14 Dec 1998 15:04:17 -0800 (PST)
          (envelope-from nobody)
Message-Id: <199812142304.PAA27210@hub.freebsd.org>
Date: Mon, 14 Dec 1998 15:04:17 -0800 (PST)
From: mjmccut@cs.ubc.ca
To: freebsd-gnats-submit@freebsd.org
Subject: 3.0-RELEASE dies starting AP on dual PPro, looks like interrupt problem.
X-Send-Pr-Version: www-1.0

>Number:         9081
>Category:       kern
>Synopsis:       3.0-RELEASE dies starting AP on dual PPro, looks like interrupt problem.
>Confidential:   no
>Severity:       critical
>Priority:       medium
>Responsible:    n_hibma
>State:          closed
>Quarter:        
>Keywords:       
>Date-Required:  
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon Dec 14 15:10:00 PST 1998
>Closed-Date:    Sat Jul 24 03:00:26 PDT 1999
>Last-Modified:  Mon Jul 23 01:10:50 PDT 2001
>Originator:     Mark McCutcheon
>Release:        3.0-RELEASE from Walnut Creek CD-ROM
>Organization:
UBC Computer Science
>Environment:
FreeBSD 3.0-RELEASE #10: Thu Dec 10 17:45:28 PST 1998
root@hozameen.cs.ubc.ca:/usr/src/sys/compile/HOZAMEEN-SMP
(NOTE: from 'strings /kernel'; kernel doesn't boot so can't do uname)
>Description:
HP Vectra XU 6/200 dual 200MHz PPro ran SMP in several v.3 snapshots
(e.g. 3.0-980107-SNAP) but hangs early in boot process with 3.0-RELEASE
kernel.  Due to early demise I haven't been able to log the bootup info
or recover it via dmesg; however, last screen displays:

ISA Pin=1, IRQ=-1, Pol=0, Trig=0
IOAPIC #0 intpint 1 -> irq -1
ISA Pin=2 IRQ=-1, Pol=0, Trig=0
IOAPIC #0 intpint 2 -> irq -1
..........
ISA Pin=15 IRQ=-1, Pol=0, Trig=0
IOAPIC #0 intpint 15 -> irq -1

Debugging message inserted in /usr/src/sys/i386/i386/mp_machdep.c just
before
  CHECK_INIT(99); /* setup checkpoints */
(approx. line 1860) prints out but none of the subsequent trace or panic
messages in the code are seen.
>How-To-Repeat:
Make kernel with:
  options SMP
  options APIC_IO
and run on dual-PPro Vectra XU 6/200.  May be related to bug report
kern/8456 (SMP boot failure on dual PPro Everex PO-6200 with
3.0-RELEASE).
>Fix:
Fix not yet determined.  There have been MANY changes and additions to
interrupt setup code in 3.0-RELEASE re. 3.0-980107-SNAP, as well as the
apparent removal of special handling for CPUCLASS_686 (in
/usr/src/sys/i386/i386/mp_machdep.c).  Possibly there are motherboard
issues but it DID work fine in earlier pre-RELEASE snapshots....
>Release-Note:
>Audit-Trail:
State-Changed-From-To: open->analyzed 
State-Changed-By: n_hibma 
State-Changed-When: Fri Jul 23 06:36:43 PDT 1999 
State-Changed-Why:  
It looks very much like 8456 which has been fixed. 


Responsible-Changed-From-To: freebsd-bugs->n_hibma-bugs 
Responsible-Changed-By: n_hibma 
Responsible-Changed-When: Fri Jul 23 06:36:43 PDT 1999 
Responsible-Changed-Why:  
remind me to close the PR in case this is the case. 

From: Nick Hibma <nick.hibma@jrc.it>
To: freebsd-gnats-submit@freebsd.org, mjmccut@cs.ubc.ca
Cc:  
Subject: Re: kern/9081: 3.0-RELEASE dies starting AP on dual PPro, looks like 
 interrupt problem.
Date: Fri, 23 Jul 1999 15:28:24 +0200

 Did you try the fix that has been posted in PR 8456?
 
 	http://www.freebsd.org/cgi/query-pr.cgi?pr=9081
 
 Have you tried newer RELEASEs or CURRENT?
 
 Nick
 -- 
 ISIS/STA, T.P.270, Joint Research Centre, 21020 Ispra, Italy
 
Responsible-Changed-From-To: n_hibma-bugs->n_hibma 
Responsible-Changed-By: n_hibma 
Responsible-Changed-When: Fri Jul 23 07:10:24 PDT 1999 
Responsible-Changed-Why:  
typing is difficult. my name doubly so. 
State-Changed-From-To: analyzed->closed 
State-Changed-By: n_hibma 
State-Changed-When: Sat Jul 24 03:00:26 PDT 1999 
State-Changed-Why:  
Tor Egge agrees that problem is probably solved by the solution  
given in 8456. 

From: Nick Hibma <n_hibma@webweaving.org>
To: freebsd-gnats-submit@freebsd.org
Cc:  
Subject: Re: kern/9081 (fwd)
Date: Sun, 15 Aug 1999 16:29:44 +0200 (CEST)

 ---------- Forwarded message ----------
 Date: Tue,  3 Aug 1999  9:09:33 UTC-0700
 From: Mark McCutcheon <mjmccut@cs.ubc.ca>
 To: n_hibma@freebsd.org
 Subject: Re: kern/9081
 
 This message bounced (user unknown) so I am resending to your
 FreeBSD email address.
 
 Forwarded IPMessage:
 
 From:           Mark McCutcheon <mjmccut@cs.ubc.ca>
 To:             <n_hibma@jrc.it>
 In-Reply-To:    FreeBSD-Bugs:235
 Message-ID:     FreeBSD-Bugs:238 15870*mjmccut@cs.ubc.ca
 Subject:        Re: kern/9081: 3.0-RELEASE dies starting AP on dual PPro,
                  looks like  interrupt problem.
 
 > Did you try the fix that has been posted in PR 8456?
 >
 >	http://www.freebsd.org/cgi/query-pr.cgi?pr=9081
 >
 
 I apologize for this late reply but I have just returned after
 2 weeks away on holiday.
 
 Yes, I have tried the PR8456 fix, and it unfortunately makes
 no difference in the failure of this machine (HP Vectra XU
 6/200) to function in SMP mode.  For whatever reason, early
 releases of 3.0 worked as expected. but since the major
 revision in SMP code which appeared at about 3.0-RELEASE it
 appears that the APICs on this motherboard are not being
 accessed correctly.
 
 > Have you tried newer RELEASEs or CURRENT?
 >
 > Nick
 > -- 
 > ISIS/STA, T.P.270, Joint Research Centre, 21020 Ispra, Italy
 
 Yes, I am tracking 3-STABLE via CVSup and making world about
 once a week.  I try an SMP kernel from time to time but this
 machine remains unbootable for the same reasons as originally
 described.
 
 > State-Changed-From-To: analyzed->closed
 > State-Changed-By: n_hibma
 > State-Changed-When: Sat Jul 24 03:00:26 PDT 1999
 > State-Changed-Why: 
 > Tor Egge agrees that problem is probably solved by the
 > solution given in 8456.
 
 I have no objection to you closing this PR as I have given up
 hope of running SMP on this machine.  At some point when we
 can afford to do so, it will be replaced by another PC with a
 known-functional motherboard.  However, please note that
 "State-Changed-Why" is incorrect in asserting that the problem
 is solved by the 8456 solution.
 
 Thanks very much,
 
 Mark McCutcheon
 
 
 
>Unformatted:
