NetBSD Problem Report #7431

Received: (qmail 24412 invoked from network); 21 Apr 1999 07:16:07 -0000
Message-Id: <199904201524.RAA01893@dachau.marco.de>
Date: Tue, 20 Apr 1999 17:24:18 +0200 (CEST)
From: Matthias Pfaller <leo@dachau.marco.de>
Reply-To: leo@dachau.marco.de
To: gnats-bugs@gnats.netbsd.org
Subject: SHARKs with a serial console and com.c 1.59.2.1 won't boot
X-Send-Pr-Version: 3.95

>Number:         7431
>Category:       kern
>Synopsis:       SHARKs with a serial console and com.c 1.59.2.1 won't boot
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    ross
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Wed Apr 21 00:20:01 +0000 1999
>Closed-Date:    Thu Apr 22 16:52:01 +0000 1999
>Last-Modified:  Thu Apr 22 16:52:43 +0000 1999
>Originator:     Matthias Pfaller
>Release:        19990420
>Organization:
Matthias Pfaller                            Software Entwicklung
marco Systemanalyse und Entwicklung GmbH    Tel   +49 8131 5161-41
Hans-Böckler-Str. 2, D 85221 Dachau         Fax   +49 8131 5161-66
http://www.marco.de/                        Email leo@dachau.marco.de
>Environment:

System: NetBSD lethe 1.4_BETA NetBSD 1.4_BETA (LETHE) #101: Tue Apr 20 16:36:22 CEST 1999 leo@lethe:/usr/src/sys/arch/arm32/compile/LETHE arm32


>Description:
	SHARKs with a serial console and com.c 1.159.2.1 won't boot.
>How-To-Repeat:
	Build a kernel with com.c 1.159.2.1. Reboot and watch it stop even
	before outputting the copyright message.
>Fix:
	Revert to 1.159 and everything works as before.
>Release-Note:
>Audit-Trail:

From: matthew green <mrg@eterna.com.au>
To: leo@dachau.marco.de
Cc: gnats-bugs@gnats.netbsd.org, thorpej@netbsd.org, ross@netbsd.org
Subject: re: kern/7431: SHARKs with a serial console and com.c 1.59.2.1 won't boot 
Date: Wed, 21 Apr 1999 19:02:58 +1000



    >Description:
    	SHARKs with a serial console and com.c 1.159.2.1 won't boot.
    >How-To-Repeat:
    	Build a kernel with com.c 1.159.2.1. Reboot and watch it stop even
    	before outputting the copyright message.
    >Fix:
    	Revert to 1.159 and everything works as before.




 i guess i'm glad i didn't update and test my kernel.


 why is it acceptable to break the shark serial console?  _twice_ now,
 it would seem.  1.59 is the version after i backed out these changes
 last time it broke the shark serial console and i had no response
 from the author of these changes about it.

 note:  while i've heard that it is a "arm32" bug that is the problem
 here, i consider that IMMATERIAL!  this driver has worked for the past
 year(?) or so on this machine and to break it DURING the 1.4 cycle is
 criminal!


 jason/ross:  please fix the com driver for the shark.  or backout these
 changes!!!! (again).



 annoyed .mrg.

From: Ross Harvey <ross@ghs.com>
To: leo@dachau.marco.de, mrg@eterna.com.au
Cc: gnats-bugs@gnats.netbsd.org, root@ihack.net, thorpej@netbsd.org
Subject: re: kern/7431: SHARKs with a serial console and com.c 1.59.2.1 won't boot
Date: Wed, 21 Apr 1999 11:17:21 -0700 (PDT)

 : From Matt "Flamer" Green, who Does Not Quite Understand What Happened

 >    >Description:
 >    	SHARKs with a serial console and com.c 1.159.2.1 won't boot.
 >    >How-To-Repeat:
 >    	Build a kernel with com.c 1.159.2.1. Reboot and watch it stop even
 >    	before outputting the copyright message.
 >    >Fix:
 >    	Revert to 1.159 and everything works as before.
 >
 > i guess i'm glad i didn't update and test my kernel.
 >
 >
 > why is it acceptable to break the shark serial console?  _twice_ now,
 > it would seem.  1.59 is the version after i backed out these changes
 > last time it broke the shark serial console and i had no response
 > from the author of these changes about it.
 >
 > note:  while i've heard that it is a "arm32" bug that is the problem
 > here, i consider that IMMATERIAL!  this driver has worked for the past
 > year(?) or so on this machine and to break it DURING the 1.4 cycle is
 > criminal!
 >
 >
 > jason/ross:  please fix the com driver for the shark.  or backout these
 > changes!!!! (again).

 Please try to understand that we need to fix the problems, not just back
 out a fix that needed more work for Shark, thus trading breakage from one
 port to another.

 Breakage attempting a fix for both ports is different than intentionally
 breaking it on one port so another port works.

 There is a difference between the way we broke it on Shark and the way you
 broke it by reverting. We broke it making forward progress, although to be
 sure, incomplete progress.  You broke it by just removing the fixes instead
 of actually helping. If you break it attempting to make it work on both,
 that's a problem, but if you break it intentionally removing fixes that's
 a bigger problem.

 I will fire up a serial console Shark tonight and come up with something
 that works on both platforms...but, everyone else, please feel free to beat
 me to it and work on com.c yourself if you are so inclined.

 For those catching up:

 	There was a race in com_common_putc() on serial console
 	machines ... after kernel output it would clear the TX
 	interrupt. This would sometimes freeze the console if done
 	concurrently with output through the driver.

 	If you revert all the changes, you reactivate that problem.

 Ross.Harvey@Computer.Org

From: Ross Harvey <ross@ghs.com>
To: leo@dachau.marco.de, mrg@eterna.com.au, ross@elbe.ghs.com
Cc: gnats-bugs@gnats.netbsd.org, root@ihack.net, thorpej@netbsd.org
Subject: re: kern/7431: SHARKs with a serial console and com.c 1.59.2.1 won't boot
Date: Wed, 21 Apr 1999 11:44:03 -0700 (PDT)

 : From mrg...
 :::
 > > note:  while i've heard that it is a "arm32" bug that is the problem
 > > here, i consider that IMMATERIAL!  this driver has worked for the past
 > > year(?) or so on this machine and to break it DURING the 1.4 cycle is
 > > criminal!
 :::

 Oh, and for anyone actually looking at what happened: the known ``arm32''
 problem was that delay() used a timer, and thus couldn't be called until
 the timer was configured. This is indeed a bug, if delay() wants to use
 a timer, it needs to know when that timer is available and have a fallback
 algorithm to use before that.

 However, when I took the delay()'s out, it apparently still didn't work
 so there is another problem, one as yet unknown.

 	Ross.Harvey@Computer.Org
State-Changed-From-To: open->feedback 
State-Changed-By: ross 
State-Changed-When: Thu Apr 22 03:13:34 PDT 1999 
State-Changed-Why:  
It's fixed. Reverting com.c would break things on alpha, and in any case 
it wasn't the problem, it just happened to uncover a bug in the arm32 
isa bus_space_barrier() implementation. 

I'm not sure if I beat the sup scan this morning; when you have: 

src/sys/arch/arm32/isa: isa_io.c revision 1.6 

...then you should be OK. 


Responsible-Changed-From-To: kern-bug-people->ross 
Responsible-Changed-By: ross 
Responsible-Changed-When: Thu Apr 22 03:13:34 PDT 1999 
Responsible-Changed-Why:  
State-Changed-From-To: feedback->closed 
State-Changed-By: ross 
State-Changed-When: Thu Apr 22 09:52:01 PDT 1999 
State-Changed-Why:  
Originator confirms successful test of fix. 
>Unformatted:

NetBSD Home
NetBSD PR Database Search

(Contact us) $NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026 The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.