NetBSD Problem Report #37930

From martin@duskware.de  Thu Jan 31 11:44:45 2008
Return-Path: <martin@duskware.de>
Received: from mail.netbsd.org (mail.netbsd.org [204.152.190.11])
	by narn.NetBSD.org (Postfix) with ESMTP id 5FE8263B8BD
	for <gnats-bugs@gnats.netbsd.org>; Thu, 31 Jan 2008 11:44:45 +0000 (UTC)
Message-Id: <20080131113912.0DCA963B8A2@narn.NetBSD.org>
Date: Thu, 31 Jan 2008 11:39:12 +0000 (UTC)
From: ad@netbsd.org
Reply-To: ad@netbsd.org
To: netbsd-bugs-owner@NetBSD.org
Subject: sparc mutex stubs are broken on MP
X-Send-Pr-Version: www-1.0

>Number:         37930
>Category:       port-sparc
>Synopsis:       sparc mutex stubs are broken on MP
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    port-sparc-maintainer
>State:          suspended
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Jan 31 11:45:01 +0000 2008
>Closed-Date:    
>Last-Modified:  Thu Feb 14 14:16:00 +0000 2008
>Originator:     Andrew Doran
>Release:        4.99.52
>Organization:
The NetBSD Project
>Environment:
n/a
>Description:
The sparc mutex stubs are broken on MP systems. Usually they
work like this:

- grab the mutex interlock
- that suceeds, so set the full owning LWP address into mtx_owner

On an MP system the following can occur:

cpu2  acquire kernel_lock
cpu1  grab mutex interlock
cpu1  device interrupt occurs before setting mtx_owner
cpu1  spin on kernel_lock trying to process device interrupt
cpu2  spin or block trying to acquire mutex 'half held' by cpu1

-> potentially deadlock
>How-To-Repeat:
Code inspection.
>Fix:
- Implement 'restart' for mutex_enter() in all the interrupt stubs.
  This would be very tricky.

or:

- Change the sparc to __HAVE_SIMPLE_MUTEXES.
- Use atomic_cas_ptr() to acquire and release mutexes.
- Work on optimizing atomic_cas_ptr().

I'll create a patch.


>Release-Note:

>Audit-Trail:
From: Andrew Doran <ad@netbsd.org>
To: gnats-bugs@netbsd.org
Cc: 
Subject: Re: port-sparc/37930: sparc mutex stubs are broken on MP
Date: Thu, 31 Jan 2008 11:59:20 +0000

 Here's a patch:

 http://www.netbsd.org/~ad/sparc.diff

Responsible-Changed-From-To: port-sparc-maintainer->ad
Responsible-Changed-By: ad@narn.netbsd.org
Responsible-Changed-When: Thu, 14 Feb 2008 14:05:00 +0000
Responsible-Changed-Why:
take


From: Andrew Doran <ad@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc: 
Subject: PR/37930 CVS commit: src/sys/arch/sparc
Date: Thu, 14 Feb 2008 14:07:35 +0000 (UTC)

 Module Name:	src
 Committed By:	ad
 Date:		Thu Feb 14 14:07:35 UTC 2008

 Modified Files:
 	src/sys/arch/sparc/conf: files.sparc
 	src/sys/arch/sparc/include: mutex.h rwlock.h

 Log Message:
 Make sparc use atomic_cas_ulong() for mutex and rwlock operations, and
 disable the custom mutex/rwlock code. PR port-sparc/37930. ok martin@


 To generate a diff of this commit:
 cvs rdiff -r1.144 -r1.145 src/sys/arch/sparc/conf/files.sparc
 cvs rdiff -r1.7 -r1.8 src/sys/arch/sparc/include/mutex.h
 cvs rdiff -r1.3 -r1.4 src/sys/arch/sparc/include/rwlock.h

 Please note that diffs are not public domain; they are subject to the
 copyright notices on the relevant files.

Responsible-Changed-From-To: ad->port-sparc-maintainer
Responsible-Changed-By: ad@narn.netbsd.org
Responsible-Changed-When: Thu, 14 Feb 2008 14:16:00 +0000
Responsible-Changed-Why:
back to port-sparc-maintainer for the time being.


State-Changed-From-To: open->suspended
State-Changed-By: ad@narn.netbsd.org
State-Changed-When: Thu, 14 Feb 2008 14:16:00 +0000
State-Changed-Why:
mrg asked me to leave this open because it could use optimization.

We could either make the custom mutex scheme work properly, or optimize
atomic_cas_ulong() so that it does not need to disable interrupts. Both of  
these would mean changing the interrupt handling code to complete and/or
restart a CAS operation if it's interrupted. I think it would be better to
optimize atomic_cas_ulong() since it would mean we're focusing one one
routine instead of many.


>Unformatted:

Home	PR Database Search