NetBSD Problem Report #14368

Received: (qmail 25946 invoked from network); 26 Oct 2001 22:24:09 -0000
Message-Id: <200110262224.f9QMO7200541@ficus.home>
Date: Sat, 27 Oct 2001 00:24:07 +0200 (CEST)
From: Anthony Mallet <anthony.mallet@useless-ficus.net>
Reply-To: Anthony Mallet <anthony.mallet@useless-ficus.net>
To: gnats-bugs@gnats.netbsd.org
Subject: sysmon_envsys.c deadlock
X-Send-Pr-Version: 3.95

>Number:         14368
>Category:       kern
>Synopsis:       sysmon_envsys.c deadlock
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    jdolecek
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Fri Oct 26 22:25:00 +0000 2001
>Closed-Date:    Sat Feb 16 10:34:16 +0000 2002
>Last-Modified:  Sat Feb 16 18:19:05 +0000 2002
>Originator:     Anthony Mallet
>Release:        -current as of October 20
>Organization:
>Environment:
System: NetBSD ficus 1.5Y NetBSD 1.5Y (FICUS) #13: Fri Oct 26 23:43:01 CEST 2001 toto@ficus:/home/src/netbsd-current/sys/arch/i386/compile/FICUS i386
Architecture: i386
Machine: i386
>Description:
	When /dev/sysmon is opened twice by two different processes, the
second process blocks infinitely, certainly because it cannot acquire the
'sysmon_envsys_lock'. That's ok.

	If the first process stops *before* the second one, no more
process can gain access to /dev/sysmon. See 'How-To-Repeat' for more
details.
>How-To-Repeat:
	1   launch `/usr/sbin/envstat -i 1` ( name it "A").
	2   launch another `/usr/sbin/envstat -i 1`  (name it "B").
	2.5 see how "B" is blocked --- ok.
	3   kill -INT the "A" process.
	3.5 see how the "B" process does *not* gain access to sysmon.
            Not ok!
	4   kill -INT the "B" process.
	4.5 see how *no more* process can access to susmon...
>Fix:
	Dunno at this time. I can compile a kernel with LOCK_DEBUG and
symbols if it is needed.
>Release-Note:
>Audit-Trail:
Responsible-Changed-From-To: kern-bug-people->jdolecek 
Responsible-Changed-By: jdolecek 
Responsible-Changed-When: Sat Nov 17 05:20:07 PST 2001 
Responsible-Changed-Why:  
I'd look at this eventually. 

From: Jaromir Dolecek <jdolecek@netbsd.org>
To: gnats-bugs@netbsd.org
Cc:  
Subject: Re: kern/14368: sysmon_envsys.c deadlock
Date: Thu, 3 Jan 2002 22:56:50 +0100 (CET)

 Hi,
 this should fix this - can you confirm whether it does for you too?
 The problem is that the code assumes device close routine is called
 for each descriptor close, which is not true.

 Jaromir

 XXXX
 Index: sysmon_envsys.c
 ===================================================================
 RCS file: /cvsroot/syssrc/sys/dev/sysmon/sysmon_envsys.c,v
 retrieving revision 1.2
 diff -u -p -r1.2 sysmon_envsys.c
 --- sysmon_envsys.c	2001/11/13 06:28:55	1.2
 +++ sysmon_envsys.c	2002/01/03 21:51:54
 @@ -69,28 +69,18 @@ u_int	sysmon_envsys_next_sensor_index;
  int	sysmon_envsys_initialized;
  struct simplelock sysmon_envsys_initialized_slock = SIMPLELOCK_INITIALIZER;

 -void	sysmon_envsys_init(void);
 +#define	SYSMON_ENVSYS_LOCK()	\
 +	lockmgr(&sysmon_envsys_lock, LK_EXCLUSIVE, NULL)

 +#define SYSMON_ENVSYS_UNLOCK()		\
 +	lockmgr(&sysmon_envsys_lock, LK_RELEASE, NULL)
 +
  int	sysmonioctl_envsys(dev_t, u_long, caddr_t, int, struct proc *);
 -int	sysmonioctl_wdog(dev_t, u_long, caddr_t, int, struct proc *);

  struct sysmon_envsys *sysmon_envsys_find(u_int);
  void	sysmon_envsys_release(struct sysmon_envsys *);

  /*
 - * sysmon_envsys_init:
 - *
 - *	Initialize the system monitor.
 - */
 -void
 -sysmon_envsys_init(void)
 -{
 -
 -	lockinit(&sysmon_envsys_lock, PWAIT|PCATCH, "smenv", 0, 0);
 -	sysmon_envsys_initialized = 1;
 -}
 -
 -/*
   * sysmonopen_envsys:
   *
   *	Open the system monitor device.
 @@ -98,18 +88,14 @@ sysmon_envsys_init(void)
  int
  sysmonopen_envsys(dev_t dev, int flag, int mode, struct proc *p)
  {
 -	int error = 0;
 -
  	simple_lock(&sysmon_envsys_initialized_slock);
 -	if (sysmon_envsys_initialized == 0)
 -		sysmon_envsys_init();
 +	if (sysmon_envsys_initialized == 0) {
 +		lockinit(&sysmon_envsys_lock, PWAIT|PCATCH, "smenv", 0, 0);
 +		sysmon_envsys_initialized = 1;
 +	}
 +	simple_unlock(&sysmon_envsys_initialized_slock);

 -	error = lockmgr(&sysmon_envsys_lock,
 -	    LK_EXCLUSIVE | LK_INTERLOCK |
 -	    ((flag & O_NONBLOCK) ? LK_NOWAIT : 0),
 -	    &sysmon_envsys_initialized_slock);
 -
 -	return (error);
 +	return (0);
  }

  /*
 @@ -121,7 +107,7 @@ int
  sysmonclose_envsys(dev_t dev, int flag, int mode, struct proc *p)
  {

 -	(void) lockmgr(&sysmon_envsys_lock, LK_RELEASE, NULL);
 +	/* Nothing to do */
  	return (0);
  }

 @@ -181,8 +167,11 @@ sysmonioctl_envsys(dev_t dev, u_long cmd
  			break;
  		oidx = tred->sensor;
  		tred->sensor = SME_SENSOR_IDX(sme, tred->sensor);
 -		if (tred->sensor < sme->sme_nsensors)
 +		if (tred->sensor < sme->sme_nsensors) {
 +			SYSMON_ENVSYS_LOCK();
  			error = (*sme->sme_gtredata)(sme, tred);
 +			SYSMON_ENVSYS_UNLOCK();
 +		}
  		tred->sensor = oidx;
  		sysmon_envsys_release(sme);
  		break;
 @@ -199,9 +188,11 @@ sysmonioctl_envsys(dev_t dev, u_long cmd
  		}
  		oidx = binfo->sensor;
  		binfo->sensor = SME_SENSOR_IDX(sme, binfo->sensor);
 -		if (binfo->sensor < sme->sme_nsensors)
 +		if (binfo->sensor < sme->sme_nsensors) {
 +			SYSMON_ENVSYS_LOCK();
  			error = (*sme->sme_streinfo)(sme, binfo);
 -		else
 +			SYSMON_ENVSYS_UNLOCK();
 +		} else
  			binfo->validflags = 0;
  		binfo->sensor = oidx;
  		sysmon_envsys_release(sme);
 XXX
State-Changed-From-To: open->feedback 
State-Changed-By: jdolecek 
State-Changed-When: Thu Jan 3 13:55:11 PST 2002 
State-Changed-Why:  
Sent e-mail with fix, avaiting confirmation it indeed fixes the problem. 

From: Jaromir Dolecek <jdolecek@netbsd.org>
To: Anthony Mallet <anthony.mallet@useless-ficus.net>
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/14368
Date: Sun, 13 Jan 2002 19:48:03 +0100 (CET)

 Anthony Mallet wrote:
 > I'll check your patch asap (maybe this WE).

 Had you had a chance to try it yet? It's been committed to -current
 meanwhile, so you need only update to newer -current.

 Jaromir

From: Anthony Mallet <anthony.mallet@useless-ficus.net>
To: jdolecek@netbsd.org
Cc: gnats-bugs@netbsd.org
Subject: Re: kern/14368
Date: Fri, 15 Feb 2002 14:50:26 +0100 (CET)

 Jaromir Dolecek wrote :
 | Had you had a chance to try it yet? It's been committed to -current
 | meanwhile, so you need only update to newer -current.

 I just compiled an 1.5ZA kernel this WE, and the deadlocks have
 disappeared :) Thanks a lot
  (and I apologize for the late feedback on this pr)
State-Changed-From-To: feedback->closed 
State-Changed-By: jdolecek 
State-Changed-When: Sat Feb 16 02:34:00 PST 2002 
State-Changed-Why:  
Fixed, thanks for report. 

From: Jaromir Dolecek <jdolecek@netbsd.org> (by way of Erik E. Fair)
To: NetBSD GNATS Problem Report Tracking System <gnats-bugs@gnats.netbsd.org>
Cc:  
Subject: Re: kern/14368
Date: Sat, 16 Feb 2002 10:17:10 -0800

 Anthony Mallet wrote:
 >   I just compiled an 1.5ZA kernel this WE, and the deadlocks have
 >   disappeared :) Thanks a lot

 Thanks for testing it :)

 >    (and I apologize for the late feedback on this pr)

 No problem at all.

 Jaromir
 -- 
 Jaromir Dolecek <jdolecek@NetBSD.org> http://www.NetBSD.org/Ports/i386/ps2.html
 -=  Those who would give up liberty for a little temporary safety deserve  =-
 -=  neither liberty nor safety, and will lose both.  -- Benjamin Franklin  =-
>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.