NetBSD Problem Report #24551

Received: (qmail 8363 invoked by uid 605); 24 Feb 2004 19:48:16 -0000
Message-Id: <20040224194812.AF95D21037@smistad.uninett.no>
Date: Tue, 24 Feb 2004 20:48:12 +0100 (CET)
From: he@netbsd.org
Sender: gnats-bugs-owner@NetBSD.org
Reply-To: he@netbsd.org
To: gnats-bugs@gnats.netbsd.org
Subject: switching with held simple_lock in sys_pipe.c on i386/MP system
X-Send-Pr-Version: 3.95

>Number:         24551
>Category:       kern
>Synopsis:       switching with held simple_lock in sys_pipe.c on i386/MP system
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    jdolecek
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Feb 24 19:49:00 +0000 2004
>Closed-Date:    Thu Feb 26 08:16:51 +0000 2004
>Last-Modified:  Thu Feb 26 22:57:00 +0000 2004
>Originator:     Havard Eidnes
>Release:        NetBSD 1.6ZK Feb 24 2004 (i.e. current sources)
>Organization:

>Environment:
System: NetBSD splitter-pine.urc.uninett.no 1.6ZK NetBSD 1.6ZK (SPLITTER-PINE) #1: Tue Feb 24 18:58:03 CET 2004  he@splitter-pine.urc.uninett.no:/usr/obj/sys/arch/i386/compile/SPLITTER-PINE i386
Architecture: i386
Machine: i386
>Description:
	A GENERIC.MP + LOCKDEBUG + DIAGNOSTIC kernel just paniced with

switching with held simple_lock 0xc24b64cc CPU 3 /usr/src/sys/kern/sys_pipe.c:359

	Console log from panic'ed system:

NetBSD/i386 (splitter-pine.urc.uninett.no) (console)

login: Feb 24 19:04:03 splitter-pine getty[537]: /dev/ttyE0: Device not configured

switching with held simple_lock 0xc24b64cc CPU 3 /usr/src/sys/kern/sys_pipe.c:359
simple_lock_switchcheck(cda7a214,0,cdab5d9c,c0362585,c07816b4) at netbsd:simple_lock_switchcheck+0x15
mi_switch(cda7a214,0,c037096c,cda7a214,0) at netbsd:mi_switch+0x4e
ltsleep(c07e1f64,10,c06e7fab,64,0) at netbsd:ltsleep+0x47b
pipelock(c24b64cc,0,c24b6268,1,0) at netbsd:pipelock+0x7f
pipe_direct_write(cdb4f4b8,c24b64cc,cdab5ec4,c2224000,c24b64e8) at netbsd:pipe_direct_write+0x24e
pipe_write(cdb4f4b8,cdb4f4e0,cdab5ec4,c2523200,1) at netbsd:pipe_write+0x197
dofilewrite(cda7ec88,6,cdb4f4b8,8057000,2000) at netbsd:dofilewrite+0x85
sys_write(cda7a214,cdab5f64,cdab5f5c,4,0) at netbsd:sys_write+0x6f
syscall_plain(cdab5fa8,1f,1f,1f,bfbf001f) at netbsd:syscall_plain+0x17e
Stopped in pid 294.1 (scp) at   netbsd:cpu_Debugger+0x4:        leave
db{3}> machine cpu 0
using CPU 0
db{3}> trace
__cpu_simple_lock(c07816b4,ccb51a7c,c2224800,c0100f66,ccb284a4) at netbsd:__cpu_simple_lock+0x1d
_simple_lock(c07816b4,c07362c0,475,c2224800,c2224800) at netbsd:_simple_lock+0x74
schedclock(ccb284a4,ccb284a4,c0100f66,ccb51acc,ccb51acc) at netbsd:schedclock+0x58
hardclock(ccb51ae0,d,c07cbb80,ccb284a4,2562) at netbsd:hardclock+0x593
lapic_clockintr(0,0,10,30,400010) at netbsd:lapic_clockintr+0x48
Xintr_lapic_ltimer() at netbsd:Xintr_lapic_ltimer+0x76
kernel: page fault trap, code=0
Faulted in DDB; continuing...
db{3}> show reg
ds          0x10
es          0x10
fs          0x30
gs          0x10
edi         0x475
esi         0xc
ebp         0xccb51a34  usb_all_tasks+0xc353ddc
ebx         0xc07816b4  sched_lock
edx         0xc07816b4  sched_lock
ecx         0xc
eax         0x1
eip         0xc040d169  __cpu_simple_lock+0x1d
cs          0x8
eflags      0x246
esp         0xcc2bbfbc  usb_all_tasks+0xbabe364
ss          0xcc2b0010  usb_all_tasks+0xbab23b8
netbsd:__cpu_simple_lock+0x1d:  decl    %eax
db{3}> machine cpu 1
using CPU 1
db{3}> trace
__cpu_simple_lock(c07816b4,cdaefc8c,cdb1be18,c03715e4,c07816b4) at netbsd:__cpu_simple_lock+0x1d
_simple_lock(c07816b4,c07362c0,475,c2222000,c2222000) at netbsd:_simple_lock+0x74
schedclock(cda7a8c8,c0784980,c0740d00,94,ffffffff) at netbsd:schedclock+0x58
hardclock(cdb1be94,403ba13f,bda42,0,0) at netbsd:hardclock+0x593
lapic_clockintr(0,0,c07e0010,cdb10030,c0370010) at netbsd:lapic_clockintr+0x48
Xintr_lapic_ltimer() at netbsd:Xintr_lapic_ltimer+0x76
--- interrupt ---
0:
db{3}> show reg
ds          0x10
es          0x10
fs          0x30
gs          0x10
edi         0x475
esi         0xc
ebp         0xcdb1bde8  usb_all_tasks+0xd31e190
ebx         0xc07816b4  sched_lock
edx         0xc07816b4  sched_lock
ecx         0xc
eax         0x1
eip         0xc040d169  __cpu_simple_lock+0x1d
cs          0x8
eflags      0x246
esp         0xcc29ffbc  usb_all_tasks+0xbaa2364
ss          0xcc290010  usb_all_tasks+0xba923b8
netbsd:__cpu_simple_lock+0x1d:  decl    %eax
db{3}> machine cpu 2
using CPU 2
db{3}> trace
acquire(c07e0d20,cdb09d9c,400000,0,600) at netbsd:acquire+0x49
_spinlock_acquire_count(c07e0d20,1,c07362c0,3d2,c24b64e8) at netbsd:_spinlock_acquire_count+0x99
mi_switch(cda7a73c,0,330,0,c24b64cc) at netbsd:mi_switch+0x1a7
preempt(1,c040e640,cdb09e60,c041a683,c0784980) at netbsd:preempt+0x75
uiomove(cdbf3000,2000,cdb09ec4,0,0) at netbsd:uiomove+0x146
pipe_read(cdb4f788,cdb4f7b0,cdb09ec4,c2523200,1) at netbsd:pipe_read+0x1fd
dofileread(cda7e8f8,0,cdb4f788,bfbfd310,2000) at netbsd:dofileread+0x80
sys_read(cda7a73c,cdb09f64,cdb09f5c,3,0) at netbsd:sys_read+0x6c
syscall_plain(cdb09fa8,80a001f,1f,1f,bfbf001f) at netbsd:syscall_plain+0x17e
db{3}> show reg
ds          0x10
es          0x10
fs          0x30
gs          0x10
edi         0
esi         0x335f4fe
ebp         0xcdb09d6c  usb_all_tasks+0xd30c114
ebx         0xc07e0d20  kernel_lock
edx         0x80000000
ecx         0xd
eax         0x480400
eip         0xc0360e1d  acquire+0x49
cs          0x8
eflags      0x246
esp         0xcc2d7fbc  usb_all_tasks+0xbada364
ss          0xcc2d0010  usb_all_tasks+0xbad23b8
netbsd:acquire+0x49:    incl    %esi
db{3}> machine cpu 3
using CPU 3
db{3}> trace
cpu_Debugger(c07816b4,c06e6fd6,cdab5d9c,c03713f9,d) at netbsd:cpu_Debugger+0x4
simple_lock_switchcheck(cda7a214,0,cdab5d9c,c0362585,c07816b4) at netbsd:simple_lock_switchcheck+0x15
mi_switch(cda7a214,0,c037096c,cda7a214,0) at netbsd:mi_switch+0x4e
ltsleep(c07e1f64,10,c06e7fab,64,0) at netbsd:ltsleep+0x47b
pipelock(c24b64cc,0,c24b6268,1,0) at netbsd:pipelock+0x7f
pipe_direct_write(cdb4f4b8,c24b64cc,cdab5ec4,c2224000,c24b64e8) at netbsd:pipe_direct_write+0x24e
pipe_write(cdb4f4b8,cdb4f4e0,cdab5ec4,c2523200,1) at netbsd:pipe_write+0x197
dofilewrite(cda7ec88,6,cdb4f4b8,8057000,2000) at netbsd:dofilewrite+0x85
sys_write(cda7a214,cdab5f64,cdab5f5c,4,0) at netbsd:sys_write+0x6f
syscall_plain(cdab5fa8,1f,1f,1f,bfbf001f) at netbsd:syscall_plain+0x17e
db{3}> show reg
ds          0x10
es          0x10
fs          0x30
gs          0x10
edi         0xcda7ec88  usb_all_tasks+0xd281030
esi         0
ebp         0xcdab5d3c  usb_all_tasks+0xd2b80e4
ebx         0xc078c800  cpu_info_primary
edx         0xc06e5f25  copyright+0x10a45
ecx         0xc0821ec0  usb_all_tasks+0x24268
eax         0x1
eip         0xc04129c0  cpu_Debugger+0x4
cs          0x8
eflags      0x202
esp         0xcdab5d3c  usb_all_tasks+0xd2b80e4
ss          0x10
netbsd:cpu_Debugger+0x4:        leave
db{3}> call simple_lock_dump
all simple locks:
0xc24b64cc CPU 3 /usr/src/sys/kern/sys_pipe.c:359
0xc07816b4 CPU 3 /usr/src/sys/kern/kern_synch.c:421
0xcdac1804 CPU 0 /usr/src/sys/miscfs/genfs/genfs_vnops.c:1180
0xc078c800
db{3}> reboot 4

>How-To-Repeat:
	I'm not sure what triggers this.

>Fix:
	I don't know how to fix this, but I think it is clear that the
	sequence of first doing simple_lock() and then doing tsleep()
	is what triggers this check with LOCKDEBUG.
>Release-Note:
>Audit-Trail:
Responsible-Changed-From-To: kern-bug-people->jdolecek 
Responsible-Changed-By: jdolecek 
Responsible-Changed-When: Wed Feb 25 08:07:08 UTC 2004 
Responsible-Changed-Why:  
I'd look at this. 

From: jdolecek@NetBSD.org (Jaromir Dolecek)
To: Havard Eidnes <he@netbsd.org>
Cc: pk@netbsd.org, fvdl@netbsd.org, gnats-bugs@gnats.netbsd.org
Subject: Re: kern/24551: switching with held simple_lock in sys_pipe.c on i386/MP
 system
Date: Wed, 25 Feb 2004 09:17:53 +0100 (CET)

 There is obvious bug in pipelock() - the simplelock
 is not released when tsleep()ing. Following should
 fix the problem.

 Index: sys_pipe.c
 ===================================================================
 RCS file: /cvsroot/src/sys/kern/sys_pipe.c,v
 retrieving revision 1.48
 diff -u -p -r1.48 sys_pipe.c
 --- sys_pipe.c	22 Feb 2004 17:51:25 -0000	1.48
 +++ sys_pipe.c	25 Feb 2004 08:15:15 -0000
 @@ -376,7 +376,8 @@ pipelock(pipe, catch)
  		 * interruptable at the start of pipe_read/pipe_write to be
  		 * beneficial.
  		 */
 -		(void) tsleep(&lbolt, PRIBIO, "rstrtpipelock", hz);
 +		(void) ltsleep(&lbolt, PRIBIO, "rstrtpipelock", hz,
 +		    &pipe->pipe_slock);
  	}
  	return (error);
  }

 -- 
 Jaromir Dolecek <jdolecek@NetBSD.org>            http://www.NetBSD.cz/
 -=- We should be mindful of the potential goal, but as the Buddhist -=-
 -=- masters say, ``You may notice during meditation that you        -=-
 -=- sometimes levitate or glow.   Do not let this distract you.''   -=-

From: Havard Eidnes <he@netbsd.org>
To: gnats-bugs@gnats.netbsd.org
Cc:  
Subject: Re: kern/24551: switching with held simple_lock in sys_pipe.c on
 i386/MP system
Date: Wed, 25 Feb 2004 10:10:34 +0100 (CET)

 Hi,

 I added a printf of the error code returned by lockmgr just
 before the tsleep() in pipelock(), and it tripped again, and the
 error code was == -3.

 Regards,

 - Havard
State-Changed-From-To: open->feedback 
State-Changed-By: jdolecek 
State-Changed-When: Wed Feb 25 12:39:19 UTC 2004 
State-Changed-Why:  
Patch supplied, please test and let me know if it's okay with the patch. 

From: Havard Eidnes <he@netbsd.org>
To: jdolecek@NetBSD.org
Cc: pk@NetBSD.org, fvdl@NetBSD.org, gnats-bugs@gnats.netbsd.org
Subject: Re: kern/24551: switching with held simple_lock in sys_pipe.c on
 i386/MP system
Date: Wed, 25 Feb 2004 14:10:45 +0100 (CET)

 > There is obvious bug in pipelock() - the simplelock
 > is not released when tsleep()ing. Following should
 > fix the problem.

 Thanks, I've applied your patch and I'm currently running what
 used to trigger it (I think): an scp of a 3GB large file to
 ftp.netbsd.org.

 Regards,

 - H=E5vard

From: Frank van der Linden <fvdl@netbsd.org>
To: Jaromir Dolecek <jdolecek@NetBSD.org>
Cc: Havard Eidnes <he@netbsd.org>, pk@netbsd.org, gnats-bugs@gnats.netbsd.org,
  blymn@netbsd.org
Subject: Re: kern/24551: switching with held simple_lock in sys_pipe.c on i386/MP system
Date: Wed, 25 Feb 2004 14:28:53 +0100

 On Wed, Feb 25, 2004 at 09:17:53AM +0100, Jaromir Dolecek wrote:
 > There is obvious bug in pipelock() - the simplelock
 > is not released when tsleep()ing. Following should
 > fix the problem.
 > 
 > Index: sys_pipe.c
 > ===================================================================
 > RCS file: /cvsroot/src/sys/kern/sys_pipe.c,v
 > retrieving revision 1.48
 > diff -u -p -r1.48 sys_pipe.c
 > --- sys_pipe.c	22 Feb 2004 17:51:25 -0000	1.48
 > +++ sys_pipe.c	25 Feb 2004 08:15:15 -0000
 > @@ -376,7 +376,8 @@ pipelock(pipe, catch)
 >  		 * interruptable at the start of pipe_read/pipe_write to be
 >  		 * beneficial.
 >  		 */
 > -		(void) tsleep(&lbolt, PRIBIO, "rstrtpipelock", hz);
 > +		(void) ltsleep(&lbolt, PRIBIO, "rstrtpipelock", hz,
 > +		    &pipe->pipe_slock);
 >  	}
 >  	return (error);
 >  }
 > 

 Looks good to me. I think this may be the issue Brett has been seeing
 on his MP amd64 system, so I'm cc-ing him; perhaps he can try to see
 if his problems go away when he applies this patch.

 - Frank

From: Jaromir Dolecek <jdolecek@netbsd.org>
To: gnats-bugs@gnats.NetBSD.org
Cc:  
Subject: pr/24551 CVS commit: src/sys/kern
Date: Thu, 26 Feb 2004 08:15:31 +0000 (UTC)

 Module Name:	src
 Committed By:	jdolecek
 Date:		Thu Feb 26 08:15:31 UTC 2004

 Modified Files:
 	src/sys/kern: sys_pipe.c

 Log Message:
 pipelock() must release the pipe simplelock during tsleep()
 fixes PR kern/24551 by Havard Eidnes


 To generate a diff of this commit:
 cvs rdiff -r1.50 -r1.51 src/sys/kern/sys_pipe.c

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

State-Changed-From-To: feedback->closed 
State-Changed-By: jdolecek 
State-Changed-When: Thu Feb 26 08:16:32 UTC 2004 
State-Changed-Why:  
Confirmed fixed. Thanks for report. 

From: Brett Lymn <blymn@baesystems.com.au>
To: Frank van der Linden <fvdl@netbsd.org>
Cc: Jaromir Dolecek <jdolecek@netbsd.org>, Havard Eidnes <he@netbsd.org>,
  pk@netbsd.org, gnats-bugs@gnats.netbsd.org, blymn@netbsd.org
Subject: Re: kern/24551: switching with held simple_lock in sys_pipe.c on i386/MP system
Date: Fri, 27 Feb 2004 09:26:15 +1030

 On Wed, Feb 25, 2004 at 02:28:53PM +0100, Frank van der Linden wrote:
 > 
 > Looks good to me. I think this may be the issue Brett has been seeing
 > on his MP amd64 system, so I'm cc-ing him; perhaps he can try to see
 > if his problems go away when he applies this patch.
 > 

 Unfortunately for me, no that did not fix my problem.  I think I have
 tracked down what my problem is - I will send you a separate mail with
 the details Frank to save spamming everyone :)

 -- 
 Brett Lymn
>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.