NetBSD Problem Report #1256
From gnats Sat Jul 22 19:19:35 1995
Received: from sonyinet.sony.co.jp (sonyinet.sony.co.jp [202.24.32.17]) by sun-lamp.pc.cs.cmu.edu (8.6.11/8.6.10) with ESMTP id TAA08472 for <gnats-bugs@gnats.netbsd.org>; Sat, 22 Jul 1995 19:19:33 -0400
Message-Id: <199507220536.OAA00258@king-show.sys.ptg.sony.co.jp>
Date: Sat, 22 Jul 1995 14:36:48 +0900
From: enami tsugutomo <enami@sys.ptg.sony.co.jp>
Reply-To: enami@sys.ptg.sony.co.jp
To: gnats-bugs@gnats.netbsd.org
Subject: panic lockholder (156) != 0
X-Send-Pr-Version: 3.2
>Number: 1256
>Category: kern
>Synopsis: 3 or more simultaneous access to CD-ROM causes panic.
>Confidential: no
>Severity: serious
>Priority: low
>Responsible: kern-bug-people
>State: closed
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Sat Jul 22 19:20:01 +0000 1995
>Closed-Date: Thu Dec 03 19:23:43 +0000 1998
>Last-Modified: Thu Dec 03 19:28:12 +0000 1998
>Originator: enami tsugutomo
>Release: arround jul 7
>Organization:
Sony Corporation
>Environment:
System: NetBSD king-show 1.0A NetBSD 1.0A (KINGSHOW) #10: Sat Jul 22 14:11:11 JST 1995 root@king-show:/usr/src/sys/arch/i386/compile/KINGSHOW i386
SCSI Adaptor: Buslogic BT-445C
CD-ROM drive: ICM CD-620S or Nakamichi MBR-7.4
>Description:
If 3 or more processes access save cdrom drive simltaneously,
kernel panics with message "lockholder (XXX) != 0" at cd9660_lock.
>How-To-Repeat:
# assume that cdrom is mounted on /cd/0
find /cd/0 -print & find /cd/0 -print & find /cd/0 -print &
>Fix:
In the function cd9660_lock, it seems that we should go back
to check if someone has lock after wakeup from sleep. Otherwise,
two or more slept processes work without locking each other.
So, I tried to put `goto start;' just after the sleep,
but it seems the panic is gone, but some processes are
deadlocked now.
>Release-Note:
>Audit-Trail:
From: Chris G Demetriou <Chris_G_Demetriou@BALVENIE.PDL.CS.CMU.EDU>
To: enami@sys.ptg.sony.co.jp
Cc: gnats-bugs@gnats.netbsd.org, mycroft@netbsd.org
Subject: Re: kern/1256: panic lockholder (156) != 0
Date: Mon, 24 Jul 1995 01:00:33 -0400
> >Description:
> If 3 or more processes access save cdrom drive simltaneously,
> kernel panics with message "lockholder (XXX) != 0" at cd9660_lock.
> >How-To-Repeat:
> # assume that cdrom is mounted on /cd/0
> find /cd/0 -print & find /cd/0 -print & find /cd/0 -print &
I cannot duplicate this; it appears to be a timing-dependent problem,
and it doesn't show up for me. however, looking at the code, there is
a bug.
> >Fix:
> In the function cd9660_lock, it seems that we should go back
> to check if someone has lock after wakeup from sleep. Otherwise,
> two or more slept processes work without locking each other.
>
> So, I tried to put `goto start;' just after the sleep,
> but it seems the panic is gone, but some processes are
> deadlocked now.
looking at the code, and comparing to UFS, 'goto start;' should work.
There appears to be some similar code in 'cd9660_ihashget'. However,
looking at it, it appears to do the right thing after sleeping.
I'd be interested to know _where_ those processes are
deadlocked...
chris
From: enami tsugutomo <enami@sys.ptg.sony.co.jp>
To: Chris G Demetriou <Chris_G_Demetriou@BALVENIE.PDL.CS.CMU.EDU>
Cc: gnats-bugs@gnats.netbsd.org, mycroft@NetBSD.ORG
Subject: Re: kern/1256: panic lockholder (156) != 0
Date: Sun, 12 Nov 1995 17:13:28 +0900
Sorry for too late reply.
> looking at the code, and comparing to UFS, 'goto start;' should work.
>
> There appears to be some similar code in 'cd9660_ihashget'. However,
> looking at it, it appears to do the right thing after sleeping.
>
>
> I'd be interested to know _where_ those processes are
> deadlocked...
I got the stack trace of them.
enami.
The kernel I used is Nov 05 kernel (NetBSD/i386 1.1 ALPHA)
with adding goto start in the function cd9660_lock, like this,
===================================================================
RCS file: RCS/cd9660_vnops.c,v
retrieving revision 1.1
diff -u -r1.1 cd9660_vnops.c
--- cd9660_vnops.c 1995/11/12 07:03:53 1.1
+++ cd9660_vnops.c 1995/11/12 07:04:21
@@ -792,6 +792,7 @@
ip->i_lockwaiter = -1;
#endif
(void) sleep((caddr_t)ip, PINOD);
+ goto start;
}
#ifdef DIAGNOSTIC
ip->i_lockwaiter = 0;
Then I ran five find process concurrently and got deadlock.
With three or four process I didn't get deadlock. I dicided
they deadlocked becase they are running more than 30 minites
but four or less process are all done in a few minutets.
The CD-ROM is mounted on /cd/0, like this:
/dev/cd0a on /cd/0 type cd9660 (local, read-only)
And I did the following commands:
% cd /cd/0
% find . -print >/dev/null & find . -print >/dev/null & find . -print >/dev/null & find . -print >/dev/null & find . -print >/dev/null &
And they are running like this:
enami 336 0.0 2.0 84 316 p0 D 4:14PM 0:00.14 . (find)
enami 337 0.0 2.0 84 316 p0 D 4:14PM 0:00.14 . (find)
enami 338 0.0 2.0 84 316 p0 D 4:14PM 0:00.14 . (find)
enami 339 0.0 2.0 84 316 p0 D 4:14PM 0:00.13 . (find)
enami 340 0.0 2.0 84 316 p0 D 4:14PM 0:00.13 . (find)
The output of ps in ddb is:
db> ps
pid proc addr uid ppid pgrp flag stat em comm wchan
340 f86c1100 f9a35000 1000 120 340 004006 3 netbsd find f8732980
339 f86d9f00 f9a32000 1000 120 339 004006 3 netbsd find f8729580
338 f86d9d00 f9a2f000 1000 120 338 004006 3 netbsd find f8729580
337 f8707d00 f9a2c000 1000 120 337 004006 3 netbsd find getblk f8ca
a15c
336 f86c1b00 f9a29000 1000 120 336 004006 3 netbsd find f8729580
:
The stack trace of each process is:
db> x/x 0xf9a3503c
0xf9a3503c: f9a36bdc
db> trace 0xf9a36bdc
bpendsleep(f86fce80,2f800,f9a36c0c,f8108adc,f8732980) at bpendsleep
bpendsleep(f8732980,8) at bpendsleep
_cd9660_ihashget(600,2f800,f86fce80,f8729580,f9a36d00) at _cd9660_ihashget+0x54
_cd9660_vget_internal(f86b0400,2f800,f9a36cdc,0,f8d9b066) at _cd9660_vget_inter
nal+0x24
_cd9660_lookup(f9a36e44,f86fce80,f9a36f24,f9a36f00,f8748020) at _cd9660_lookup+
0x932
_lookup(f9a36f00,f870bf00,f9a36f00,f86c1100,241ed) at _lookup+0x257
_namei(f9a36f00,f870bf00,f86c1100,4,f81a971c) at _namei+0x16b
_sys_chroot(f9a36f00,f86c1100) at _sys_chroot+0x8d
_sys_chdir(f86c1100,f9a36f88,f9a36f80,0,9c00) at _sys_chdir+0x37
_syscall() at _syscall+0x139
--- syscall (number 12) ---
0x100395e7:
db> x/x 0xf9a3203c
0xf9a3203c: f9a33c28
db> trace 0xf9a33c28
bpendsleep(f8729580,f86fce80,f9a33c54,f810b8d2,f8729580) at bpendsleep
bpendsleep(f8729580,8) at bpendsleep
_cd9660_lock(f9a33c64,ef2b,f81a96f0,f86fce80,f9a33e14) at _cd9660_lock+0x22
_vget(f86fce80,1,f86d8780,f9a33f24,f9a33f10) at _vget+0x87
_cd9660_lookup(f9a33e44,f86d8780,f9a33f24,f9a33f00,f8748820) at _cd9660_lookup+
0x196
_lookup(f9a33f00,f86d1b00,f9a33f00,f86d9f00,241ed) at _lookup+0x257
_namei(f9a33f00,f86d1b00,f86d9f00,4,f86d1b00) at _namei+0x16b
_sys_chroot(f9a33f00,f86d9f00) at _sys_chroot+0x8d
_sys_chdir(f86d9f00,f9a33f88,f9a33f80,0,12400) at _sys_chdir+0x37
_syscall() at _syscall+0x139
--- syscall (number 12) ---
0x100395e7:
db> x/x 0xf9a2f03c
0xf9a2f03c: f9a30e30
db> trace 0xf9a30e30
bpendsleep(f8729580,f86fce80,f9a30e5c,f810b8d2,f8729580) at bpendsleep
bpendsleep(f8729580,8) at bpendsleep
_cd9660_lock(f9a30f14) at _cd9660_lock+0x22
_sys_getdirentries(f86d9d00,f9a30f88,f9a30f80,0,7040) at _sys_getdirentries+0xc
5
_syscall() at _syscall+0x139
--- syscall (number 196) ---
0x1004e02b:
db> x/x 0xf9a2c03c
0xf9a2c03c: f9a2db34
db> trace 0xf9a2db34
bpendtsleep(f8caa15c,0,f86a3c74,0,f9a2db70) at bpendtsleep
bpendtsleep(f8caa15c,11,f8127d9d,0,f9a2dc48) at bpendtsleep
_getblk(f86fce80,0,800,0,0) at _getblk+0x7d
_bread(f86fce80,0,800,ffffffff,f9a2dbd0) at _bread+0x21
_cd9660_blkatoff(f9a2dc48,f86fce80,f9a2de64,f9a2de50,32000) at _cd9660_blkatoff
+0x4c
_cd9660_lookup(f9a2dda8,f86fce80,f9a2de64,f9a2de40,f86eb820) at _cd9660_lookup+
0x3f7
_lookup(f9a2de40,f7bfdaac,f8707d00,8,f8748000) at _lookup+0x257
_namei(f9a2de40,f7bfdaac,f8707d00,8,f9a2df88) at _namei+0x16b
_sys_lstat(f8707d00,f9a2df88,f9a2df80,0,f7bfdac8) at _sys_lstat+0x50
_syscall() at _syscall+0x139
--- syscall (number 190) ---
0x10041fa3:
db> x/x 0xf9a2903c
0xf9a2903c: f9a2ab5c
db> trace 0xf9a2ab5c
bpendsleep(f8729580,f86fce80,f9a2ab88,f810b8d2,f8729580) at bpendsleep
bpendsleep(f8729580,8) at bpendsleep
_cd9660_lock(f9a2ab98,ef2b,f81a96f0,f86fce80,f9a2ad48) at _cd9660_lock+0x22
_vget(f86fce80,1,f8709b80,f9a2af0c,f9a2aef8) at _vget+0x87
_cd9660_lookup(f9a2ad78,f8709b80,f9a2af0c,f9a2aee8,f8748420) at _cd9660_lookup+
0x1f5
_lookup(f9a2aee8,0,138c0,f9a2aee8,f8732980) at _lookup+0x257
_namei(f9a2aee8,0,138c0,f8722d40,f874e000) at _namei+0x16b
_vn_open(f9a2aee8,5,60,f7bfd990,f86c1b00) at _vn_open+0x17c
_sys_open(f86c1b00,f9a2af88,f9a2af80,0,13880) at _sys_open+0x91
_syscall() at _syscall+0x139
--- syscall (number 5) ---
0x10055383:
db> c
State-Changed-From-To: open->closed
State-Changed-By: bouyer
State-Changed-When: Thu Dec 3 11:23:43 PST 1998
State-Changed-Why:
I can't reproduce it any more ...
>Unformatted:
(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.