Date: Sun, 1 Mar 1998 16:25:09 +0100 (CET) From: Andre Albsmeier <andre.albsmeier@mchp.siemens.de> To: FreeBSD-gnats-submit@FreeBSD.ORG Subject: kern/5886: 2.2-STABLE crashes when unmounting a busy ufs fs Message-ID: <199803011525.QAA25028@intern>
next in thread | raw e-mail | index | archive | help
>Number: 5886
>Category: kern
>Synopsis: 2.2-STABLE crashes when unmounting a busy ufs fs
>Confidential: no
>Severity: critical
>Priority: high
>Responsible: freebsd-bugs
>State: open
>Quarter:
>Keywords:
>Date-Required:
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Sun Mar 1 07:30:01 PST 1998
>Last-Modified:
>Originator: Andre Albsmeier
>Organization:
>Release: FreeBSD 2.2.5-STABLE i386
>Environment:
FreeBSD 2.2.5-STABLE, recent build
>Description:
let's assume the following (done as root):
mount /dev/sd3c /mnt
tar cf /mnt/bla.tar some_big_directory
While the tar is running we do the following in another shell:
umount /mnt
If we see a "Device busy" error we just repeat the umount again and again.
Suddenly the system freezes for about 10 seconds and reboots. During the
bootstrap the following message appears:
bali daemon.alert savecore: reboot after panic: vinvalbuf: dirty bufs
All that happens on a FreeBSD 2.2.5-STABLE system.
Here is a mail that I received after posting the problem to the mailinglists:
----------------------- snip --------------------------------------
Return-Path: <j@bug.fe.up.pt>
Date: Fri, 27 Feb 1998 20:21:09 +0000 (WET)
From: freebsd@bug.fe.up.pt
X-Sender: jmg@slug.EUnet.pt
cc: freebsd-questions@FreeBSD.ORG, hackers@FreeBSD.ORG, jkh@time.cdrom.com
Subject: Re: 2.2.5-STABLE: is this a bug in umount?
In-Reply-To: <199802271717.SAA00672@intern>
Message-ID: <Pine.BSF.3.96.980227201611.251A-200000@slug.EUnet.pt>
MIME-Version: 1.0
Hi!
I did a kernel debug on it and here is the log of the debug (I've
included an attachment with the full log):
(kgdb) down
#9 0xf01318fc in unmount (p=0xf1509800, uap=0xefbfff94, retval=0xefbfff84)
at ../../kern/vfs_syscalls.c:280
280 return (dounmount(mp, uap->flags, p));
(kgdb) list
275 * Don't allow unmount of the root filesystem
276 */
277 if (mp->mnt_flag & MNT_ROOTFS)
278 return (EINVAL);
279
280 return (dounmount(mp, uap->flags, p));
281 }
282
283 /*
284 * Do the actual file system unmount.
(kgdb) exit
The solution is to add an if before the return checking if there is
activity in that partition. The way to do that I'll leave to the
core team or someone with commit priviliges.
Jorge
---------- and the log ------------------------------------
Script started on Fri Feb 27 20:03:53 1998
slug# cd /sys/compile/SLUG
slug# gdb -k
GDB is free software and you are welcome to distribute copies of it
under certain conditions; type "show copying" to see the conditions.
There is absolutely no warranty for GDB; type "show warranty" for details.
GDB 4.16 (i386-unknown-freebsd), Copyright 1996 Free Software Foundation, Inc.
(kgdb) symbol-file kernel.debug
Reading symbols from kernel.debug...done.
(kgdb) exec-file /var/crash/kernel.1
(kgdb) core-file /var/crash/vmcore.1
IdlePTD 1de000
current pcb at 1c2718
panic: vinvalbuf: dirty bufs
#0 boot (howto=256) at ../../kern/kern_shutdown.c:266
266 dumppcb.pcb_cr3 = rcr3();
(kgdb) where
#0 boot (howto=256) at ../../kern/kern_shutdown.c:266
#1 0xf0111f32 in panic (fmt=0xf012feb9 "vinvalbuf: dirty bufs")
at ../../kern/kern_shutdown.c:390
#2 0xf012ff64 in vinvalbuf (vp=0xf15a1880, flags=1, cred=0x0, p=0x0,
slpflag=0, slptimeo=0) at ../../kern/vfs_subr.c:481
#3 0xf016f89d in vm_object_terminate (object=0xf15f3f00)
at ../../vm/vm_object.c:397
#4 0xf016f74b in vm_object_deallocate (object=0xf15f3f00)
at ../../vm/vm_object.c:353
#5 0xf0130889 in vflush (mp=0xf152bc00, skipvp=0x0, flags=0)
at ../../kern/vfs_subr.c:989
#6 0xf015f4dc in ffs_flushfiles (mp=0xf152bc00, flags=0, p=0xf1509800)
at ../../ufs/ffs/ffs_vfsops.c:725
#7 0xf015f3aa in ffs_unmount (mp=0xf152bc00, mntflags=0, p=0xf1509800)
at ../../ufs/ffs/ffs_vfsops.c:664
#8 0xf01319ca in dounmount (mp=0xf152bc00, flags=0, p=0xf1509800)
at ../../kern/vfs_syscalls.c:312
#9 0xf01318fc in unmount (p=0xf1509800, uap=0xefbfff94, retval=0xefbfff84)
at ../../kern/vfs_syscalls.c:280
#10 0xf0181933 in syscall (frame={tf_es = 39, tf_ds = 39, tf_edi = -272640576,
tf_esi = 168358, tf_ebp = -272639440, tf_isp = -272629788, tf_ebx = 0,
tf_edx = 1, tf_ecx = 58, tf_eax = 22, tf_trapno = 7, tf_err = 7,
tf_eip = 11541, tf_cs = 31, tf_eflags = 582, tf_esp = -272640628,
tf_ss = 39}) at ../../i386/i386/trap.c:914
---Type <return> to continue, or q <return> to quit---
#11 0x2d15 in ?? ()
#12 0x12d3 in ?? ()
#13 0x107e in ?? ()
(kgdb) down 10
#0 boot (howto=256) at ../../kern/kern_shutdown.c:266
266 dumppcb.pcb_cr3 = rcr3();
(kgdb) up 10
#10 0xf0181933 in syscall (frame={tf_es = 39, tf_ds = 39, tf_edi = -272640576,
tf_esi = 168358, tf_ebp = -272639440, tf_isp = -272629788, tf_ebx = 0,
tf_edx = 1, tf_ecx = 58, tf_eax = 22, tf_trapno = 7, tf_err = 7,
tf_eip = 11541, tf_cs = 31, tf_eflags = 582, tf_esp = -272640628,
tf_ss = 39}) at ../../i386/i386/trap.c:914
914 error = (*callp->sy_call)(p, args, rval);
(kgdb) up
#11 0x2d15 in ?? ()
(kgdb) down
#10 0xf0181933 in syscall (frame={tf_es = 39, tf_ds = 39, tf_edi = -272640576,
tf_esi = 168358, tf_ebp = -272639440, tf_isp = -272629788, tf_ebx = 0,
tf_edx = 1, tf_ecx = 58, tf_eax = 22, tf_trapno = 7, tf_err = 7,
tf_eip = 11541, tf_cs = 31, tf_eflags = 582, tf_esp = -272640628,
tf_ss = 39}) at ../../i386/i386/trap.c:914
914 error = (*callp->sy_call)(p, args, rval);
(kgdb) down
#9 0xf01318fc in unmount (p=0xf1509800, uap=0xefbfff94, retval=0xefbfff84)
at ../../kern/vfs_syscalls.c:280
280 return (dounmount(mp, uap->flags, p));
(kgdb) list
275 * Don't allow unmount of the root filesystem
276 */
277 if (mp->mnt_flag & MNT_ROOTFS)
278 return (EINVAL);
279
280 return (dounmount(mp, uap->flags, p));
281 }
282
283 /*
284 * Do the actual file system unmount.
(kgdb) exit
slug#
Script done on Fri Feb 27 20:13:59 1998
>How-To-Repeat:
mount a ufs drive (in my case a jaz drive)
do something on it (here a tar which writes a big archive)
unmount it while it's busy (maybe have to try multiple times)
>Fix:
Maybe the stuff above helps
>Audit-Trail:
>Unformatted:
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-bugs" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199803011525.QAA25028>