NetBSD Problem Report #41725

From apb@cequrux.com  Tue Jul 14 18:57:36 2009
Return-Path: <apb@cequrux.com>
Received: from mail.netbsd.org (mail.netbsd.org [204.152.190.11])
	by www.NetBSD.org (Postfix) with ESMTP id 8184563BC2A
	for <gnats-bugs@gnats.NetBSD.org>; Tue, 14 Jul 2009 18:57:36 +0000 (UTC)
Message-Id: <20090714185417.E0802100AD57@apb-laptoy.apb.alt.za>
Date: Tue, 14 Jul 2009 18:54:17 +0000 (UTC)
From: apb@cequrux.com
Reply-To: apb@cequrux.com
To: gnats-bugs@gnats.NetBSD.org
Subject: uvm fault and problems unmounting file sytems at shutdown
X-Send-Pr-Version: 3.95

>Number:         41725
>Category:       kern
>Synopsis:       uvm fault and problems unmounting file sytems at shutdown
>Confidential:   no
>Severity:       critical
>Priority:       high
>Responsible:    dyoung
>State:          closed
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Tue Jul 14 19:00:00 +0000 2009
>Closed-Date:    Wed Feb 10 08:58:29 +0000 2010
>Last-Modified:  Wed Feb 10 08:58:29 +0000 2010
>Originator:     Alan Barrett
>Release:        NetBSD 5.99.15
>Organization:
Not much
>Environment:
System: NetBSD 5.99.11 i386
>Description:
        During an attempt to shutdown, the system crashes with a uvm
        fault, and appears to detach devices before unmounting file
        sytems that use those devices.

>How-To-Repeat:
	The root file system is a memory disk, provided using a custom
	miniroot.kmod, loaded from the boot loaded via "load /miniroot.kmod".

        The memory disk file system contains an /etc/rc script that
        configures a cgd device, mounts an ffs+wapbl file system from
        the cgd device, sets sysctl -w init.chroot=/cgd1a, and exits.
        All userland activity then proceeds inside the chroot.

	Press the power button.

	The event is passed to powerd, which performs "/sbin/shutdown -p
	now".

	shutdown(8) executes all the rc.d shutdown scripts.  Note that
	this is still inside the chroot.

	The kernel detaches all or most devices, ending with:

	ppb2: detached
	ppb1: detached
	ppb0: detached
	audio1: detached
	md0: detached

	unmounting 0xd0358605 /cgd1a/ (/dev/cgd1a)...
	unmounting 0xd022f770 / (/dev/md0a)...
	forcefully unmounting /cgd1a/ (/dev/cgd1a)...forcefully unmounted
	/dev/cgd1a on /cgd1a/ type ffs

	unmounting 0xd022f770 / (/dev/md0a)...
	forcefully unmounting / (/dev/md0a)...uvm_fault(0xd07606a4, 0, 1)
	-> 0xe
	fatal page fault in supervisor mode
	trap type 6 code 0 eip c0521e4c cs 8 eflags 10246 cr2 60 ilevel 0
	kernel: supervisor trap page fault, code=0
	stopped in pid 10670.1 (halt) at netbsd:mdstrategy+0x3c: movl
	0x60(%eax),%eax
	db{0}> bt
	mdstrategy ...
	bdev_strategy ...
	spec_strategy ...
	VOP_STRATEGY ...
	bwrite ...
	ffs_sbupdate ...
	ffs_cgupdate ...
	ffs_unmount ...
	VFS_UNMOUNT ...
	dounmount ...
	vfs_unmount_forceone ...
	cpu_reboot ...
	sys_reboot ...
	syscall ...

>Fix:
        Unknown.  There seem to be at least three things wrong with
        this picture:  (1) why was md0 detached before the file system
        on md0a was unmounted, (2) why was it necessary to forecefully
        unmount the cgd1a and md0a file systems instead of gracefully
        unmounting them, and (3) why was there a uvm fault?

        I don't know whether cgd1 was detached too early; if it was,
        then the relevant line had scrolled off the screen before I saw
        the problem.

>Release-Note:

>Audit-Trail:

Responsible-Changed-From-To: kern-bug-people->dyoung
Responsible-Changed-By: dyoung@NetBSD.org
Responsible-Changed-When: Sat, 25 Jul 2009 23:26:55 +0000
Responsible-Changed-Why:
I see the bug in md(4).  I know how to fix it.


State-Changed-From-To: open->feedback
State-Changed-By: dyoung@NetBSD.org
State-Changed-When: Mon, 08 Feb 2010 23:23:25 +0000
State-Changed-Why:
I think that this bug is fixed.  Is it, Alan?

Dave


State-Changed-From-To: feedback->closed
State-Changed-By: apb@NetBSD.org
State-Changed-When: Wed, 10 Feb 2010 08:58:29 +0000
State-Changed-Why:
This problem was fixed in 2009.


>Unformatted:

Home	PR Database Search