Subject: kern/36705: fast vnconfig/mount/umount/vnconfig -u panics NetBSD
To: None <kern-bug-people@netbsd.org, gnats-admin@netbsd.org,>
From: None <jan.m.danielsson@gmail.com>
List: netbsd-bugs
Date: 07/30/2007 09:30:00
>Number:         36705
>Category:       kern
>Synopsis:       fast vnconfig/mount/umount/vnconfig -u panics NetBSD
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Jul 30 09:30:00 +0000 2007
>Originator:     Jan Danielsson
>Release:        NetBSD/amd64 3.1
>Organization:
La Cosa Nostra
>Environment:
NetBSD nl102-238-202.student.uu.se 3.1 NetBSD 3.1 (ANCA) #4: Tue Jul 17 18:57:21 CEST 2007  root@nl102-238-202.student.uu.se:/usr/src/sys/arch/amd64/compile/ANCA amd64

>Description:
I have a script which builds a custom boot image. It runs:
# vnconfig -c vnd0 boot.img
[..initialize boot record and file system..]
# mount /dev/vnd0a /mnt/img
[..copy stuff to image..]
# umount /mnt/img
# vnconfig -u vnd0

When this script is run, it may cause the system to panic. I have seen this problem before (under 3.0 I blieve), and thought it was just some random behavior. But it turns out it's pretty reproducible. When it hangs, I just have time to read that vnconfig -c can not run, because the device already exists. Though I believe that it only panics on the "vnconfig -u" or the "umount".

It doesn't happen every time.

Unfortunately, I currently don't have the time to provoke the reaction to diagnose it further, but I will when I have more spare time in case someone needs be to.

>How-To-Repeat:
Apparently configure a vnd, mount, copy stuff to, unmount, and unconfigure it.

>Fix:
Not a fix, but adding a bunch of "sleep 1" makes the panics go away. I suspect that's just makes it less likely for the problem to appear though.