Subject: kern/9483: LKM interface serious broken
To: None <gnats-bugs@gnats.netbsd.org>
From: Matthias Scheler <tron@lyssa.zhadum.de>
List: netbsd-bugs
Date: 02/24/2000 14:51:45
>Number:         9483
>Category:       kern
>Synopsis:       LKM interface serious broken
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people (Kernel Bug People)
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Feb 24 14:50:59 2000
>Last-Modified:
>Originator:     Matthias Scheler
>Organization:
Matthias Scheler                            http://www.sighardstrasse.de/~tron/
>Release:        2000/02/24
>Environment:
System: NetBSD lyssa.zhadum.de 1.4T NetBSD 1.4T (LYSSA) #1: Thu Feb 24 18:17:13 CET 2000 tron@lyssa.zhadum.de:/src/sys/compile/LYSSA i386


>Description:
For some time I'm experiencing strange problems on my current system:

1.) My system crashes in soft dependence code (see PR kern/9446) while
    it seems to work fine on other peoples' systems.

2.) On one out of three reboots I get a message like this ...

Feb 24 03:59:08 lyssa mountd[322]: Can't delete exports for /tmp

    ... and afterwards the affected filesystem seems to be empty. After
    another reboot all the files and directories are back again unless
    the above problem occured again.

3.) With very recent kernel sources it gets worse. The system always gets
    stuck on the final "mount -a" and I have to press CTRL-C to get into
    multi user mode. After the system is up for a few minutes the contents
    of complete filesystems vanishes e.g. my biff program terminates
    because the "/var" filesystem is complete empty. If I try to umount
    such a filesystem I get an EINVAL. After a reboot to single user mode
    the contents of the filesystem is back.

Because the problems always started after mounting LKM filesystems I built
a kernel with support for all required filesystems and disabled the LKMs
in "/etc/lkm.conf". With this kernel all of the above problems are gone.
I suppose there is a serious bug in the current LKM handling which
caused some kind of memory corruption in the kernel.

>How-To-Repeat:
1.) Compile a kernel without IP filter, MFS, KERNFS and PROCFS support.
2.) Load IP filter and the above filesystems as LKMs.
3.) Try to use those filesystems.

>Fix:
Sorry, I don't now.
>Audit-Trail:
>Unformatted: