On Sun, Apr 06, 2003 at 10:05:33PM +0200, Frank Kardel wrote:
> On Sun, Apr 06, 2003 at 06:43:19PM +0200, Frank van der Linden wrote:
> > There was a time window during which the sequence:
> > 
> > 	* boot new kernel, write fs
> > 	* boot old kernel OR use old fsck
> > 	* boot new kernel
> > 
> > ..would get the block counts wrong, possibly creating trouble.
> > 
> > I've fixed this problem in sys/ufs/ffs_vfsops.c yesterday, I
> > suggest using a kernel with that fix. There's still a potential
> > problem with new fsck and an old kernel, I'll fix that shortly.
> Probably i picked up an on fsck during the phase with the ext2fs
> panic before i could upgrade to the new user-land.

What I'm advising everyone who ran a kernel with sys/ufs/ffs/ffs_vfsops.c
with rev. 1.110 (i.e. the version of the UFS2 commit), is to:

	* upgrade your kernel to the latest sources
	* boot single user
	* force an fsck on all filesystems
	* reboot

After that, things should be ok. I had the problem of a corrupted
filesystem, but attributed it to other problems (the machine in
question has run a lot of experimental kernels). However, the
problem went away after I went through the steps above.

- Frank

-- 
Frank van der Linden                                            fvdl@netbsd.org
===============================================================================
NetBSD. Free, Unix-like OS. > 45 different platforms.    http://www.netbsd.org/