Subject: Re: Nightmare File Corruptions with 1.6Q - FFS?
To: None <>
From: Frank van der Linden <>
List: current-users
Date: 04/12/2003 23:11:07
On Sat, Apr 12, 2003 at 10:34:36AM +0200, Martin Husemann wrote:
> On Sat, Apr 12, 2003 at 01:15:04AM -0700, Greywolf wrote:
> > Most systems are probably not affected because most people were afraid
> > of exactly what's happening to the few, the proud, the silly who
> > ate the upgrade.

> It did not happen to me. Neither did it happen to Frank, who thoroughly tested
> the code before the commit.

There's basically a matrix of all different parameters to try, and I tried
quite a few of them (all of them I thought), doing various tests, and running
builds on the filesystems.

Unfortunately the problem people saw was kind of obscure, and appeared more
quickly with certain block/fragment sizes. I tested a few different
block and fragment sizes, but not the combination that made the problem
appear easily. So I never saw the problem.

I believe it to be fixed now. If you have an up to date revision of
sys/ufs/ffs/ffs_vfsops.c, things should be ok. Although you may have
to run an fsck -fy first to get rid of the problem (after booting
the new kernel).

- Frank

Frank van der Linden                                  
NetBSD. Free, Unix-like OS. > 45 different platforms.