Subject: Re: Interesting disk/filesystem anomaly addendum
To: Bri <itai@replic.net>
From: Luke Mewburn <lukem@wasabisystems.com>
List: port-i386
Date: 12/27/2001 19:34:21
On Thu, Dec 27, 2001 at 12:23:59AM -0800, Bri wrote:
  |  Interestingley, yes. There is about 360Kbyte/sec of activity on the disk
  | with 99/100% usage according to systat.
  | 
  |  As far as how nested, it isn't. I was making the directory right in the
  | top level of where it was mounted. The root of that drive has 4 folders
  | including the test one that I used to get the time.


What does "dumpfs" on the raw device return (e.g, "dumpfs /dev/rwd2a") ?

There is known pathological behaviour in directory creation on
(large) partitions with certain parameters. This was fixed in
NetBSD-current earlier this year, and that fix was subsequently
replaced by the "new" ffs_dirpref() routine. The latter code
has been pulled up to the NetBSD-1-5 CVS branch (available via
AnonCVS, or ftp at ftp://ftp.netbsd.org/pub/NetBSD/NetBSD-release-1-5/)
and will be in NetBSD-1.5.3 (if/when that's released).

I suggest trying a kernel compiled from the latest sources in the
NetBSD-1-5 branch. If that works, update your userland as well because
tools like fsck_ffs and newfs have been updated (although the old
versions should work, I'd still update).

Luke.