Subject: UFS2 status
To: None <>
From: Frank van der Linden <>
List: current-users
Date: 04/15/2003 11:26:12
Just to recap some things that I mentioned with regard to UFS2 and
FFS stability that I mentioned before, but were buried in various
threads on this list:

	* There were some fsck_ffs problems, most notably:

		* The oldest one: if you have an old fsck_ffs,
		  it will incorrectly signal a mismatch with
		  the first alternate, because it compares
		  unused fields. Nothing can be done about this
		  anymore, and old binary is an old binary.
		  1.6.1 has a forward compatibility fix for this.

		* It saw empty blocks in large directories, but
		  this wasn't true. Fixed on April 9th.

		* If it had to use an alternate superblock, it
		  failed to update the standard one. Fixed april 14th.

	* There were some kernel problems:

		* When switching back and forth between old and new
		  kernels (and using an old fsck perhaps), the
		  sum totals would get out of sync. Fixed April 5th.

		* Because of a buffer cache issue with finding
		  superblocks, the bitmaps in the first cylinder
		  group could get corrupted. This showed up
		  as 'map corrupted' panics and 'Bad file descriptor'
		  errors. Fixed April 12th.

There are a few open PRs, but they do not describe problems affecting
normal NetBSD usage. If you have run a kernel compiled from sources
between April 2nd and April 12th, please update. Using an up to date
fsck_ffs is also recommended. After updating both, run fsck -f
on all your filesystems just to make sure no corruption ocurred
(on a lot of systems you won't have problems, others may show
several problems, depending on your block/frag size).

After doing this, things should be fine.

- Frank

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