On Mon, Aug 11, 2003 at 11:09:07AM +1000, Luke Mewburn wrote:
> On Sun, Aug 10, 2003 at 10:58:47PM +0100, David Laight wrote:
>   | This has to be related to a smilar bug with raidframe.
>   | 
>   | The disk geometry is reported as 119150 cylinders, 16 heads, 63 sectors.
>   | I suspect this means that the ufs filesystem has a very large number
>   | of cylinder groups.
>   | 
>   | The directory placing algorithm doesn't work well under these circumstances!
>   | 
>   | Recent changes to newfs (with the ufs2 support) mean that it ignores
>   | the physical geometry so these problems won't happen any more!
> 
> newfs was changed a While ago to use the minimal number of cylinder
> groups that the geometry supports (rather than the default of `32').
> If I recall correctly, that change was made after this PR was
> submitted.
> 
> 
>   | Anyone know if the modified newfs is actually compatible with the old
>   | filessytem code?

In the mean time the machine has been updated to NetBSD-1.6, and the
problem went away. No newfs was necessary.

Nevertheless, thanks for looking into this PR.


--Lex