On Thu, Nov 09, 2000 at 07:52:08AM +0100, Pållen wrote:
> On Wed, 8 Nov 2000, Ignatios Souvatzis wrote:
> 
> > On Wed, Nov 08, 2000 at 10:04:08AM -0800, Bill Studenmund wrote:
> > > NetBSD had no problem. We made 32k/4k file systems on them and went to
> > > town.
> > 
> > > fsck was SLOW.
> > what cpg?
> 
> Where can we get info on how to set such parameters?

Uh, the newfs(8) manual page?

Personally, I think the current defaults in newfs are poor, to say the
least.  An effort should be made to minimize cpg, which is almost always
excessive for modern disks.

It's a shame that the filesystem data structures would have to be modified
to allow reasonable cpg values with the default blocksize; many of my
machines run 32k/4k or 16k/2k filesystems for exactly this reason.

Charles mentioned something a long time back about how the "can't fragment
beyond 1/8 the blocksize" restriction could easily be killed, and that would
seem to offer a reasonable solution, too, but I can't recall what his idea
was.  That way we could maximize cpg by using large blocks, but not have to
waste so much space by having correspondingly large fragments.