On Sat, Oct 18, 2003 at 05:27:23AM -0400, der Mouse wrote:
> >> Can I inquire, while we're here, *why* a user-tunable cpg has vanished?
> > Probably because the value read from the disklabel had a nasty habit
> > of being completely wrong.
> 
> That's a reason to stop paying attention to the disklabel value, not a
> reason to break -c.

Ah - someone changed it to "max. blocks per group".
It is currently ignored (for compatibility).

-C <count> could be used for the minimum number of cylinder groups.

> > Why do you want to set cpg?
> 
> Personally, I like to set it to as large a value as it will let me, so
> that as little space as possible is wasted on per-cg guff.  (In my
> experience, yes, "wasted" is an appropriate term.

That is the default - except that it tries to put 4 cgs in a small
filesystems (and a hack to ensure there is only 1 cg in the memfs
filesystem created by init).

> I can't remember the last time I actually used any alternative[%]
> superblock aside from the one in cg 0.)

Yes, it would actually be nice if fsck checked that all the alternat{iv}e
superblocks were present and correct before destroying the world when
the system has found an old disklabel (and a pile of crap it assumed
was a disklabel).
(actually checking 'first few' and 'last' might be good enough, and would
allow the others to be used for files)

I did also wonder whether mfs needs any of the alternates at all!

	David

-- 
David Laight: david@l8s.co.uk