On Sat, 14 Aug 1999, Martin Husemann wrote:

> I'm replacing one of my NetBSD machines with a new one. As always, disk's 
> have grown big and cheap. This time I got this disk and controller:

> I don't need many partitions, so I created this disklabel:
> 
> # /dev/rwd0d:
> type: unknown
> disk: IBM-DTTA-371440 
> label: IBMDTTA371440 
> flags:
> bytes/sector: 512
> sectors/track: 63
> tracks/cylinder: 16
> sectors/cylinder: 1008
> cylinders: 16383
[snip]
> 5 partitions:
> #        size   offset     fstype   [fsize bsize   cpg]
>   a:   193536        0     4.2BSD     1024  8192    16   # (Cyl.    0 - 191)
>   b:   532224   193536       swap                        # (Cyl.  192 - 719)
>   c: 28229040        0     unused        0     0         # (Cyl.    0 - 28004)
>   d: 28229040        0     unused        0     0         # (Cyl.    0 - 28004)
>   e: 27503280   725760     4.2BSD     1024  8192    16   # (Cyl.  720 - 28004)
                                                      ^^   Try 32 or 64

> Right now "newfs" is doing it's job on /dev/rwd0e - and it's dumping block
> numbers for super block backups since several minutes.
> 
> Just to be sure - ffs won't do updates on all this blocks everytime the 
> filesystem is unmounted? Or even at sync time? I understand this block is
> important, but are  thousands of copies realy necessary?

It shoves one in each cylinder group. To make that number variable would
complicate the code even more. :-)

Just up the number of cyl/group. That will directly reduce the number of
superblock copies. At 32 c/g, you'll have about 511 cg's. At 64 c/g,
you'll have ~256. That's still a lot. :-)

Take care,

Bill