On Tue, 11 Feb 2003, RJ45 wrote:

[R: hello, on my ultra10 with NetBSD disks looks like this:
[R:
[R: 8 partitions:
[R: #        size    offset     fstype  [fsize bsize cpg/sgs]
[R:  a:   1049328         0     4.2BSD   1024  8192    86   # (Cyl.    0 -
[R: 1040)
[R:  b:   2097648   1049328       swap                      # (Cyl. 1041 -
[R: 3121)
[R:  c:  78165360         0     unused      0     0         # (Cyl.    0 -
[R: 77544)
[R:  d:  16778160  45090864     4.2BSD   1024  8192    86   # (Cyl. 44733 -
[R: 61377)
[R:  e:  16296336  61869024     4.2BSD   1024  8192    86   # (Cyl. 61378 -
[R: 77544)
[R:  g:  41943888   3146976     4.2BSD   1024  8192    86   # (Cyl. 3122 -
[R: 44732)
[R:
[R: disk is a western digital 40GB
[R:
[R: what eactly means cpg (cylinders per group?)
[R: I noticed this number is 16 on my DEC3000 runining NetBSD and is 16 on my
[R: Digital Personal Workstation running FreeBSD.
[R: does this number has any real importance or can I modify it from 86 to 16?
[R: Will it have inpact on performance?
[R: should I leave it 86 or change it ??
[R: Actually I can't figure out while it was set to 86...

I believe that newfs attempts to figure out the "optimal" cpg based on the
size of the partition one is creating relative to the size of the disk
on which it is being created.

Now, that aside, *I* have a difficulty understanding why I cannot
force newfs to use what is on the disklabel after I have made it instead
of watching newfs pretend it's smarter and more clever and has a better idea
of how I want my filesystem created than I do.  If the information is
present on the disklabel -- whether put there by newfs or whether I PUT IT
THERE when I ran disklabel -i or disklabel -e -- it should respect it.

				--*greywolf;
--
NetBSD: the free unix for the rest of us.