Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 24 May 95 12:07:26 MDT
From:      terry@cs.weber.edu (Terry Lambert)
To:        pb@fasterix.freenix.fr (Pierre Beyssac)
Cc:        freebsd-current@freefall.cdrom.com
Subject:   Re: newfs weirdness...
Message-ID:  <9505241807.AA08459@cs.weber.edu>
In-Reply-To: <199505240059.CAA12436@fasterix.frmug.fr.net> from "Pierre Beyssac" at May 24, 95 02:59:51 am

next in thread | previous in thread | raw e-mail | index | archive | help
> newfs doesn't seem to use the number of sectors by cylinder it gets
> from the disklabel. It correctly warns that it will use 4096 and that
> it differs from the disklabel value, though.
> 
> Is that the expected behaviour ?

Yes; modern SCSI drives use ZBR (Zone Boundry Recording), which means
that the geometry is translated.  Since it's in hardware (it can be
recovered using SCSI II commands only), it's largely irrelevent to
most operations; however, the head seek and rotational latencies are
then wrong in UFS/FFS everywhere except near the fron of the disk, and
even then they could be shifted.  The result is that UFS/FFS optimizations
for linearly laid out disks (ESDI/MFM/RLL) are actually causing slower
performance by expecting seeks when there are non, etc.
the 4096/1 geometry is a way of "shutting off" most of the UFS/FFS
optimizations.

If you are using a non-ZBR disk, you can overrride the 4096/1 with a
parameter to newfs.  Note: I'm not sure if it uses the BIOS fictional
geometry (which would have the same deleterious effects as a ZBR geometry),
or if it uses the one reported by the controller when you do the override.
This should probably be investigated.


					Terry Lambert
					terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9505241807.AA08459>