On Mon, May 19, 2003 at 09:52:16AM +0200, Stefan Sonnenberg-Carstens wrote:
> Here is my original posting to netbsd-users,
> for which I did not receive any comments (question too stupid?!?!?) :

No too hard!

> Hi list,
> I got one of the 3ware escalade ide raid controllers with two
> 75 GB IBM DTLA drives attached and configured as RAID1.
> NetBSD 1.6.1 sees a ld0 drive which a configure with
> fdisk and disklabel.
> But when I made the entries for /,/usr etc and it comes to
> newfs sysinst stops and tells me I have to set bsize to 32768 and fsize =
> to 8192.
> If I do so, I can install and boot NetBSD but newfs tells me that the =
> performance of the system can be degraded.
> Is this a bug (twe driver), anything normal ?
> P.S.:
> It does not matter, if I enter the geometry by hand.

IIRC there are some lurking restrictions in the UFS disk layout that
cause problems with disks with very large numbers of cylinders.
(It is actually quite surprising that it allows for disks being 1000
times larger than anyone thought were possible when they wrote the code!)

Any solution would make the filesystem incompatible with earlier
code and other OS....

Large blocksizes can cause problems with the boot code (OTOH at the moment
you can't boot from a raid set - certainly not software raid) because
there just isn't enough space below 640kb for the number of filesystem
blocks that get buffered.

	David

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