On Mon, May 19, 2003 at 10:04:58AM +0100, David Laight wrote:
> 
> 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.
> 
Is this in any way related to the rather peculiar results I get
from installing NetBSD-1.6.1 on a fasttrak100 (TM) host?  It's a
(GigaByte?) GS-SR101 rackmount server using the 6VTXDR motherboard.

I had it working by turning off the RAID capability and wiring the
second drive directly to the master controller, under 1.6.  I was
hoping 1.6.1 would actually manage it, but whereas I can install,
it seems that the kernel and the RAID controller (or is it the
install kernel and the generic kernel?) have different views of
the configuration and once installation is complete rebooting almost
succeeds, except the mounting process seems driven by an unrecognisable
/etc/fstab (the root directory remains read-only and a warning/error
message complains about some mount point inconsistency).  I have
a suspicion that drive selection is going wrong, so I'm now mirroring
what I hope is the drive I originally installed to (note, it ought
to have installed to both and it certainly didn't) to the other
one, to see if somehow it makes a difference.

I would dearly like to sort this out, I'm scared I might lose the
small cabling I have to remove to make only half the system work.
If I need to do some experimenting, I'll happily do it while I have
access to the hardware (today and the next two days).

++L