Hi,

I've recently had the pleasure of trying to install my own
snapshot of late July on an ASUS P2L97-DS system board equipped
with a 16GB IBM IDE disk and 256MB of memory.

It appears that no matter how I try I can't make the darn thing
boot from its IDE disk.  Here are the various things I have
tried, in no particular order:

 o Using the "default" geometry (2055c/255h/63s), newfs complains
   loudly that with that large a cylinder group, I would have to
   jack up the block size from 8K to 32K and the fragment size
   from 1K to 4K.  Even though the disk is large I'm not
   particularly keen on wasting that much disk space.

 o First, I tried to set the BIOS geometry to 16h/63s.  For some
   reason the "addressing type" would have to be "auto" in that
   case, and on boot-up it showed "LBA".  "Fine", I thought, "that
   should be giving me a nice half-MB per cylinder and no 32k/4k
   nonsense."  I put a disklabel on the drive with this geometry
   before starting sysinst (to "prime" sysinst, which appears
   necessary, otherwise it'll think it's an 8GB drive...).  However,
   it appears that at this time (possibly due to leftovers bits from
   previous attempts?) the geometry in the MBR partition table says
   255h, and "pfdisk" appears to be unable to set that value back to
   16.

During this process I have done "fdisk/mbr", tried putting the
NetBSD portion of the disk from sector 0, tried zeroing out the
initial part of the disk to start afresh, tried installing OS-BS as
the first-stage loader etc. etc.

The current state is that both the MBR partition table, the BIOS and
the NetBSD disklabel all say 16h, but *still* when it tries to
execute the boot code it simply clears the screen (jumps into
nowhere-land, I guess) and the PC restarts the BIOS boot sequence.

I think the primary source of this sort of trouble is the fact that
NetBSD still pretends to know something about the real geometry of a
drive, and tries to optimize its seek patterns accordingly.  While
some form of locality of reference on the drive is still beneficial
to performance, I don't think this sort of restriction makes much
sense anymore with both IDE and SCSI drives using logical block
addressing and where the drive doesn't even have a fixed number of
sectors per track.

I suspect that if I could just install with 255h and still use an
8k/1k file system, nothing of this mess would have happened.

I will also note that it's not entirely the first time I install
NetBSD either, and I think this sort of nonsense should be dealt
with better than it is now in order not to put off new (or old!)
users with reasonably modern hardware.

Hmm...  I guess my next try will be to install with 255h and with a
file system layout of 32k/4k to see whether that will work any
better.


Sorry 'bout the rant, good hints and ideas welcome.

- H=E5vard