On Mon, 25 Dec 2000, Frank van der Linden wrote:
> On Fri, Dec 22, 2000 at 11:35:13PM -0600, Richard Rauch wrote:
> > I have a hard drive that is 20GB (give or take a little).  It is generally
> > recognized as such by NetBSD.  However, note the following:

 [...disklabel/fdisk report about 7 or 8 GB of physical space...]

> There are 2 limitations with PCs and IDE drives. The first one is the
> BIOS interface limitation when using CHS (cylinder/head/sector) addressing.
> The limit there is 1024/255/63. The other limit is the old controller
> interface limit, which is 16383/16/63.

Ah...this lays to rest my concerns then.

However, when I was upgrading to NetBSD 1.5, the installer remarked that
it was _quite_ sure that it had positively identified my hard drive and
knew what the geometry should be.  I don't remember which set of numbers
it gave me (1024/255/63, or the other).  There was something about
``unless these numbers are obviously wrong'', I should accept them.  
Well, they _were_ obviously wrong, and I suspected that it was BIOS
limitations creeping in---but I wasn't sure.  Since fdisk was reporting
these numbers prior to upgrading (and really this just deals with marking
the disk for BIOS's sake), I figured that if it ran safely under
1.5_ALPHA, it would continue to run safely under 1.5.

Conceivably, a new NetBSD user wouldn't have any reason to believe that
this was ``okay'', though.

Perhaps the message (in sysinst?) would benefit from being rewritten?  
Something along the lines of adding, ``...unless your drive is larger than
about 8GB, in which case these numbers are necessarily bogus to the point
of not even match your disk's real size.''


> Due to these limits, reported CHS values will be smaller than the total
> size of the disk if it's > 8G. Don't worry about this. Ignore the

Thanks.  It's good to have confirmation of what I suspected.  (^&


  "I probably don't know what I'm talking about." --rauch@eecs.ukans.edu