Re: Trouble partitioning disk

[Stephen Borrill <netbsd%precedence.co.uk@localhost>]

On Wed, 9 Jul 2008, Jonathan A. Kollasch wrote:
> On Tue, Jul 08, 2008 at 01:16:08PM -0700, James Colannino wrote:
> > Hey everyone.  I'm attempting to install the x86_64 version of NetBSD 4.0.
> > I have a 1.4TB disk, and the installer seems to recognize the geometry of
> > the disk ok.  However, when it gets to the point where I need to create
> > partitions, it says that the total amount of space on the disk is negative,
> > which tells me there's some integer value that's overflowing.
> >
> > I'm guessing that others have installed NetBSD on volumes larger than this,
> > so does anyone know why I'm having problems?  Thanks!
>
> Well, the on-disk MBR and NetBSD disklabel structure uses uint32_t
> to store partition location and size as a LBA.  2^32 sectors
> works out to 2TiB of storage.  So, the only issue is that signed
> and unsigned quantities are being mixed in sysinst.  Fixing
> that wouldn't solve the real issue though.
>
> Easiest thing would probably be to use a smaller (under 1TiB) disk for
> the system drive.  GPT could be used on non-boot drives.

I can confirm that I've successfully use 2TB drives (either RAIDframe or 
on hardware raid) at least since 3.1. However, I have manually 
disklabeled and newfsed them and I wasn't using them as the root 
partition. With the hardware raid (ld0 being 2TB), I stuck with a separate 
small drive for the root partition (it was some time ago, but this may 
well be because I ran into exactly the same problem as you).

For the RAIDframe method, I had either 4x750GB or 4x1TB disks, but created 
a small (2GB) RAID-1 root partition (raid0) and swap (raid1), with the 
remainder as RAID-5 (raid2). If raid2 < 2TB, it could be disklabeled and 
then raid2a newfsed as expected. If raid2 > 2TB, then (with a small 
tweak I made to the kernel array size printing) it is possible to newfs 
the raw device raid2d. This works with 3.1 and 4.0.

--
Stephen