Subject: Re: newfs can't make filesystems over 1TB in size
To: Brian Buhrow <firstname.lastname@example.org>
From: Bill Studenmund <email@example.com>
Date: 12/10/2002 18:45:06
On Tue, 10 Dec 2002, Brian Buhrow wrote:
> Just to be clear about the problem we're talking about. The problem is
> that NetBSD currently cannot address disk sectors above 2 TB of data.
> Worse, because the disk info is printed as signed integers when being shown
> to users, disks larger than 1TB show up with a negative number of
> cylinders, heads and sectors.
> Bill is proposing that we convert our disk labels
> and internal disk addressing schemes to use quad_t, which is a signed
> 64-bit value.
Well, I was first proposing adjusting the in-kernel structures to deal
with 64-bit block counts/offsets.
The disklabel would be a subsequent step, since we have to support
old-style disklabels, and it'd be a good time to revamp dislabel handling.
> That doesn't fix the filesystem size problem, but
> does allow us to populate multi-tb disks with 1-tb filesystems.
Actually, as I understand ffs, it will support 2^31 fs-blocks (fragments).
Thus ffs will get us farther than 1-tb. Not much, since I don't think we
want to go TOO much farther than 4k fragments.
> I'm in favor of this approach, but would now like guidence on how
> folks think it should be undertaken.