Subject: Re: newfs can't make filesystems over 1TB in size
To: Brian Buhrow <firstname.lastname@example.org>
From: Luke Mewburn <email@example.com>
Date: 12/08/2002 11:22:15
On Sat, Dec 07, 2002 at 11:01:46AM -0800, Brian Buhrow wrote:
| Hello Luke. I think I also pointed out that the FreeBSD folks seem to
| define ffsize as a quad_t, a signed 64-bit value. How hard would it be to
| convert NetBSD to use this as the underlying access method? This seems
| like it would buy us some time in terms of what size filesystems we could
| support without having to change the underlying size of the sector.
A significant chunk of the existing file system code in the kernel
uses daddr_t (or derived types, such as ufs_daddr_t, which is defined
to be int32_t for on-disk purposes, in case daddr_t changes size)
to refer to underlying disk sectors.
I'm not sure of the exact effort involved in changing this; maybe
someone with more experience in the underlying file system structure
(for example, Bill Studenmund or Chuck Silvers...) can comment here.
| On Dec 8, 6:50am, Luke Mewburn wrote:
| } The kernel uses daddr_t to access underlying disk sectors.
| } daddr_t == int32_t == 2^31.
| } With 512 byte (2^9) sectors, 2^9 x 2^31 == 2^40 == 1TB.
| } If I recall correctly, we don't cleanly support larger sectors yet.
| } Luke.
| >-- End of excerpt from Luke Mewburn