Subject: Re: Typo in the RaidFrame guide.
To: Marcin Jessa <lists@yazzy.org>
From: David Laight <david@l8s.co.uk>
List: netbsd-docs
Date: 05/06/2006 15:14:23
On Fri, May 05, 2006 at 11:43:01PM +0000, Marcin Jessa wrote:
> > 
> > FFSv1 should be able to handle filesystems with up to 2^32 fragments,
> > With 8k fragments I make that 32TB.
> 
> Isn't sectors size is a standard - 512 bytes, giving:
> 2^32*512 == 2TB
> ?

The sector size isn't relevent to the limiting size of an FFSv1 filesystem.
FFSv1 uses 32 bit numbers for file system fragments, these can be between
512 bytes and 64k bytes - although, since the upper limit for file system
blocks is also 64k, the usual limit for fragments is 8k.
So FFSv1 could support 2^32*64k = 256TB filsystems.

FFSv2 doubles the size of the on-disk inode (from 128 to 256 bytes), and
increases 'fragment numbers' to 64 bits (and a few other things that
aren't relevant for large filesystems).  This doubles the size of the
red-tape (see the recent thread about being unable to load large kernels
from FFSv2 using grub because you hit the 2nd level indirect blocks sooner).

	David

-- 
David Laight: david@l8s.co.uk