Re: Loadlin and Squeeze kernel 2.6.32

[Dom <toyer@rpdom.net>]

On 16/07/12 15:41, Camaleón wrote:
> On Sun, 15 Jul 2012 12:14:10 -0400, Tom H wrote:
>
> > On Sun, Jul 15, 2012 at 10:12 AM, Camaleón<noelamac@gmail.com>  wrote:
>
> > > Which, generally speaking, it translates into...? I mean, what are
> > > those "block lists" and how are they effectively affecting the boot
> > > process from a user's point of view?
> >
> > Let's assume that grub1/grub2 have to load "/boot/grub/camaleon" in
> > order to boot.
> >
> > If they're using block lists, they'll locate that file as starting on
> > xxx block of yyy partition.
> >
> > If they're using an intermediate stage (grub1's stage 1.5 or grub2's
> > core.img), they'll locate that file by name.
> >
> > Block lists are supposed to be less reliable/more fragile/(fill in with
> > the negative flavor that suits you).
>
> I'm not sure to had get it (sorry, I must be a bit dense...). Can you
> provide a user case for someone using block lists and another case when
> they're not in use?

As I understand it, when GRUB is installed in the MBR it installs a bit 
more code in the "spare" space between the MBR and the first partition. 
This includes code that recognises various file system formats (read 
only), so it can work out where "/boot/vmlinux-xxx" is.

When installed on a partition boot sector it doesn't have that spare 
space, so needs to have the location of the kernel/initrd hard-coded in 
to it as a list of physical disk blocks.

If, for any reason the files moved (say you resize the partition 
containing /boot, or backup/delete/restore /boot), then that block list 
won't match the actual location of the files *unless* you run update-grub.

In practice this won't happen very often, if at all.

--
Dom