Re: [PATCH] force load of lvm and raid before entering normal mode (Re: LVM on debian)

[Robert Millan <rmh@aybabtu.com>]

On Wed, Jan 09, 2008 at 02:07:20PM +0100, Jan Nieuwenhuizen wrote:
> Robert Millan writes:
> 
> > The problem with that is it can't really solve the problem, only provide a
> > half-baked solution.  If you include pc and gpt, you're misusing space in
> > a component where it's really critical, and you still don't support all
> > partitions GRUB might be installed on.  For that, you'd have to load _all_
> > partition maps, which means even more space.
> 
> I understand, and I've been wondering about this.  As I do not have much
> if any knowledge for the sizing constraints of a bootloader, so I cannot
> appreciate the efforts of reducing space.

The mission of core.img is being able to access /boot/grub, from where you can
load anything you want.  It has to be as small as possible, because if it won't
fit in partition boot record, this can be a significant inconvenient.

Ironicaly, the big problem is for LVM/RAID users themselves:

  if (must_embed && !able_to_embed)
    grub_util_error ("Can't embed the core image, but this is required when\n"
                     "the root device is on a RAID array or LVM volume.");

> As a (blonde) user,

Uhm how is the color of your hear related to this? :-)

> I would think that you'd want to use *as much*
> rather than as little space you can.  Why not (from a sizing pov,
> stability could be another issue), after core.img contains
> all necessary modules, fill up any `remaining' space (if that is how it
> works) with extra non-essential modules?  Little harm in the ability
> to read more devices, I would say?

No, the goal in core.img is not to provide useful features, just access
/boot/grub.  It has to be as small as possible.  The smaller, the more
chances it'll work.

-- 
Robert Millan

<GPLv2> I know my rights; I want my phone call!
<DRM> What use is a phone call, if you are unable to speak?
(as seen on /.)