r7484 - in branches/udev_update/BOOK: . chapter01 chapter08

Archaic archaic at linuxfromscratch.org
Fri Mar 31 10:49:16 PST 2006

On Fri, Mar 31, 2006 at 09:51:06AM -0700, ken at linuxfromscratch.org wrote:
> Author: ken
> Date: 2006-03-31 09:51:06 -0700 (Fri, 31 Mar 2006)
> New Revision: 7484
> Modified:
>    branches/udev_update/BOOK/chapter01/changelog.xml
>    branches/udev_update/BOOK/chapter08/grub.xml
>    branches/udev_update/BOOK/general.ent
> Log:
> Populate /dev/ sufficiently for grub.

We should continue this discussion because I think this commit should be
reverted. The reason is one of a "big picture" view. We currently do
various mounting and fake mounting in more than 1 location. My proposal
(tested and working fine here on several builds) creates a one-stop
location for mounting.

1) Mounting, IMO, should *not* be done in chroot when it is avoidable
   because it doesn't mimic the boot process. The whole of chapter 5 is
   to create a viable environment for chapter 6. The proper mounting of
   filesystems (pseudo or otherwise) is part of that environment. The
   boot process sets up an environment, too, and it is done before a
   shell is started.

2) Moving all the mount commands to the same spot removes the bug about
   properly regenerating the environment if the system is rebooted
   before LFS is finished. We can simply put a note in one page that
   says "Do this again...".

3) The devices present to the host will be the same as for the new
   build WRT hard drives that grub needs, so no change.

4) The host's kernel still controls /dev/pts and /dev/shm regardless of
   when mounted or by which toolset, so no change.

Comments or questions?


Want control, education, and security from your operating system?
Hardened Linux From Scratch

More information about the lfs-book mailing list