Linux-2.6.7 Nitpick (Chapter 8)

Nathan Coulson conathan at conet.dyndns.org
Thu Jul 15 23:32:01 PDT 2004


> The following command is still in Chapter 8/kernel.html:
>
> sed -i 's@/sbin/hotplug@/bin/true@' kernel/kmod.c
>
> Perhaps this should be removed or changed?

Usually, I would advocate that we should only support what is in the LFS
book at that time, but I am going to disregard it here.

Earlier, when we first started playing with udev/hotplug in the LFS
unstable branch, we started running into problems where someone's reiserfs
system was mounted readwrite at boottime, and hotplug was installed.  It
would generate the hotplug events, which of course call udev, which
creates a file called /dev/.udev.tdb (I may have it wrong).

Anyway, later along in rcsysinit.d/S20udev [again, may be the wrong
filename], and if [ -f "/dev/.uded.tdb" ] exists, it'll exit without a
message.  You may also remember that the udev bootscript is the thing that
mounts a ramfs on /dev, and then runs udevstart.


To make a long story short, this person now has some extra files in /dev,
and their system wont mount a ramfs on /dev, and populate it full of
files.  (meaning your /dev directory is almost empty and probably wont be
a functional system until you remove /dev/.udev.tdb).


Now,  the LFS Bootscripts have taken it upon themselves to set what agent
to use for hotplug events when the time comes.  [S20udev, it'll setup
/sbin/udevsend, and if hotplug's installed later, it'll setup
/sbin/hotplug].

Therefore, taking out the above sed, people may have nonbootable systems
if hotplug is ever installed.



More information about the lfs-dev mailing list