> I strongly recommend that, when using development kernels, you link
> init statically and put copies in all the directories checked, i.e.,
> /sbin/init /bin/init /usr/bin/init....
There's no need for that. You're pretty unlikely to break init upgrading
your kernel, and a statically linked copy is fairly unlikely to save you
if you do.
Far better is to always keep an older kernel around that you can boot in
an emergency.
> If you can get to the root mount point from another kernel, you can
> modify /etc/rc.d/rc.S to:
>
> /bin/bash
>
> ... on its first line.
>
> This will get you up to a very dumb mode with a shell. Just exit the
> shell to continue.
If you _do_ manage to break init and are unable to boot your system, it's
worth knowing that if you take init away altogether the kernel will just
run a shell for you.
phil