[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: An initrd proposal



Sorry for the late response, but I am not subscribed to the list and
only read the message because of weekly news.

> 2) get the installation binaries
>    - load necessary modules
>    - configure userspace programs if necessary
>    - access the medium
>    - get a single archive from the medium and unpack it to the ramdisk

> To do something useful in 2), we will need at least a shell, an init
> program and the c library (unless we use only static binaries which I doubt
> would save us something and is too unflexible anyway) We probably want a
> user interface at this point so an ncurses library is necessary.
> All this together with the modules to access the installation binaries
> needs to fit on the bootimage. Let me show how this is possible:
As a email of mine has been referenced in this mail, you might know that
I have a /working/ linurc solution (my system boots with ide support as
a module from an ide hdd).
I do not need a shell nor an init program. The only binaries I need on
the initrd image are 

-rwxr-xr-x   1 root     root       230243 Oct  6 16:20 linuxrc
-rwx------   1 root     root       338720 Oct  6 16:20 sbin/insmod 

A few device files in dev, the modules in lib and that's it. The
binaries are statically linked at the moment and therefore I think we
could save space with a minimal c library and dynamical linking.
What is not there (my initrd image is intended for non-interactive
booting when the system has already been installed and could be used by
Debian after the installation process - just to make it easier for
people) is a user interface for selecting the modules, but here we could
use the detect library for autodetection (maybe).

just a thought,
Rene


Reply to: