better initrd for potato (was Re: debian 2.1r3 needed and release team coordination)
On Sun, Apr 11, 1999 at 04:35:58PM +0200, Sven Rudolph wrote:
> Enrique Zanardi <ezanardi@ull.es> writes:
>
> > The way our current kernel is built was the only way to go before
> > initrd support appeared, but initrd has been with us for a long
> > time,
>
> As I remember we already use the initrd mechanism. The syslinux config
> file says so. When I tried to get initrd working properly I found that
> the busybox init consumes the initrd ticket very early, so I couldn't
> change the root to the real root again. I have seen people using
> chroot for such purposes; I don't remember the drawbacks compared to
> initrd.
Yes, we use the initrd mechanism of the rescue floppy, but we make a very
limited use of it. The kernel still ships with everything compiled into,
we reboot to get into the installed system instead of changing to the
real root... And of course, no initrd is used once the system is installed.
> > That will give the user a smaller kernel, that loads faster and uses less
> > memory, and probably will fit his needs without having to recompile.
> > That also helps while building different rescue disks.
>
> Currently we run the policy that the kernel on the install system is
> the same as theone on the finally installed system. This means we have
> to provide tools to create the initrd image reliably; and this tool
> has to be used on about every kernel upgrade.
Indeed. That's what I would like us to develop for potato.
--
Enrique Zanardi ezanardi@ull.es
Reply to: