Re: RFC: live-initramfs 2.x features
Tom Deblauwe wrote:
> - You should be able to boot a live system and it should be able to
> mount certain partitions that it reads from some config file in the
> binary filesystem.
that live-initramfs mounts certain partitions automatically means, that
/etc/fstab needs to have a configuration mechanism (already on the point).
that you can configure that options for fstab from within the binary
image is already listed under the following point:
One time argument handling (with proper respect of live.conf, also from
> - The ability to set the minimum size of the resulting binary image.
> When I build, I want the result to be 200MB, even if the total binary
> size is only 135MB. This is useful if you want to save some
> configuration information on the binary filesystem when using an USB
> stick to boot. This brings me to my next point...
ack, but that's a live-helper feature (build-time), not live-initramfs
> - The ability that the USB stick you boot from gets mounted under some
> directory so that it is accessible to write on.
this is already the case (/live/image), but it's ro by default. however,
you're right, we should provide a mechanism to allow to have it rw (if
fs and medium permits, of course).
> - When the host has e.g. 2.6.26 as kernel, and the image you are
> building contains a 2.6.30 kernel, the uname -a function executed within
> the chroot needs to be replaced if you want to compile kernel modules
> inside the chroot, otherwise you get 2.6.26 as a result. Maybe there are
> other commands that suffer from this?
this is a build-time problem, but it's a non-issue since we build from
debian packages, where the target kernel version can be specified at
module build time, see our example hooks for building various kernel
modules in an automated way.
Address: Daniel Baumann, Burgunderstrasse 3, CH-4562 Biberist