I wonder if part of the reasoning behind moving klibc into the Kernel itself and integrating it with kbuild has to do with creating a common base system for the initramfs that is meant to work across all Linux distros? I'm very skeptical that such a thing can be created in a way that would satisfy everyone, and maintained peacefully with no forks. I've been helping Jeff Bailey of Canonical with the 'initramfs-tools' package for Ubuntu. We've got a modular design that is extensible by way of installing packages and by the local admin. The initramfs is to be generated when the kernel image is installed via 'dpkg'. Scripts run at that time, and others are dropped into the image itself. It also has a facility for appending a set of .cpio.gz (overlay) files found in a certain directory. You may view the source via 'bazaar-ng' (bzr branch http...) at: http://people.ubuntu.com/~jbailey/bzrtree/initramfs-tools at: http://archive.ubuntu.com/ubuntu/pool/main/i/initramfs-tools/ ... or most simply, by logging into your favorite Ubuntu workstation and running: apt-src install initramfs-tools The package is not done yet. It will need to be gone over by the 'usplash' people, to ensure that the hooks they'll need are in place. (usplash is a user-space boot splash package) -- Karl Hegbloom <hegbloom@pdx.edu>
Hi Karl.> I wonder if part of the reasoning behind moving klibc into the Kernel > itself and integrating it with kbuild has to do with creating a common > base system for the initramfs that is meant to work across all Linux > distros? I'm very skeptical that such a thing can be created in a way > that would satisfy everyone, and maintained peacefully with no forks.It is my understanding that the main rationale here is to move functionality from kernel space to user space. So functionality usually provided in kernel space today but that really belongs to early userspace will be target for kernel inclusion. So consider what is included in the kernel as core functionality, what you add on top of this is what you use to diverge from the rest. Sam
Karl Hegbloom wrote:> I wonder if part of the reasoning behind moving klibc into the Kernel > itself and integrating it with kbuild has to do with creating a common > base system for the initramfs that is meant to work across all Linux > distros? I'm very skeptical that such a thing can be created in a way > that would satisfy everyone, and maintained peacefully with no forks.No, the reason is to be able to unload functionality from kernel code. This will establish, of course, a baseline set of initramfs code that will be always present, but will not replace current initrd/initramfs functionality. -hpa