Re: Hurd advocacy? Bootable (Knoppix-like) CD?
On Thu, Aug 21, 2003 at 02:15:24PM +0200, Niels M?ller wrote:
> Philip Charles <philipc@copyleft.co.nz> writes:
>
> > For a GNU live CD to work a hurdish ramdisk would have to be created and
> > setup at the initial boot.
>
> I think that's what the gzip store type is for; it's a store that is
> initialized by unpacking gzipped data in an underlying store. It
> should be possible to use it with the plain ext2 translator, although
> it might be inefficient compared to a specialized ramdisk fs. Hmm, we
> probably don't even need the gzip store, as GRUB can inflate the image
> for us.
>
> I'm a little confused about the Hurd boot procedure in general, but I
> guess it should work like this:
>
> GRUB loads the kernel, the ext2 root translator, any other servers
> (proc? ld.so?) which are needed for boot,
only ld.so, which then starts exec, which it can read from ext2fs. From
there, ext2fs starts init, which then starts the rest (auth, proc).
> and the gzipped ext2-image
> (which must contain enough free blocks so that new files can be
> written to it).
>
> The kernel wraps the loaded filesystem image into a memory object.
Which is currently not supported :)
> Then the ext2 server must be started with arguments that tells it to
> use a supplied memory image as the underlying store.
Theoretically possible.
> When the root filesystem is running, control is passed on to some boot
> script (/etc/runsystem?), which can install whatever additional
> symlinks and translators that aren't present already in the filesystem
> image, and start the rest of the system.
>
> I'm not sure what the missing pieces (if any?) are. The memory object
> store type?
That exists, I think. Probably untested.
> The passing of the memory object between kernel and the
> root ext2 server? The creation of a memory object from a grub module?
Right, those are missing. Probably not too hard.
> Another issue is that as we have no writable disks, the system must be
> reasonably stable without swap, or one has to fake some swap by by
> somwhow putting a swap file inside the ramdisk (ugly and wastes some
> more memory).
Yeah.
Well, I think that the above procedure is mostly useful for diskless
booting. For the CD Rom, a unionfs seems to be the better approach to me,
as it saves a lot of RAM.
Thanks,
Marcus
--
`Rhubarb is no Egyptian god.' GNU http://www.gnu.org marcus@gnu.org
Marcus Brinkmann The Hurd http://www.gnu.org/software/hurd/
Marcus.Brinkmann@ruhr-uni-bochum.de
http://www.marcus-brinkmann.de/
Reply to: