On Wed, Jan 11, 2006 at 05:19:14PM +0100, Joerg Schilling wrote: > thomas riddle <Tom.Riddle at Sun.COM> wrote: > > > Suggestion of using inetboot was expected to provide issues as > > tracibility is difficult and UFSboot was thought to be an easier route. > > > > Since we have OF implementation on ODW and the capabilities of the > > Genesi team to enhance it, the choices came down to > > ------- > > Use a TFTP boot to load a simplified PPC flavored multiboot and the > > boot_archive. > > * this is the most straighforward approach and least code intensive, > > small changes to current multiboot to remove BIOS calls. Not the best > > longterm solution. > > Yesterday, I did talk with the person who did make the last changes > to ufsboot. > > ufsboot includes a semi complete ELF interpreter and knows that krtld > is the load-interpreter for /platform/*/kernel/unix (see elfdump > /platform/*/kernel/unix). > > ufsboot then loads krtld and krtld loads again /platform/*/kernel/unix > and then the files from the NEEDED tags. > > If (what I expect) multiboot on x86 does the same, it would be a relatively > simple task to hack multiboot to match the PPC needs in case the ELF code > if already OK for PPC.
I don't think the pegasos OF does (yet) support the NEEDED tag, but i can implement this, or we could use an external tool to deal with this. The other question i have about this, is where are those NEEDED files loaded ? Randomly, and then ELF relocated to some fixed address ? What about the kernel itself ? > the OBP then would only need to load/execute the modified multiboot program > and > to load the boot archive. Oh, you would still do ELF handling of NEEDED tags from the multiboot program ? /me still a bit lost about what exactly multiboot is, and where i can find the sources of it to become more knowledgeable. The source or some good documentation. Friendly, Sven Luther
