On 28 Jun 2001, Eric W. Biederman wrote:
> Steven James <[EMAIL PROTECTED]> writes:
>
> > Greetings,
> >
> > It sounds like we're on nearly the same path. I am using the ide driver from
> > TIARA also. I'm also using a special raw partition and header. On init, the MBR
> > from each IDE is fetched, looking for partition type FC. The first block in such
> >
> > a partition is a structure with a magic signature, and room for 3
> > kernel/initrd/commandline specs (any or all can be empty). The idea is to
> > provide a simple mechanism for failsafe kernels and the ability to boot from any
> > functional drive.
>
> Please tell me what is wrong with ELF header as the special header
> header. I didn't pick it because the standard tools use it. But
> instead because it is multiple platform, multi os, and darn trivial.
>
> I already support command lines, ramdisks, and initrd's with
> mkelfImage.
>
> I already support loading memtest86, etherboot, and linux directly.
> And to some extent I already support alpha.
>
> I also already have code that adds a self decompressor to ELF images,
> that works on both x86 and alpha.
>
> I can boot these images directly from the linux kernel of from linuxbios.
>
> The only part that I haven't completed is a set of standard tables to
> replace the ACPI, and irq routing tables. Maybe I'll get there maybe
> not but presenting a standard interface should be much less work in
> the long run.
That doesn't sound that bad to me. Since I just wanted to pass over
the sizes and kernel parameters, I also thought using a small
self-constructed header will do it. I don't know much about ELF,
but if this can be done it makes much sense to me.
After a short look into your mkelfimage, I think it is a great idea.
Armin
