[syslinux] Proposal for Chain-Loadable LDLINUX.SYS

[H. Peter Anvin]

On 01/01/2013 06:32 PM, Shao Miller wrote:
> On 12/31/2012 16:02, Gert Hulselmans wrote:
>> A LDLINUX.SYS that confirms to the (a) multiboot specification might
>> be a possibiltity to consider.
>> The multiboot entry point can be a different one than the normal entry
>> point (e.g doesn't need to load
>> the rest of LDLINUX.SYS from disk).
>
> This is true, but there's still the complication of the "missing"
> ldlinux.bin VBR code.  If this "missing" VBR code could be added to
> ldlinux.sys, then such a multiboot entry-point could rebuild ldlinux.bin
> in memory the same way as the current flow yields.
>
> I believe that the build process currently simply chops ldlinux.bin up
> to get ldlinux.sys.  It probably wouldn't be terrifically complicated to
> chop it up and then put it together again in a different order to result
> in an ldlinux.sys with the VBR code somewhere in it, but my original
> proposal ends up having ldlinux.bin == ldlinux.sys, so no more chopping
> (except in the installers).
>
> Time to review the (a) multiboot specification, anyway.  Thanks. :)
>

The installers is where the chopping happens anyway.  However, the 
chainloaded ldlinux.sys doesn't want to run the VBR; instead it should 
run a different piece of code in lieu of the VBR and the disk-loader 
code.  In Syslinux 5+ there really isn't a lot that piece of code would 
have to do.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.