[syslinux] *.c32 for efi64 and efi32?

[Gene Cumm]

On Tue, Apr 22, 2014 at 8:15 PM, Steven Shiau <steven at nchc.org.tw> wrote:
>
> Dear Syslinux developers,
> I'd like to continue the discussion about this:
> http://www.syslinux.org/archives/2014-February/021659.html
> i.e. different directories for *.c32 files of BIOS, EFI32, and EFI64.
> I am wondering why we can not have
> *.c32 for the COMBOOT files of BIOS,
> *.e32 for the COMBOOT files of EFI32
> *.e64 for the COMBOOT files of EFI64
> As now the ldlinux file of syslinux 6.0x has, i.e.
> ldlinux.c32: BIOS
> ldlinux.e32: EFI IA32
> ldlinux.e64: EFI X64

I recall a discussion on this but having difficulty finding it.

> Because from the doc:
> http://www.syslinux.org/wiki/index.php/SYSLINUX#COM32_image
> it mentions:
> .c32            COM32 image (32-bit COMBOOT)
> Then it makes sense the *.c32 is for 32-bit. With EFI64, shall we have different name for that?
> With the different suffixes for 3 different architectures, we can put all the COMBOOT files in the same tftp/ftp/http dir. Then these 2 issues could be solved:
> 1. File name extension matches to its arch.
> 2. We could handle BIOS and EFI clients in one server, i.e.
>    make the goal of "One DHCP/PXE config for BIOS, EFI32, and EFI64 clients":
>    http://www.syslinux.org/archives/2013-September/020848.html

The resulting config would require suffix-less module references, i.e.
"UI menu" or "COM32 ls".

Additionally, I documented the basics of my test system here:

http://www.syslinux.org/archives/2014-February/021740.html

Bear in mind, by "URL-like file locations", I mean that if we have
"bios/pxelinux.0", "efi32/bootia32.efi" and "efi64/bootx64.efi" as
paths relative to the root and a common kernel "shared/vmlinuz", you'd
use "::shared/vmlinuz" or "tftp://<my-server>/shared/vmlinuz" in your
configs instead of "../shared/vmlinuz" (which is prone to issues).

Bear in mind, the only chance of a universal Linux kernel binary is
with an x86 (32-bit) kernel with BIOS and EFI hooks.  I can't be
certain if Syslinux EFI64 can boot such a kernel yet.

> Or maybe there are some key reasons I have missed so we can not have this?
> My 2 cents.

-- 
-Gene