[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Initrd support & ext2fs bugfix
From: |
Jeroen Dekkers |
Subject: |
Re: Initrd support & ext2fs bugfix |
Date: |
Fri, 16 Jan 2004 22:35:06 +0100 |
User-agent: |
Mutt/1.5.4i |
On Fri, Jan 16, 2004 at 10:11:03PM +0100, Johan Rydberg wrote:
> Jeroen Dekkers <address@hidden> wrote:
>
> : * loader/i386/pc/linux.c (pupa_rescue_cmd_linux): Check whether
> : pupa_file_read fails.
> : (pupa_rescue_cmd_initrd): Implement.
>
> I've been thinking about PUPA a bit lately. The loaders go into sysdep
> directories as of now. This will result in a lot of duplicated code when
> PUPA gets ported to other platforms. Though, it minimizes the need for
> ifdefs for a perticular platform (which is a good thing, IMHO. )
>
> Most of the code in a ELF loader is generic. The only platform-dep code
> is validity checks and the invoktion of the kernel.
>
> As soon as PUPA gets ported to, for example, PowerPC, new constrains will
> be put on the loader. On the PPC platform, the bootloader must pass the
> entry point to the IEEE 1275:1994 firmware. So if the loader is generic,
> there must be a hook for this.
>
> Say that there is a PUPA module that implements OpenFirmware, then that
> needs to pass its entry point to the loaded kernel. So there is need
> for hooks for that aswell.
>
> So what do you say. Is it worth it to make (some) loaders generic?
Yes and no. First of all multiboot is x86-only so it was logical to
be put there. Almost all Linux stuff is also very x86-specific, I got
all my info from Documentation/i386/boot.txt and I didn't saw many
architecture-independent things.
I think the only thing which should be easily sharable is the
yet-to-be-created architecture indepedent Multiboot specification.
But I don't have any knowledge nor experience with non-x86 hardware
so I don't really know what's needed for that.
Jeroen Dekkers