Update Multiboot spec

Update Multiboot spec

[James Buchanan]

Is anyone working on an updated Multiboot specification?

If no, I think we should start one and make a page for it on the GRUB-2 
devel wiki.  It should have extra options to scan for the MP structure 
on IA-32 machines, generalize the video options as much as possible 
(give framebuffer address and device and manufacturer IDs, bus and 
function number, something like that -- would it work for PowerPC and 
other machines?), options for detecting all hardware (e.g. linked 
structures saying I found so-and-so PCI cards, so-and-so type of 
graphics/sound/ethernet cards, monitor type and ID, this system has 
ACPI, CPUID for IA-32, things like that.)

I'm just thinking out loud here.  Does anyone want to help write a new 
Multiboot spec?

Thanks,
James

Re: Update Multiboot spec

[Yoshinori K. Okuji]

On Friday 08 July 2005 06:31, James Buchanan wrote:
> Is anyone working on an updated Multiboot specification?

I don't think so.

> If no, I think we should start one and make a page for it on the GRUB-2
> devel wiki.  It should have extra options to scan for the MP structure
> on IA-32 machines, generalize the video options as much as possible
> (give framebuffer address and device and manufacturer IDs, bus and
> function number, something like that -- would it work for PowerPC and
> other machines?), options for detecting all hardware (e.g. linked
> structures saying I found so-and-so PCI cards, so-and-so type of
> graphics/sound/ethernet cards, monitor type and ID, this system has
> ACPI, CPUID for IA-32, things like that.)

In principle, we do not want to use the Multiboot Specification as helper 
functions. It is a boot protocol, but not an OS library. For example, CPUID 
is not necessarily useful to probe in a boot loader, since operating systems 
can probe it after the control is passed. There is no reason why a boot 
loader must handle it.

The most important goal of the next version of Multiboot Specification is 
portability. There are some ideas about this, but not decided yet. I hope you 
could help this.

Okuji

Re: Update Multiboot spec

[James Buchanan]

Yoshinori K. Okuji wrote:
> In principle, we do not want to use the Multiboot Specification as helper 
> functions. It is a boot protocol, but not an OS library. For example, CPUID 
> is not necessarily useful to probe in a boot loader, since operating systems 
> can probe it after the control is passed. There is no reason why a boot 
> loader must handle it.
> 
> The most important goal of the next version of Multiboot Specification is 
> portability. There are some ideas about this, but not decided yet. I hope you 
> could help this.

You're right -- I got carried away there.  I looked at GRUB's own 
support code and forgot to separate the two.

I'd be glad to help if I can.  I'm working on some GRUB-2 code at the 
moment, studying it and trying to figure out some video card stuff.

James

Re: Update Multiboot spec

[Yoshinori K. Okuji]

On Friday 08 July 2005 08:57, James Buchanan wrote:
> I'd be glad to help if I can.  I'm working on some GRUB-2 code at the
> moment, studying it and trying to figure out some video card stuff.

I think we need to list up what information must be passed by a boot loader on 
every architecture and each architecture first. I made two new pages on the 
wiki:

http://www.autistici.org/grub/moin.cgi/MultibootRequirements
http://www.autistici.org/grub/moin.cgi/MultibootIdeas

Feel free to edit these pages.

I think some survey is required. It would be worth checking how existing 
systems deal with each architecture (such as Linux, NetBSD and L4KA).

Okuji

Re: Update Multiboot spec

[Cyril Plisko]

Yoshinori K. Okuji wrote:
> On Friday 08 July 2005 08:57, James Buchanan wrote:
> 
>>I'd be glad to help if I can.  I'm working on some GRUB-2 code at the
>>moment, studying it and trying to figure out some video card stuff.
> 
> 
> I think we need to list up what information must be passed by a boot loader on 
> every architecture and each architecture first. I made two new pages on the 
> wiki:
> 
> http://www.autistici.org/grub/moin.cgi/MultibootRequirements
> http://www.autistici.org/grub/moin.cgi/MultibootIdeas
> 
> Feel free to edit these pages.
> 
> I think some survey is required. It would be worth checking how existing 
> systems deal with each architecture (such as Linux, NetBSD and L4KA).
> 

I am not sure how applicable it is here, but when I read about tags
I immediately thought of IEEE1212 directory format. Anyone think
that can be useful here ?

Cyril

Re: Update Multiboot spec

[Hollis Blanchard]

On Jul 9, 2005, at 12:19 AM, Cyril Plisko wrote:

> Yoshinori K. Okuji wrote:
>> On Friday 08 July 2005 08:57, James Buchanan wrote:
>>> I'd be glad to help if I can.  I'm working on some GRUB-2 code at the
>>> moment, studying it and trying to figure out some video card stuff.
>> I think we need to list up what information must be passed by a boot 
>> loader on every architecture and each architecture first. I made two 
>> new pages on the wiki:
>> http://www.autistici.org/grub/moin.cgi/MultibootRequirements
>> http://www.autistici.org/grub/moin.cgi/MultibootIdeas
>> Feel free to edit these pages.
>> I think some survey is required. It would be worth checking how 
>> existing systems deal with each architecture (such as Linux, NetBSD 
>> and L4KA).
>
> I am not sure how applicable it is here, but when I read about tags
> I immediately thought of IEEE1212 directory format. Anyone think
> that can be useful here ?

I haven't heard of IEEE 1212, and of course I can't read it without 
paying.

The tagged-structure approach is a very good idea here.

-Hollis