Re: [coreboot] grub as coreboot payload

[Robert Millan]

On Mon, Oct 26, 2009 at 09:20:53PM +0100, Andreas B. Mundt wrote:
> Hi,
>
> first, congratulations to the 1.97-release!
>
> I vainly tried to run the latest grub2 (Revision: 2663) as payload to
> coreboot (Revision: 4852) following the wikipage:
>  http://grub.enbug.org/CoreBoot
>
> Is this page still up to date and does anybody use grub2 as payload
> successfully? How?

Hi,

It appears that since r4534 (move to Kconfig), Multiboot information is no
longer built in by default.  You have to enable it in "System tables" /
"Generate Multiboot tables (for GRUB2)", then GRUB works fine (at least
on QEMU, which I just tested).

Coreboot developers: would you consider enabling it again?  The overhead
is minimal, and it would make this less confusing for users.

grub as coreboot payload

[ron minnich]

Like this?

ron

grub as coreboot payload

[Robert Millan]

On Wed, Oct 28, 2009 at 05:27:12PM -0700, ron minnich wrote:
> Like this?

Looks fine to me, thanks.

Acked-by: Robert Millan
> Leave it on by default, it's cheap.
>
> Signed-off-by: Ronald G. Minnich
> Index: src/Kconfig
>
> --- src/Kconfig	(revision 4887)
> +++ src/Kconfig	(working copy)
> @@ -319,7 +319,7 @@
>
> config MULTIBOOT
>    bool "Generate Multiboot tables (for GRUB2)"
> default n
> default y
>
> config GENERATE_ACPI_TABLES
>    depends on HAVE_ACPI_TABLES

grub as coreboot payload

[Uwe Hermann]

On Thu, Oct 29, 2009 at 01:42:11AM +0100, Robert Millan wrote:
> On Wed, Oct 28, 2009 at 05:27:12PM -0700, ron minnich wrote:
> > Like this?
>
> Looks fine to me, thanks.
>
> Acked-by: Robert Millan
> > Leave it on by default, it's cheap.
> >
> > Signed-off-by: Ronald G. Minnich

Thanks, acked and committed in r4929.

Uwe.

grub as coreboot payload

[Stefan Reinauer]

Robert Millan wrote:
> On Mon, Oct 26, 2009 at 09:20:53PM +0100, Andreas B. Mundt wrote:
> > Hi,
> >
> > first, congratulations to the 1.97-release!
> >
> > I vainly tried to run the latest grub2 (Revision: 2663) as payload to
> > coreboot (Revision: 4852) following the wikipage:
> >  http://grub.enbug.org/CoreBoot
> >
> > Is this page still up to date and does anybody use grub2 as payload
> > successfully? How?
>
> Hi,
>
> It appears that since r4534 (move to Kconfig), Multiboot information is no
> longer built in by default.  You have to enable it in "System tables" /
> "Generate Multiboot tables (for GRUB2)", then GRUB works fine (at least
> on QEMU, which I just tested).
>
> Coreboot developers: would you consider enabling it again?  The overhead
> is minimal, and it would make this less confusing for users.

Just checked, they're placed somewhere in the F segment.

That means they'll be overwritten when SeaBIOS is used as payload. Not
tragic, but is there a better place we can put them? or would this even
be unhealthy or useless?

grub as coreboot payload

[Robert Millan]

On Thu, Oct 29, 2009 at 09:58:30AM +0100, Stefan Reinauer wrote:
> Robert Millan wrote:
> > On Mon, Oct 26, 2009 at 09:20:53PM +0100, Andreas B. Mundt wrote:
> > > Hi,
> > >
> > > first, congratulations to the 1.97-release!
> > >
> > > I vainly tried to run the latest grub2 (Revision: 2663) as payload to
> > > coreboot (Revision: 4852) following the wikipage:
> > >  http://grub.enbug.org/CoreBoot
> > >
> > > Is this page still up to date and does anybody use grub2 as payload
> > > successfully? How?
> >
> > Hi,
> >
> > It appears that since r4534 (move to Kconfig), Multiboot information is no
> > longer built in by default.  You have to enable it in "System tables" /
> > "Generate Multiboot tables (for GRUB2)", then GRUB works fine (at least
> > on QEMU, which I just tested).
> >
> > Coreboot developers: would you consider enabling it again?  The overhead
> > is minimal, and it would make this less confusing for users.
>
> Just checked, they're placed somewhere in the F segment.
>
> That means they'll be overwritten when SeaBIOS is used as payload. Not
> tragic, but is there a better place we can put them? or would this even
> be unhealthy or useless?

It is useless.  It would become useful when SeaBIOS supports Multiboot.

But then, in order to get the main benefit of that (being loadable from
PC/BIOS + GRUB stack), it can't be linked in F segment anyway.

(Btw, I think a Multiboot SeaBIOS would be really interesting, specially
on non-BIOS x86 platforms like OFW or EFI)