[LinuxBIOS] GA-M57SLI - Nvidia X Server

Peter Stuge peter at stuge.se
Sun Jul 29 16:57:21 CEST 2007


On Sun, Jul 29, 2007 at 11:33:15AM +0200, Stefan Reinauer wrote:
> > I think it would be easy enough to change the initialization so that
> > ACPI is not required.
> 
> And it means the kernel driver has to maintain alll the tables for
> all CPU revisions and options.

Yep.


> > AMD: Are these tables board dependent? Why not put the tables,
> > arranged by cpuid, into the powernow driver and do without ACPI?
> 
> I am not AMD, but I guess it is preferred to do the efford once
> instead of once per OS.

I agree that there should be only one such source table. It should be
in a machine-readable format so that it can be converted to fit each
codebase that wants to use it.


> Fixing things that we do wrong in Linux is a bad thing.

But I think Linux is doing the wrong thing. (ACPI)


> It will only help us with Linux and lock out all others. Did we do
> this whole free bios thing to make it easier to lock others out?
> Ouch.

Agreed. This won't be "fixed" over night.


> > Another option is of course for someone to create the needed table,
> > store it in the right place and reserve that memory so Linux doesn't
> > use it - but I don't like that solution too much.
> 
> Why not?
> 
> I sense a large antipathy against ACPI, and I agreed to that before
> I wrote the initial ACPI implementation in LinuxBIOS.

I don't know ACPI well enough to criticise it technically but I'll
FUD a bit and say that others have done so already. :p

I see a big philosophical problem with ACPI though. I truly want the
boot software to be fire-and-forget and I want the operating system
to be able to actually drive the entire system. I want each OS to
have a power management subsystem with native drivers for all
neccessary hardware just like they do for all other hardware - only
then will the OS be able to make good decisions about how to control
the hardware.

It's just a variation on ditching BIOS interrupt services in favor of
native drivers.


> But now we have a framework since a couple of years and instead of
> improving that framework we constantly talk ACPI down. It really is
> a mixture between our device tree and our lbtable. Ok. The standard
> is broken in many parts, but it does not force people to use those
> broken parts.

True, but if not using the broken parts means reduced functionality,
we all lose. Users will be confused and upset and developers will be
annoyed and upset.


> One thing is pretty absolute sure though:
> 
> No ACPI -> No Laptops.

Compromise in the short term.


> Except we have a team on the Linux side fixing things up for us.
> Like the OLPC did. While that works fine for a philosophic and
> academic project like the OLPC, it is by no means an option to
> support a wide range of systems.

I hope and believe that open source in general and the OLPC project
in particular will educate not just kids but in fact the entire world
about how computers should work and what kind of quality any user
should demand.

I'm just worried it's going to take one generation too long..


//Peter




More information about the coreboot mailing list