[LinuxBIOS] Will linuxBIOS Work on My Machine?

Jordan Crouse jordan.crouse at amd.com
Wed Dec 12 18:00:34 CET 2007

On 12/12/07 09:40 +0900, Jun Koi wrote:
> On 12/10/07, Uwe Hermann <uwe at hermann-uwe.de> wrote:
> > On Wed, Dec 05, 2007 at 12:31:54PM -0800, Wakefield, John  wrote:
> > > I have an Acer 5102WLMi Notebook with Dual Core AMD TL50s, which are 64
> >
> > I'm afraid LinuxBIOS doesn't yet work on laptops, those are much harder
> > to support than "normal" PCs for various reasons. It's on our TODO list to
> > support laptops too sooner or later, but it's not a trivial task.
> >
> Uwe, could you elaborate a bit here? Why supporting laptops is
> technically much harder than desktop??

There are a number of reasons why laptops are considerably more complex:

1) The embedded controller.  As Gregg alluded to, embedded
8051 microcontrollers manage a huge chunk of the laptop, from interfacing
to the LPC, inputs and outputs, and even power control.  No BIOS
implementation for a laptop can avoid interfacing with the embedded
controller.    The problem is that embedded controllers are notorious
black box devices.  They are most often specific to individual platform
designs, and deeply proprietary.  Without in depth knowledge of the specific
embedded controller at hand, getting it to work is a very long shot.  If you
want some idea of how tightly coupled the EC is to the BIOS, take a look
at the OLPC implementation - you will see the OFW and the kernel 
littered with embedded controller references, and bear in mind that during
that project, the OLPC team had unprecedented visibility into the EC code.
They _are_ that evil.

2) Laptops have different design constraints then desktops - this might
seem like a nobrainer, but it does affect us negatively.  With small form
factors, ROMs are usually soldered down (sockets take up too much room).
Legacy devices such as serial ports are omitted, as are JTAG connectors
and LPC headers and other things that we could use in a pinch to get access
to the machine.  Laptops will normally use different revisions of silicon
to save money, power or real estate, and usually what gets left on the floor
are the tools we can use to get at the system core.  Most of the devices
are soldered down, and there are no PCI or PCIe busses that we can tie into.
Unlike PCI cards, mobile GPUS and NICs won't have a companion EEPROM to 
store optionROM data; instead that will be carried on the ROM, which makes
it that much more difficult to replace the firmware image with LinuxBIOS.
And last but not least, it is rather difficult to take a laptop apart to get
at the motherboard, yet still keep everything working.

In short, the downside is huge, the upside is minimal, and few people are
willing to donate their laptops to potential bricking.


Jordan Crouse
Systems Software Development Engineer 
Advanced Micro Devices, Inc.

More information about the coreboot mailing list