Assertion of an OP on the fedora users list

[JD]

This is from a thread  "UEFI bootkit" on fedora-users mailing list

-------------------------
On 09/20/2012 05:09 AM, Alan Cox wrote:
> The required information for almost all X86 devices is not available.
> You can't build an open firmware for most x86 platforms from public
> information.
>
> Alan

-------------------------
Is the above assertion true?

[Tarl Neustaedter]

On 2012-Sep-20 13:01 , JD wrote:
> This is from a thread  "UEFI bootkit" on fedora-users mailing list
>
> On 09/20/2012 05:09 AM, Alan Cox wrote:
> > The required information for almost all X86 devices is not available.
> > You can't build an open firmware for most x86 platforms from public
> > information.
> >
> > Alan
>
> Is the above assertion true?

Replacing "X86 devices" with "X86 systems", it's pretty close. A lot of 
the initialization of various Intel and AMD CPU chipsets is not publicly 
available, and the BIOS vendors work very closely with the chip vendors 
during initial fabrication stages to get their code right. As I 
understand it from the people down the hall, It's not so much that the 
initialization is secret as much as it's difficult and poorly 
documented. And neither chip vendor is willing to take the effort to 
make the information readily available.

Most cases I've seen of people using openbios on real hardware have 
allowed the standard BIOS to initialize the chipsets and then take over 
before booting (in some cases, they boot openbios as a binary from a 
virtual floppy).

[JD]

On 09/20/2012 11:49 AM, Tarl Neustaedter wrote:
> On 2012-Sep-20 13:01 , JD wrote:
> > This is from a thread  "UEFI bootkit" on fedora-users mailing list
> >
> > On 09/20/2012 05:09 AM, Alan Cox wrote:
> > > The required information for almost all X86 devices is not available.
> > > You can't build an open firmware for most x86 platforms from public
> > > information.
> > >
> > > Alan
> >
> > Is the above assertion true?
>
> Replacing "X86 devices" with "X86 systems", it's pretty close. A lot 
> of the initialization of various Intel and AMD CPU chipsets is not 
> publicly available, and the BIOS vendors work very closely with the 
> chip vendors during initial fabrication stages to get their code 
> right. As I understand it from the people down the hall, It's not so 
> much that the initialization is secret as much as it's difficult and 
> poorly documented. And neither chip vendor is willing to take the 
> effort to make the information readily available.
>
> Most cases I've seen of people using openbios on real hardware have 
> allowed the standard BIOS to initialize the chipsets and then take 
> over before booting (in some cases, they boot openbios as a binary 
> from a virtual floppy).

So if the initialization code is not publicly available, but is
made available to Bios makers, could or would the devs
of openbios.org contact AMD and Intel, sign NDA's and implement
the initialization code in openbios?
Or is there more to it than that?

[Tarl Neustaedter]

On 2012-Sep-20 15:05 , JD wrote:
> >
>
> So if the initialization code is not publicly available, but is
> made available to Bios makers, could or would the devs
> of openbios.org contact AMD and Intel, sign NDA's and implement
> the initialization code in openbios?
> Or is there more to it than that?

As I understand, it's mostly a matter of effort. The initialization 
sequences are a mixture of stuff the chip vendors expected, the BIOS 
vendors understood, and a good chunk of stuff found out live once the 
first chips come off the testers. There's a lot of back-and-forth which 
never gets fully documented anywhere. There may be empiric limitations 
on what actually can be done during initialization which wasn't 
originally part of the specification, but the BIOS implementations end 
up using because it works.