Re: [coreboot] Coreboot Haswell Support

[Patrick Georgi]

Am 2013-09-17 09:36, schrieb Stojsavljevic, Zoran:
> I am not sure what is your scenario here. If you are attempting to
> bring HSW to grub, and then to Linux, Coreboot is not required, as far
> as you are using classical EFI BIOS or new UEFI BIOS (AMI, Phoenix,
> Insyde).

There's nothing said about GRUB or Linux in his mail - since he didn't 
go into detail on what he's planning, for all we know his project might 
be fundamentally incompatible with EFI, "Unified" or not.

Anyway, reasons why UEFI might not be suitable for users (no matter if 
GRUB or not):
- The IBVs's track record in product quality is rather mixed (Samsung's 
UEFI variable issue, for example. More examples: all the BIOS 
workarounds within Linux)
- IBV's products aren't open source
- UEFI's footprint is at least five times that of a coreboot solution in 
terms of lines of code (yes, I measured it)
- the previous two issues make any quality and/or security assessment of 
UEFI solutions harder than of a coreboot solution

I didn't think that these types of arguments need any more mention on 
the coreboot mailing list, but here we are...
> As my best understanding is, mrc.bin was released by Google in the
> best effort to support minimalistic approach (mrc.bin equivalent to
> x-loader) with Coreboot SNB (I guess, applies for IVB as well) support
> (Google did created mrc.bin with some internal INTEL group support).

Yes, and Google's "best effort" only had to happen because Intel isn't 
willing to support their customers' wishes.
> All these are now called Firmware Support Packages (FSPs), and they
> have locked IPs, sole property of INTEL. Please, read the following,
> widely available on INTEL external web site:

"locked IPS, sole property of INTEL" is probably not a very appropriate 
slogan when interacting with an open source community.
> http://www.intel.com/fsp
> http://www.intel.com/content/dam/www/public/us/en/documents/training/intel-f...

Since nothing but whitepapers is published on FSP yet, this can be 
safely considered vaporware.
> I am not sure about HSW FSP roadmap, I think HSW FSP is officially not
> released yet.

Indeed.

Maybe you can figure out a schedule for when we can expect an FSP binary 
that we can host in our 3rdparty repository under the same terms as the 
current MRC binaries?
That is, if FSP ever gets released at all (even Duke Nukem Forever did 
eventually happen, so let's hope)

Regards,
Patrick

Coreboot Haswell Support

[Stojsavljevic, Zoran]

Hello Patrick,

Do not want to reply to your (rather angry) arguments, I just can help a lot. I am one at INTEL who constantly and steadily fights for what you said below, but... INTEL business managers with numbers are far more important than I am. Allow me freedom not to go to details.

Let me try to help. Three points:
[1] You writes: Yes, and Google's "best effort" only had to happen because Intel isn't willing to support their customers' wishes.
    Here I am, rather very voluntarily, listening to customer wishes, and trying to push them through with INTEL structures.
[2] You writes: "locked IPS, sole property of INTEL" is probably not a very appropriate slogan when interacting with an open source community.
    Agree. I could not (!) explain to you what I did within INTEL Inside to change that. So far, I am facing the same Iron Wall as you do.
[3] You write: Maybe you can figure out a schedule for when we can expect an FSP binary that we can host in our 3rdparty repository under the same terms as the current MRC binaries?
    So far, I do not know exact date. I see that the schedule dates date supposed to be released, and it will be certainly HSW FSP + Coreboot (from INTEL public document).

I relentlesly expect that Sage upstreams IVB Coreboot changes to Coreboot Open Source, and these to become to be part of everyday Coreboot generic x86 framework.

Stay tuned.

Best Regards,
Zoran
_______
Most of The Time you should be "intel inside" to be capable to think "out of the box".

-----Original Message-----
From: [e-mail] [mailto:[e-mail]] On Behalf Of Patrick Georgi
Sent: Tuesday, September 17, 2013 1:43 PM
To: [e-mail]
Subject: Re: [coreboot] Coreboot Haswell Support

Am 2013-09-17 09:36, schrieb Stojsavljevic, Zoran:
> I am not sure what is your scenario here. If you are attempting to 
> bring HSW to grub, and then to Linux, Coreboot is not required, as far 
> as you are using classical EFI BIOS or new UEFI BIOS (AMI, Phoenix, 
> Insyde).

There's nothing said about GRUB or Linux in his mail - since he didn't go into detail on what he's planning, for all we know his project might be fundamentally incompatible with EFI, "Unified" or not.

Anyway, reasons why UEFI might not be suitable for users (no matter if GRUB or not):
- The IBVs's track record in product quality is rather mixed (Samsung's UEFI variable issue, for example. More examples: all the BIOS workarounds within Linux)
- IBV's products aren't open source
- UEFI's footprint is at least five times that of a coreboot solution in terms of lines of code (yes, I measured it)
- the previous two issues make any quality and/or security assessment of UEFI solutions harder than of a coreboot solution

I didn't think that these types of arguments need any more mention on the coreboot mailing list, but here we are...
> As my best understanding is, mrc.bin was released by Google in the 
> best effort to support minimalistic approach (mrc.bin equivalent to
> x-loader) with Coreboot SNB (I guess, applies for IVB as well) support 
> (Google did created mrc.bin with some internal INTEL group support).

Yes, and Google's "best effort" only had to happen because Intel isn't willing to support their customers' wishes.
> All these are now called Firmware Support Packages (FSPs), and they 
> have locked IPs, sole property of INTEL. Please, read the following, 
> widely available on INTEL external web site:

"locked IPS, sole property of INTEL" is probably not a very appropriate slogan when interacting with an open source community.
> http://www.intel.com/fsp
> http://www.intel.com/content/dam/www/public/us/en/documents/training/i
> ntel-firmware-support-package-technical-training.pdf

Since nothing but whitepapers is published on FSP yet, this can be safely considered vaporware.
> I am not sure about HSW FSP roadmap, I think HSW FSP is officially not 
> released yet.

Indeed.

Maybe you can figure out a schedule for when we can expect an FSP binary that we can host in our 3rdparty repository under the same terms as the current MRC binaries?
That is, if FSP ever gets released at all (even Duke Nukem Forever did eventually happen, so let's hope)

Regards,
Patrick