[coreboot] BBC EFI story

[Corey Osgood]

On Sat, Oct 2, 2010 at 3:49 AM, ali hagigat <hagigatali at gmail.com> wrote:
> Dear Ron Minnich,
> I started with Wiki pages of Coreboot and i found Kontron, 986LCD-mITX
> as a supported mother board. I though its documentation is open
> because Coreboot is open source

Don't blame coreboot for your own misunderstanding. Just because a
project is open source doesn't mean every bit of data on how the code
was written has public documentation. The linux kernel (and I imagine
BSD, solaris, etc) is also open source, but some of the hardware
supported within doesn't have public datasheets, or the datasheets
that are public are incomplete.

> and ordered 4 motherboards, over 11000$.
> After investigating more about Intel manuals I found out that many
> registers of North bridge, 82945 have not been expressed!! and Intel
> gives this information to big BIOS companies ONLY.

This is incorrect. Intel will give this info to anyone who can present
them with a good reason for needing it. For whatever reason, they
don't want their competition knowing how to program the registers in
their chipsets. This is unfortunate, but a fact of life. You can
either deal with it and try to write a port without this
documentation, hire someone to write it for you, return your 4
motherboards in favor of AMD hardware, or live with the BIOS that was
presumably included with your boards.

> At least you could update Wiki pages to state this fact clearly to
> stop people waste money.

Most people don't insist on understanding every single line of chipset
code to port a motherboard. If your chipset is supported, it should be
relatively easy to port a board to it. Coresystems even went out of
their way to make sure the 945 port included the revision-specific
fixes for every revision of the 945, and also to support variants of
the 945 that most likely were not necessary for the board they were
working with. Instead of complaining about the lack of
chipset-specific documentation, why don't you dive in and try writing
a port based on the existing 945 boards?

> How people can contribute the code while they do not have
> documentation and necessary information?

The person who wrote the chipset port obviously couldn't, which means
that a company went to Intel with a contract/business case, acquired
the datasheets under NDA, wrote the code, and Intel then approved it
for public release. I understand that this can be frustrating, but
it's something that the coreboot project has no control over. If we
could make the datasheets for every chipset publicly available, we
would, but that's not how it works. NDA stands for Non-Disclosure
Agreement, which means that whoever gets that documentation does so
bound by law not to reveal the information in it, except what's
approved by e.g. Intel for release.

However, coreboot is written in such a way that it's theoretically
possible to write a port for a motherboard without ever looking at the
chipset code. As I look through the various boards that are already
using the 945 port, I don't see any of those nasty undocumented
registers being touched in mainboard code, that's all tucked away
inside the chipset code, and so you shouldn't need to touch it. So,
once again I urge you to stop asking/expecting the impossible, and
instead focus on using the resources available to reach your goal.

-Corey

>
> On Sat, Oct 2, 2010 at 12:22 AM, ron minnich <rminnich at gmail.com> wrote:
>> This story is full of errors. I thought I'd point out a few.
>>
>> "New PCs could start in just seconds, thanks to an update to one of
>> the oldest parts of desktop computers."
>>
>> I've got news for the UEFI forum: OLD computers, starting 10 years
>> ago, have been starting in seconds, thanks to the coreboot project. We
>> first showed a 12 second boot at the Atlanta Linux Symposium in Oct.
>> 2000. It is hardly news that one can boot a computer quickly. The
>> project then was called LinuxBIOS; the project today is called
>> Coreboot. Coreboot works on embedded systems, desktops, laptops, and
>> supercomputers. It has run in the iRobot packbot for 10 years,
>> searching mines and saving lives. Some of the biggest supercomputers
>> in the world have run coreboot. Work on coreboot was, in fact, funded
>> first by the US Gov't (Dept. of Energy) and more recently by the
>> German Government (see, for example,
>> http://www.coreboot.org/pipermail/coreboot-announce/2010-May/000007.html)
>>
>> So why, might you ask, did vendors not pick this technology up 10
>> years ago? Technology that worked on x86, 64-bit x86, Power PC, and
>> DEC Alpha? The reason is simple: it's open source. And, while vendors
>> finally did decide that an open source operating system might be
>> acceptable, they have had a lot of trouble accepting an open source
>> BIOS. They feel that too much information is divulged if the BIOS is
>> open source. They make a lot of excuses, but in the end, they finally
>> admit that the issue is that they don't want the hardware to be that
>> open.
>>
>> "The upgrade will spell the end for the 25-year-old PC start-up
>> software known as Bios that initialises a machine so its operating
>> system can get going."
>>
>> The BIOS could have been ended ten years ago, but for a simple fact:
>> many customers don't much like EFI. It's clumsy, slow, and closed. And
>> it's hard to work around, as it is designed to hide information.
>>
>> "The code was not intended to live nearly this long, and adapting it
>> to modern PCs is one reason they take as long as they do to warm up."
>>
>> There's a lot more to it than that. The closed nature of the BIOS
>> software made it very hard to replace. And, again, the vendors have
>> shown time and again that they prefer a closed, proprietary solution
>> to an open source solution. That's the real problem.
>>
>> "Alternatives to UEFI, such as Open Firmware and Coreboot, do exist
>> and are typically used on computers that do not run chips based on
>> Intel's x86 architecture."
>>
>> This statement is completely wrong. Coreboot has run on x86 systems
>> from the start and, in fact, only runs on x86 systems now. Open
>> Firmware also runs on x86 systems and is in fact the BIOS for the One
>> Laptop Per Child project -- an x86 system.
>>
>> ""At the moment it can be 25-30 seconds of boot time before you see
>> the first bit of OS sign-on," he said. "With UEFI we're getting it
>> under a handful of seconds.""
>>
>> It's nice to see UEFI catching up only 10 years later; the first
>> versions took 10 minutes to boot. Automobile computers, using ARM
>> processors, and an open source BIOS called U-boot, boot Linux in 8/10
>> of a second. So, while UEFI is where coreboot was ten years ago, we've
>> all moved on; seconds is kind of slow nowadays.
>>
>> "He said that 2011 would be the year that sales of UEFI machines start
>> to dominate."
>>
>> Dominate what? Certainly not cell phones. Certainly not the ipad. In
>> fact, UEFI is going to dominate a segment that matters less and less
>> nowadays -- PC-compatible desktops and laptops.
>>
>> I think you can do better than this article; the BBC is one of the
>> finest news organizations in the world. It seems you took a puff-piece
>> from the UEFI group and removed the quotes. I'm very disappointed in
>> the BBC.
>>
>> Thanks
>> Ron Minnich
>> Founder, LinuxBIOS, which is now coreboot.
>>
>> --
>> coreboot mailing list: coreboot at coreboot.org
>> http://www.coreboot.org/mailman/listinfo/coreboot
>>
>
> --
> coreboot mailing list: coreboot at coreboot.org
> http://www.coreboot.org/mailman/listinfo/coreboot
>