[coreboot] Fitlet-B

[Kyösti Mälkki]

On ti, 2015-06-16 at 13:07 +0200, Xan wrote:
> Hi,
> 
> I adquired a Compulab Fitlet B machine [http://www.fit-pc.com/web/products/fitlet/]. I contacted to Compulab in order to know how to install CoreBIOS to it. They sent me this message:
> 
> > Hi Xavier,
> > 
> > We’re in contact with Coreboot community in Finland that are testing and
> > checking feasibility of the fitlet with their code.
> > The moment they provide us with answers we will definitely publish them.
> > 
> > Thanks,
> > Maxim.
> 
> Can you confirm that? In which point is this porting?
> 
> Thanks,
> Xavi
> 

Yes that is true, I have acquired one fitlet here.

For some part, development is stuck due to AMD not supplying the
coreboot community with sources of crucial binaryPI to debug the chipset
initialisation. This is a motivational, technical and legal issue.

It took several months for AMD partners to deliver a very old revision
of a low-cost AMD Steppe Eagle db-ft3b-lc reference board, that had an
old coreboot port available for free download from Sage website. On my
approach, this would be the basis for Fitlet ports. I am waiting for a
permission to upstream my current work of db-ft3b-lc, as "those who pay
get to say on the schedule".

While that reference board has a connector compatible with Sage
SmartProbe, that tool will not be able to provide a coreboot log. I was
told AMD Platform Security Processor prevents any x86 debugging over
JTAG. There is no serial port, but there is a high-pitch LPC connector
to snoop on port 0x80. As for usbdebug, it is easy enough to hack it to
work, to some extent, when you know what you are doing.

So why was db-ft3b-lc of any relevance here? To verify we have a
toolchain to build a working firmware in the first place. Sadly it is
almost impossible to derive a new board or even do a rebuild of the
existing db-ft3b-lc firmware from this freely available "Open Source
Package". I suggest anyone interested requests a download and then
complains loud enough of the lack of some crucial files (like Makefile,
Kconfig etc.) in the package. I say it is only almost impossible,
because it can be done after you extract necessary files from the image
and make some assumptions which header files available elsewhere might
be the correct ones to use. You get the picture already?


So yes, there has been work done, but it has been more frustration than
what I would call efficient development so far. There is no contract or
funding for the fitlet port at the moment, so there is no schedule
either, at least not from my direction.

All in all, I think it is a nice compact hardware with a couple of
different but similar enough models to make ports for. I am actually
hoping this OEM might be the one to catch enough sales volume to have
something to say to AMD about this binaryPI thing we constantly struggle
with.


Kind Regards,
Kyösti Mälkki