On Wed, Feb 14, 2007 at 03:11:56PM +0300, Al Boldi wrote:
Your research is correct. The EPIA-SP boards use a lot of hardware different from the EPIA-MII and most of it is yet unsupported in LinuxBIOS.
It seems that this is a recurring problem with many boards.
Yes, it is.
Would it be possible to reduce dependencies, by having a minimal/blind CPU/memory bootstrapper with a kernel payload to discover the rest of the system?
This is indeed exactly what LinuxBIOS is.
LinuxBIOS is by definition quite hardware specific. It's task is to know about all the tweaks and quirks of a certain set of hardware, and tickle the hardware so that it behaves in a standard way.
We appreciate any help we can get in writing code for unsupported hardware. This is becoming an increasingly complex task with new technologies emerging every other month.
That said, I think that the V3 design and structure makes it easier to structure the code further and have more general code. There has already been some discussion about how to best generalize RAM initialization e.g. which would be one good way to make porting a little bit easiser and faster.
There's also the information problem, it's difficult to impossible for LinuxBIOS as a group to acquire the neccessary datasheets that detail how to set up the hardware, meaning many ports depend on individuals who do have access to the documentation but are allowed to release code.
And even AMD who support LinuxBIOS and contribute regularly in a stellar manner have needed a lot of time for their legal department to approve a code release.
It may sound like a crazy uphill battle, but I think the project is doing well climbing up the hill, slow as it may be. :)
//Peter