Hammer Documentation is now Available at AMD's website

[Eric W. Biederman]

"Thomas J. Merritt" <tjm at codegen.com> writes:

> |<><><><><> Original message from Eric W. Biederman  <><><><><>
> |"Thomas J. Merritt" <tjm at codegen.com> writes:
> |
> |> |<><><><><> Original message from Eric W. Biederman  <><><><><>
> |> |So far I have just skimmed the Documentation but it appears that
> |> |AMD has now publicly published everything needed to do  Hammer
> |> |LinuxBIOS port.
> |> |
> |> |No great surprise but it is nice to actually see it !
> |> |
> |> |Yeah!  The chaos with the Dual Athlons has finally paid off :)
> |> 
> |> It does appear that they have released all of the documentation 
> |> necessary.  The learning curve is pretty steep though.  There are
> |> a lot of registers to program to make a box sing.  We have been
> |> able to bring SmartFirmware up on our AMD Solo box.  It has taken
> |> about 5 months to get to the point that we can now boot NetBSD.
> |> We're currently unable to boot Linux, but making progress on that front.
> |
> |Hmm.  I wonder if that is because of the open firmware chaos.
> 
> Open Firmware is pretty much out of the picture when the kernel takes 
> over.  The primary problem is the wired assumptions in the legacy PC
> chaos.

I guess that is what I meant the chaos of having Open Firmware or
anything else besides exactly what the kernel was expecting.
 
> |So far I have not seen anything out of the ordinary in the way of
> |a learning curve, and I just booted a linux kernel.
> 
> Keep in mind that our background is primarily PowerPC, Sparc, MIPS and ARM
> processors and we've avoided x86 for the most part.  Learning the
> Hammer side of things has been relatively easy, it has been more 
> painful learning the legacy PC stuff that hangs out in the 8111
> I/O Hub.

Not a problem.  The comment was meant more for than you to clarify
that the Hammer was exceptionally complex, for an x86 cpu.
 
> |> On the LinuxBIOS front we're planning to leverage our x86-64 experience
> |> and provide commercial support for LinuxBIOS on AMD64 processors.  
> |> With AMD's public release of the processor and chipset documentation we
> |> should be releived of most of our NDA obligations, and this will now
> |> become possible.  If anyone has hardware and is contemplating bring up
> |> LinuxBIOS please let me know, since I don't want to duplicate effort if
> |> we can avoid it.
> |
> |You have heard the talk about the freebios2 tree?
> |
> |Anyway Stefan Reinauer at SuSe and I are busily bringing it up.
> |So far so good.  We are still in the stage where you hard code things,
> |so you know you understand what registers need to be programmed,
> |but everything has been straight forward so far.
> 
> As straight forward as a PC can I would say.  Given that you have
> experience with prior AMD chipsets, the Hammer should look pretty 
> familiar.  We've reviewed about 5000 pages of documentation in the
> last several months and have most of it digested now.

Yep.  There can certainly be a lot to review and digest if you are new to x86.
The only good thing is that basic assumption stay the same for a long time.

Eric