[LinuxBIOS] epia

[Tyler Pohl]

http://linuxbios.org/Supported_Motherboards

I see the epia is "Broken".  No wonder I could never get it to work
properly.  Is anyone ever going to get this to work?  What's broken anyhow?
I know the processor is missing the CMOV instruction.

Thanks
Tyler Pohl

[Corey Osgood]

Tyler Pohl wrote:
> http://linuxbios.org/Supported_Motherboards
>
> I see the epia is "Broken".  No wonder I could never get it to work
> properly.  Is anyone ever going to get this to work?  What's broken
> anyhow?  I know the processor is missing the CMOV instruction.
>
> Thanks
> Tyler Pohl

Take a peak at the northbridge (vt1211?) code, ram initialization isn't
working correctly (not sure if it's in getting the spd data or setting
the registers correctly). The last person to work on it patched it up to
work on his specific configuration, but it needs work to work on a more
generic basis (ie different ram configurations).

-Corey

[Ben Hewson]

Corey Osgood wrote:
> Tyler Pohl wrote:
> > http://linuxbios.org/Supported_Motherboards
> >
> > I see the epia is "Broken".  No wonder I could never get it to work
> > properly.  Is anyone ever going to get this to work?  What's broken
> > anyhow?  I know the processor is missing the CMOV instruction.
> >
> > Thanks
> > Tyler Pohl
>
> Take a peak at the northbridge (vt1211?) code, ram initialization isn't
> working correctly (not sure if it's in getting the spd data or setting
> the registers correctly). The last person to work on it patched it up to
> work on his specific configuration, but it needs work to work on a more
> generic basis (ie different ram configurations).
>
> -Corey

The main problem with the EPIA board is that ROMCC misses the next line
after any #if or #else directive with a C++ style comment when
compiling. It may also miss the line after a #endif as well, I didn't
check that. There are quite a few in raminit.c that set various
registers. Needless to say missing lines of code is not good.

If you either take out the C++ comments or change them to standard C
comments the raminit does mostly work.

There are some problems which I havent got to the bottom of yet.

Firstly the first spd read nearly always fails. It doesnt matter so much
as it is in a function that tries to automatically set the memory timing
registers, but it doesnt actually set anything, the code is commented out.

Secondly sometimes the system will just hang while doing SPD stuff. Now
there are timeouts in all functions that could hang, so I am not sure
what is going on there. Possibly something with either the northbridge
or southbridge is not setup quite right.

Other than that I have my board booting linux using filo.

[Uwe Hermann]

On Tue, Nov 28, 2006 at 05:17:53PM +0000, Ben Hewson wrote:
> The main problem with the EPIA board is that ROMCC misses the next line
> after any #if or #else directive with a C++ style comment when
> compiling. It may also miss the line after a #endif as well, I didn't
> check that. There are quite a few in raminit.c that set various
> registers. Needless to say missing lines of code is not good.
>
> If you either take out the C++ comments or change them to standard C
> comments the raminit does mostly work.

Can you send a patch for this?

Thanks, Uwe.

[ron minnich]

If I am feeling brave I may modify romcc to disallow C++ comments. I
realize I should fix this 'the right way' but i don't feel like
becoming a romcc expert.

ron

[Alex L. Mauer]

Ben Hewson wrote:
> If you either take out the C++ comments or change them to standard C
> comments the raminit does mostly work.

For what it's worth, this did not work for me.  I also had to add the
extra line after any #if or #else even if there was no c++-style comment
to get it to work at all (I also experience the apparent SPD lockups).

-Alex Mauer "hawke"

[Stefan Reinauer]

* Ben Hewson  [061128 18:17]:
> The main problem with the EPIA board is that ROMCC  misses the next line after
> any #if or #else directive with a C++ style comment when compiling. It may also
> miss the line after a #endif as well, I didn't check that. There are quite a
> few in raminit.c that set various registers. Needless to say missing lines of
> code is not good.
>
> If you either take out the C++ comments or change them to standard C comments
> the raminit does mostly work.
>
> There are some problems which I havent got to the bottom of yet.

Can you verify how far you get with the latest version of romcc?