Thinking about this some more, whats needed is a lot of grunt work that doesn't necessarily have to be done by the same person who fixes the ram init code.
Here's TODO from my past emails..
- I named both the northbridge and the southbridge i440bx. While I
don't think it s a huge problem. The southbridge code needs to be changed to piix4 since thats really what the part is. I'll hopefully be able to do that soon.
This has been done. It's now i82371eb.
- I seemed to have based the framework on one of the VIA chips. But
looking at things now I see a much better choice would have been the Intel e7501. Hopefully I can find some time to work on this a bit this week.
This still outstanding. If someone is feeling spunky they could go through the code re-do most of the stuff thats there.
The only thing functional in the current code is the boot stuff and the core SPD read functions, everything else is up for grabs. Its possible, (but I won't promise anything) that if someone were to get the framework all setup that I could come in and spend an hour or 2 comparing the ram init code to V1 and fixup the missing details.
One thing that can be done that doesn't require you to grok the ram init code but is still necessary is verifying the SPD support code functions are spiting out the right values.
The e7501 raminit code is _very_ well commented and pretty easy to follow even if you don't understand whats happening. A lot of the registers also seem very familiar to what I remember from the 440bx data sheet.
So the first step here is to start patching in the e7501 code and auditing it against the 440bx datasheet and V1 and then checking to see if the sane values are popping out of the various spd_support functions. Things like number of dram sticks installed, what the min cas timing is across all sticks, etc.
None of this is really hard just very time consuming. That would go a _long_ way in helping me make this chipset fly again.
Hi,
On Sun, Nov 05, 2006 at 10:07:58AM -0600, Richard Smith wrote:
- I named both the northbridge and the southbridge i440bx. While I
don't think it s a huge problem. The southbridge code needs to be changed to piix4 since thats really what the part is. I'll hopefully be able to do that soon.
This has been done. It's now i82371eb.
Why not 82371eb (that's the official name AFAIK)? Is there some requirement that directories have to start with a letter?
The e7501 raminit code is _very_ well commented and pretty easy to follow even if you don't understand whats happening. A lot of the registers also seem very familiar to what I remember from the 440bx data sheet.
I think I'll have a look at this in the next few days. But I might use the 855pm code (derived from e7501) as a basis, that looks even better.
Uwe.
Uwe Hermann wrote:
This has been done. It's now i82371eb.
Why not 82371eb (that's the official name AFAIK)? Is there some requirement that directories have to start with a letter?
Its a C thing. You can't start a variable with a number. That allows you to match the directory name to the struct statements in the code.
I think I'll have a look at this in the next few days. But I might use the 855pm code (derived from e7501) as a basis, that looks even better.
the e7501 datasheet has a date on it thats almost a full year earlier than the 855pm. I think the register names and operation may be closer to the 440bx. But feel free to use whatever. Its mostly about framework.
This has been done. It's now i82371eb.
Why not 82371eb (that's the official name AFAIK)? Is there some requirement that directories have to start with a letter?
Its a C thing. You can't start a variable with a number. That allows you to match the directory name to the struct statements in the code.
Why not call it piix4e (or just piix4), much easier to type and proofread and understand in general?
Segher
* Segher Boessenkool segher@kernel.crashing.org [061105 21:57]:
This has been done. It's now i82371eb.
Why not call it piix4e (or just piix4), much easier to type and proofread and understand in general?
because we called all the others after their part numbers rather than their code names, too.
Why not call it piix4e (or just piix4), much easier to type and proofread and understand in general?
because we called all the others after their part numbers rather than their code names, too.
But the PIIX4 family has 12-or-so part numbers and only very minor differences between those parts, they could all be handled by one and the same driver?
Segher
-
Richard Smith -
Segher Boessenkool -
Stefan Reinauer -
Uwe Hermann