Dual Athlon Progress. (fwd)
Eugene.Leitl at lrz.uni-muenchen.de
Wed Jun 13 02:14:12 PDT 2001
-- Eugen* Leitl
ICBMTO : N48 10'07'' E011 33'53'' http://www.lrz.de/~ui22204
57F9CFD3: ED90 0433 EB74 E4A9 537F CFF5 86E7 629B 57F9 CFD3
---------- Forwarded message ----------
Date: 13 Jun 2001 02:58:17 -0600
From: Eric W. Biederman <ebiederman at lnxi.com>
To: LinuxBIOS <linuxbios at lanl.gov>
Subject: Dual Athlon Progress.
I'm making head way.
I have solved an annoying bug where the system would somtimes
crass at boot. The fix was to initialize some more registers.
Totally unitialized regiterst that if you reboot fast enough
from since your last boot are fun.
I have tracked down my IOAPIC problems and it appears the proper
IOAPIC initialization is to explicitly disable all of it's interrupts.
I'd verify that by booting into linux but something I did has broken
DMA... And since one of the drivers I have compiled in does DMA
during it's initialization it fails... Has anyone seen linuxBIOS to
break DMA before?
Anyway I'm going through everything one step at a time making certain
not to leave any details left undone. Hopefully I'll get to the DMA
issue soon. There is so much left to do in the ram initialization
that it may take a while.
I'm making progress in finish the SPD code. Ram sizing has been done
for a while, but now I need to get all of the timing registers
programmed right. The Athlon is annoying you can't initialize the
cache and use it for ram. There has to be something better than
assembly to write this mess in.
Anyway. I have have registered versus non registered DIMMS handled.
The first pass at automatic CAS latency detection, written. And front
side bus clock speed changes written. There are still probably a
dozen more timing registers to go, and a lot of error cases to handle
yet. With some luck I should get this done tommorrow. Currently
raminit.inc is at 1000 lines and growing...
DDR SDRAM is just like SDRAM for as far as the SPD bytes are
concerned. But with stronger requirements that you use the SPD bytes.
I'm currently using 17 SPD bytes, and it looks like I'll probably wind
up using at least 6 more. I don't think I even want to try fitting
this code into a doc and only 512 bytes...
My totally unreasonable goal is to have the code in good enough shape
to start committing it into the linuxBIOS tree by friday.
The current outstanding bugs are:
- The second CPU doesn't seem to want to start up.
I've traced this pretty thoroughly but I'm not certain what is wrong
yet. I might have to implement cpu reset code in linuxBIOS.
- DMA doesn't work.
Now to go to bed so I am worth something tommorrow.
More information about the Beowulf