[LinuxBIOS] EPIA-M Code

[Jonathan McDowell]

On Mon, Jun 20, 2005 at 08:38:30AM +0400, Anton Borisov wrote:

> 	I've played with EPIA-M tree recently. Interesting facts:

Which tree is this? From your output it looks a little like mine?

> 1) I'm always turning to "fallback" image. How LB decides which image
> to load? I guess this is done with CMOS entry (which one?), but I want
> boot VIA's AwardBIOS as well as LB (by switching Flash Chips in BIOS
> Savior).

> 2) I included file minicom.capped1.bz2. It's a minicom's output. I bet
> jmp_to_elf_entry() do the wrong staff (src/arch/i386/boot/boot.c) thus
> not loading payload at all.
> 3) Moreover. Looking at minicom.capped2.bz2. I've changed
> jmp_to_elf_entry() a little (put the code from PPC part). Ok, loading
> FILO now works. But reverting to "Normal" image. And finally, "SMBUS
> controller not found" because of wrong DEV (DEV = 0xFF).

What changes did you make for jmp_to_elf_entry?

How much memory is on your board? I think the RAM detection stuff is
wrong and that's possibly causing FILO problems; from your
minicom.capped2.bz2:

I would set ram size to 0x3fc000 Kbytes

and then later, in FILO:

collect_sys_info: RAM 4032 MB

> P.S. Booting in cold restart loading "Normal" image. "Normal" image
> don't boot - it stops at enable_smbus(). 

Adam Talbot and I have been looking at this; it doesn't seem to be a
specific piece of code that's crashing, as we've seen it hang at various
points around the initialisation code. My current guess is that some
interrupt that we're not handling yet is occurring, or that we need to
do some further chipset initialisation before trying to enable the
SMBUS.

I've got a list of chipset registers the Award BIOS seems to twiddle, so
I might add those to the enable_mainboard function to see if that at
least gets us further.

J.

-- 
        Thesaurus: ancient reptile with an excellent vocabulary