Re: P4 with i845E not booting with 2.4.19 / 3.5.31

[Jens Wiesecke <j_wiese@hrzpub.tu-darmstadt.de>]

Justin Heesemann wrote:
>>I tested:
>>2.4.19-pre6 -> boots with mem=512M parameter
>>2.4.19-pre7 -> didn't boot at all
>>2.4.19-pre7 with arch/i386/kernel/setup.c from 2.4.19-pre6 -> boots with
>>mem=512M parameter
> 
> 
> I have another update:
> some might know the memtest86 utility.
> 
> it has three ways of detecting the memory size "BIOS - Std", "BIOS - All" and 
> "Probe"
> 
> on my Athlon Mainboard (which has no Problems with 2.4.19 or later kernels) 
> all methods show the same (or nearly same) ammount of memory.
> 
> on my Epox 4G4A+ i845g Mainboard, which has all the Problems Jens Wiesecke 
> has, it's different:
> BIOS - Std : 640k (!!)
> BIOS - All : 4091M (!!!!)
> Probe: 511M (which seems correct as I have 512 MB Ram and 1 MB is shared 
> graphics ram)

ok. I tested my Chaintech 9EJL i845E mainboard the same way and I have 
exactly the same output with memtest86

BIOS - Std : 640k
BIOS - All : 4091M (memtest stops working afterwards)
Probe: 512M (since I have no shared graphics ram)

so the problem seems to be BIOS related. But what I don't understand is 
that since I tell the kernel to use 512 MByte of RAM (mem=512M) and 
kernels up to 2.4.19pre6 can handle this:

What changed and is there a workaround for kernels newer than 2.4.19pre6 
(for example telling the kernel not to rely on the memory information of 
the BIOS e820 procedure) ?

I tried to compile 2.4.20pre5-(ac) with the arch/i386/kernel/setup.c 
from 2.4.19pre6 but that didn't work.

Best regards
-- 
Jens Wiesecke
Institute for Macromolecular Chemistry
Darmstadt - Germany
e-mail: j_wiese@hrzpub.tu-darmstadt.de