hybrid-graphics-linux team mailing list archive

["Ruslan N. Marchenko" <ruff@xxxxxxxxxxx>]

On Mon, 19 Jul 2010, Matthias Fulz wrote:
> Coutld it be perhaps, that the VGA ROM is directly mapped to the address
> by the System Bios?
>
> Perhaps its not even a real RAM Array, instead it's a RAM address which
> leads to the VGA ROM?
>
> Any way to read this address under windows?
> Perhaps the BIOS is remapping the ATI BIOS to the address under windows,
> I would check it, if anyone knows a way to read the addresses under windows.
Debugger? Will try at home, though consider the fact that vanilla catalyst 
from ati is not able to work with switchable graphics either, you need to 
have installed OEM version of driver (from acer in my case) which can 
afterwards be upgraded with vendor's driver.

> And here is (I think the problem we have with the RO MEM Array):
>
> http://www.pcguide.com/ref/ram/umaMap-c.html
> http://www.pcguide.com/ref/ram/umaShadowing-c.html
>
> So I think we need to find a way to bypass the BIOS Shadowing.
> Best would be to do it (if possible) inside the Linux Kernel.
At first i think we need to confirm whether we really need it. I.e. the 
problem we're trying to solve with this - to use catalyst drivers. Now, 
fglrx Xorg driver is not able to load bios, which can be caused by -
a) it is dumb and just tries to read bios via /dev/mem which is intel vga
b) it fetches bios correctly but unable to parse it because of 
modifications done by oem manufacturer
c) something else?

Bioses differs with only 3 bytes - is there any checksum for integrity 
verification? even one byte difference should change the whole checksum 
field, given it exists.

---
 Looking forward to reading yours.
   Ruslan N. Marchenko