Hello all

Following the discussion with Robert Love, here is a patch that :
	- provides AGP support for i820 chipset
	- fixes several incorrect reads/writes to the APSIZE register. According 
to the Intel specs, this register has a size of 8 bits for the 820, 840, 
845, 850 and 860 chipsets. As a consequence, the types for the aperture 
info has been changed to aper_Size_info_8
	- the '*tlbflush' was incorrect. The Intel specs give only one bit of the 
AGPCTRL to be set to flush the GTLB. The generic function writes this 
bit, along with two others (why ?)


The patch has been made against 2.4.13-ac7. I have another one available 
for the 2.4.9 Redhat's kernel at 
http://ltswww.epfl.ch/~aspert/patches/patch-agp_i820-2.4.9-6_fix_apsize
This one works fine with me (I have i820) _but_ the other chipsets have 
not been tested. Please report any bug/silly things you may find. I have 
not tested the patch for 2.4.13-ac7, but it should not break up too many 
things :-)

Nicolas.

PS: I am not subsribed to linux-kernel, so CC to me your 
answers/proposals/whatever ;-)
-- 
Nicolas Aspert      Signal Processing Laboratory (LTS)
Swiss Federal Institute of Technology (EPFL)