Re: Progress report

[Jos H. <jo...@st...>]

On Mon, 3 Mar 2003, Nicolas Souchu wrote:
> I also tried Matrox load:
>
> G200_AGP0: <KGI G200 AGP driver> mem 0xdc000000-0xdc7fffff,0xdb000000-0xdb003fff,0xdf000000-0xdfffffff irq 9 at device 0.0 on pci1
> <2> scanning pcicfg space:
> <2> scanning device 102b 521, subsystem 102b ff03
>
> <2> scanning device 1002 4758, subsystem 0 0
>
> <1> subsystem autodetect failed.
> <2> scanning pcicfg space:
> <2> scanning device 102b 521
>
> <1> found device 102b0521 at 1000000
> Matrox G200 revision 1
> <1> board is a primary card - already initialized by the BIOS
> <2> mem_check_region('Matrox VGA text16 aperture', base 0xb8000, size 32768): ok
> <2> mem_check_region('Matrox BIOS', base 0x0, size 65536): failed
> Matrox BIOS region already served! (base_io: 0)

Yeah, you run into an error that is no error, but a very old issue. KGI
provides no mechanism to take over the primary video board, just claims
the region, maps its own consoles and hopes no other driver accesses the
card anymore. This works on linux for the linux doesn't know memory
resource management (i.e. on linux the mem_check_region functions always
passes), and the boot vga driver relies entirely on the BIOS/VESA in real
mode, i.e. it doesn't have to claim the VGA IO region nor the BIOS region

In short: the fact it doesn't work on BSD, is due to a flaw in Linux :) We
never had the need to fix this.

Jos