On 21 Nov, Stefan Reinauer wrote:
Hi Daniel,
first, I think, we should discuss this on the list. I'd like to forward the postings if you haven't anything against this (without the binaries, they will go to the webserver...
Ok, openbios CC:ed
Ok. that's pretty good. Did you use the version including my try to include the menuconfig stuff?
No, I though about it but I decided that having one CONFIG_ variable per chip set is wrong, it should be one variable which have the chip set name as a value (have more than one chip set configured make no sense).
I don't see this. I've looked at the pci chipset stuff and it's really possible to access and probe the pci bus on newer boards without even havinng setup the memory controller.
Yep, but have you tried to program without working RAM? No stack, only 8 registers to keep stuff in? - Believe me the first thing you wanna do is get at least the first 1MB of RAM working...
I've talked with this about many people and having support for more than one intel chipset in one, for example, would be a knockout fact for award bios in large compuyter pools.
And there are MB specific issues such as how the SIMM/DIMM sockets are wired, how the AUX chip-select signals found on most super I/O and South Bridges are used, how the super I/O is implemented, which way to ROM is mapped after reset (do we look at the top 64k or the bottom in a 128k ROM).
So I'll tell you - a BIOS image is not only chip set specific, but it is mother board specific. It might be possible to make a BIOS which works for several configurations, but that will not be as simple as putting support for several SCSI HA's or NIC's in a Unix/Linux kernel.
/Daniel --