On Sun, Feb 24, 2008 at 08:30:50PM +0100, Peter Stuge wrote:
On Sun, Feb 24, 2008 at 01:34:29PM -0500, Kevin O'Connor wrote:
I do wonder... if coreboot builds pirq, mptable, ACPI, and e820 tables. Why not have it also populate that other bios "table" - the 64K one at 0xf0000-0xfffff?
If by "table" you mean code the answer is that coreboot does not like callbacks by design.
Well, it wouldn't be coreboot that got the callbacks, it would be that binary blob at 0xf0000.
Seriously though, I don't think implementing the real-mode irq handlers is that technically challenging - I'd guess it is no harder than implementing filo was. The calling interface is stable - implement it once and there's not likely to be a lot of maintenance.
So, I'd look at it as coreboot initializing and loading another "table" at a certain area of memory. Some OSs will make use of it - others wont. I don't see a lot of down sides to deploying it.
The big upside is that, with the proper handlers, all modern OSes will load out of the box. This will be a big selling point to hardware vendors - especially those that want to deploy a single solution for both mass-market and niche markets.
-Kevin