ADLO committed (NO FUNCTION TABLES!)
Ronald G. Minnich
rminnich at lanl.gov
Tue Nov 26 14:34:01 CET 2002
On 26 Nov 2002, Christer Weinigel wrote:
> "Ronald G. Minnich" <rminnich at lanl.gov> writes:
>
> > One thing we should keep in mind for the shadow ram thing:
> > Just about every chipset I have seen has shadow ram registers that can
> > correctly be set with the following info:
> >
> > VendorID, DeviceID, Function, register, AndMask, OrMask.
> >
> > This rapidly leads to a simple table something like this:
> >
> > struct shadowram{
> > int physaddress;
> > u16 VendorID, DeviceID, Function;
> > u8 register;
> > u8 AndMask, OrMask};
> >
> > struct shadowram s[] = {
> > {0xf0000, 0xabcd, 0x1234, 0x15, 0x40, 0xf8, 0x1},
> > { etc.}
> > {0,}
> > };
> >
> > I haven't stumbled across a chipset yet for which this would not
> > work.
>
> Except for the MediaGX/Geode which has this in the GX_BASE memory
> region which is not a pci configuration register. :-(
well, looks like my beautiful theory just got killed by a brutal gang of
facts. Now it's somebody else's turn to figure this out :-)
ron
More information about the coreboot
mailing list