ollie <[EMAIL PROTECTED]> writes: > ollie wrote: > > > Jeremy Jackson wrote: > > > >> Ok put another way: 0xf000 - 0xffff from who's > >> point of view, CPU, PCI bus, or ISA bus? > >> > >> Ok... I see ipl.S for sis630 does a far call right at the > >> reset_vector: so the i386 switches from the dummy > >> segment descriptor it gets at reset 0xfffff000 to > >> 0xfe00. > >> > >> Sorry kinda thinking outloud here. I think the PIIX4 makes 0xf000-0xffff > >> ram, which the > >> std BIOS shadows itsself into before jumping > >> there, but I'm just guessing. What does the 630 do? > >> > > > > This depends on your southbridge and superio chipset. For SiS 630 it does jump > > > to F segment as its first instruction.The chipset will decode F segment > address > > > to the Flash socket.
Is there any reason we need to make a far call? My gut reaction is that it is easier to make a near call and use the high alias. For the normal LinuxBIOS case this works quite well. > BTW, after everything is done. The nvram_on in southbridge.c make the F segment > mapped > to RAM (for DoC cases). This is because LinuxKernel look F segment for PIRQ > table. Nice. Ollie if you are regularly doing this that makes at least one thing easier. If Ron keeps playing with pcbios compatibility I'll have to move the linuxbios table to segment F. Eric
