Re: bioscall and EFI

[David Laight <david%l8s.co.uk@localhost>]

On Tue, Oct 08, 2019 at 09:20:34AM -0400, Thor Lancelot Simon wrote:
> On Mon, Oct 07, 2019 at 01:49:50PM +0000, Emmanuel Dreyfus wrote:
> > Hello
> > 
> > If I understand correctly, there is no hope to survive a bioscall()
> > when we booted from an EFI system. The first one happens in 
> > start -> init386() -> mca_busprobe()
> > 
> > How is it supposed to work? I would have expected an alternate path
> > for EFI bootstrap here...
> 
> This has me thinking that, actually, what happens early in boot does
> tell us quite a bit about the possible hardware configuration of the
> machine and could - should? - let us skip a bunch of probing later on.
> 
> Obviously you cannot have a machine with EFI firmware and MicroChannel.
> But I bet it goes a lot further than that (and I bet there are other BIOS
> calls that ought to be skipped too).

Even on a system with a traditional bios, it is probably best to
'execute' the bios code using an x86 cpu emulator.
This solves all the problems of getting back into real mode,
problems with the bios enabling interrupts, and lets the kernel
police the memory/io addresses the bios actually uses.

	David

-- 
David Laight: david%l8s.co.uk@localhost