Am 13/01/2013 23:12, schrieb Fredrik Holmqvist:
2013/1/13 Axel Dörfler <axeld@xxxxxxxxxxxxxxxx>:On 01/13/2013 10:38 PM, François Revol wrote:Is this mostly correct? Should we use another entry in haiku_loader for EFI based systems?I don't see why you need yet another stage just for that.Indeed. It also cannot work this way without a BIOS emulation, unless you actually let haiku_loader have an EFI backend. With a standard EFI entry point; there is no need to load the haiku_loader from a BFS disk.Do you mean having a haiku_loader built with the efi api and ABI? That will make things more complicated. Or am I misunderstanding you?
You will need to access the disk drives to boot Haiku from the boot loader (unless you boot like CDs boot, or how network boot is done), so the proper solution is to go the way all the other platforms have gone, and use the API of the firmware to get Haiku to boot.
If we need to do that perhaps do enough in the efi application and just pass the info needed to haiku_loader? (I rather try to keep work at a minimum for the first version.)
I can't see why haiku_loader should not be the EFI application, or how your way can be any easier than this one. But maybe you can enlighten me :-) AFAICT the point of gnu-efi is to let you compile EFI apps from Linux without headaches as well.
Bye, Axel.
