Stefan Reinauer wrote:
Is there a flaw somewhere in my reasoning ?
Unfortunately, yes. The problem is that video BIOSes run in real mode and tend to rely on legacy system BIOS services. These _can_ be rewritten but noone on this list is very interested in doing so because it is a LOT of work that might never be worthwhile anyway. On the other hand, if it DOES get written it might also be used for booting other, legacy, operating systems like MS-DOS or maybe older Windows systems, both of these cases have been mentioned on the list, but I do not believe either has very high priority.
Couldn't LinuxBios simply call the appropriate init code in a graphics card Bios in order to get a text console, and then carry on afterwards in the same way a normal boot loader would ?
There's already code in linuxbios to do something like this. (as well as a userspace emulation program, which is kind of incomplete) The problem is that you have to emulate quite some legacy pc bios functionality to be able to do this. As that's all 16bit code (and LinuxBIOS is completely 32bit) this is not always trivial.
There's quite a bit of work going on that will support this over at:
http://www.missl.cs.umd.edu/Projects/sebos/winint/
Windowz and VGA support isn't very high priority for many on this list that mainly work with clusters but there is a great deal of interest in embedded and industrial computing for an open source BIOS that can support VGA and legacy operating systems. For many embedded projects the costs of securing source code and the royalty payments for a legacy spaghetti-code BIOSes are higher than the hardware development costs.
Bari