Ollie Lho <[EMAIL PROTECTED]> writes: > > I will used it to load some HW test program (like simplified memtest86 > before loading Linux Kernel. It also can potentially solve the GPL/NDA > conflict to some extend. Our customer may have some HW which they don't > want even us to know (they probably don't know it niether), if that part > of code is in an ELF image and loaded by LinuxBIOS then there will be > no GPL violation.
What you are asking for is very reasonable. The binary only code problem may be a little more interesting... Where I am at is that I really want a firm line saying the linuxbios core ends here. Doing that will keep out feature creep, so we don't end up evolving linuxBIOS into a messed up operating system. And I really want that line to be exactly loading one ELF image, from one place. Beyond that I don't see a problem putting a bootloader in the linuxBIOS tree. And sharing code with that bootloader. The bootloader can even have it's own set up CMOS parameters.... Specing out a return discipline for the bootloader also shouldn't be difficult. I just want the line so that someone who wants a really tight linuxBIOS, can count on it being there. If we try putting a bootloader in the tree and it is a horrible failure I'll drop the idea. But until we try I'll keep moving in that direction. Originally I really wanted to use a bootloader from another project like RedBoot, or etherboot, and to some extent that has worked. Using preexisting bootloaders really hasn't met the need. So I think we need something in the linuxBIOS tree that we have full control over. But at the same time I haven't given up on the seperate binary idea. Eric
