Carl-Daniel Hailfinger wrote:
The most ideal course of action would be to fit coreboot v3 and your ADLO effort together. The design of v3 is still flexible and maybe we can stuff a payload library in v3 which has functionality like generation of e820 tables etc. That would save payload authors from duplicating code needlessly.
I wonder whether we would want to create e820 tables directly in coreboot, instead of putting it into a payload library. It's something every OS needs, and requiring a payload to call a function like "make_e820_from_cbtable()" unconditionally makes little sense. As a comparison, we don't leave pirq, mptable and ACPI to the payload, either. Making e820 an exception here is one of the design inconsitencies we are suffering from. It also bites a couple of projects, like open firmware on coreboot. Torsten Duwe ran into funny issues here.
Stefan