On 01.12.21 11:28, Ivan Ivanov wrote:
Good thing about what I've suggested - "different places for v3 and v4 allocators" - is that it's (almost?) already done ;-)
It would mean a constant burden for the project. The resource allocation is kind of the heart of coreboot's ramstage, it's strongly tied to our devicetree model. Many changes in that area would have to be done twice, and changes to the older code are more expensive than changes to the newer, cleaner code (there were reasons to rewrite it!). Also, we'd be either constantly looking for testers for changes to the old code, or risk to break it which is why branching is so much cheaper for the community as a whole. Basically, what you are asking is that people who want to advance coreboot in that area should work twice as hard (mostly in their limited spare time).
Trust me, it's infeasible. You can't talk people into maintaining some- thing for you when you ask them to work on something very pesky for free.
Maybe worth to mention again: The platforms that don't work with the v4 allocator don't work because their code is broken. It's subtle bugs that were not visible with the v3 allocator by coincidence. You're also as- king to keep the code buggy.
Nico
PS. If you have any hope of open-source coreboot for newer platforms, you shouldn't make it harder for coreboot to advance.