urghh..
edit: been reading more and it seems there is more needed directly from amlogic. could anyone point me to something i can read that shows what exactly is needed, and how the drivers on that link aren't whats needed?
If you have to ask how ARM video licensing works (technically) there is no chance of you making a positive and insightful contribution to the "what's needed" debate. So I will summarise the problem and options (again):
Problem: The open-source drivers require a closed-source blob provided by ARM that is tethered to an unknown combination of hardware identifiers specific to its IP licensee (Amlogic). Options are:
a) Amlogic opens their chequebook and licenses the closed-source libmail.so library from ARM. This is not cheap. In Chinese culture it is impolite to say no, which is probably why the CTO of Amlogic smiled nicely, didn't say no, and also didn't provide any affirmative answers when I explicitly asked about plans to license S912 fbdev drivers in a face-to-face meeting earlier this year.
b) ARM suddenly develops a sense of philanthropy and gives us a free universal libmali.so driver to use; thus breaking the entire ARM licensing business model. I'm an optimist, but I also have a tiny suspicion it's not going to happen.
c) Reverse engineering of libmali.so allows us to defeat the licensing system used by ARM and hack a works-on-all library. It's not theoretically impossible and we already poked sticks at things. Despite being an optimist I suspect the $billion-dollar ARM licensing machine knows more about protecting its IP from reverse engineering than we know about reverse engineering.
d) The fully open-source "lima" driver manages to advance a decade in code maturity and provides a viable alternative driver. Considering the lima project is basically dead for the last two years; this is something I'm not optimistic about.