On Tuesday 7. February 2017 17.17.14 Luke Kenneth Casson Leighton wrote:
On Tue, Feb 7, 2017 at 4:01 PM, Jonathan Frederickson
Sure, but is that a result of the hardware actually being that different or is it just because manufacturers aren't actually upstreaming their code? In the latter case, everyone would have to keep rewriting driver code, but not because it doesn't already exist - just because they can't cooperate.
it's a byproduct of the N (designs) times M (processors) problem. Fabless Semi A flatly refuses to mainline their kernel because it costs them too much time and money to do so, passes it OEM B who just compiles it maybe makes a few (GPL-violating modifications) and chucks product out the door.
Interestingly, although Ingenic (mentioned earlier) have always seemed to target older kernels - you can't really blame them if it's what their engineers know, they can get stuff out the door easily, and the Linux speedboat is not exactly their fault, anyway - at least they released all the code, and the various jz47xx SoCs appear to be supported pretty well in the mainline, thanks to various people forward-porting those code drops.
(I recently promised to test mainline kernels on the Ben NanoNote, and I mustn't forget to get back to that.)
So although there's limited libre operating system distribution support for those SoCs, mostly because of a lack of commitment from those distributions, fuelled by a perceived lack of available hardware to build them natively, the kernel support is probably pretty good. Of course, you can't guarantee that stuff won't be thrown out of the speedboat, but the situation has been surprisingly good for a couple of years at least, I'd say.
Paul