USB charging, part 2: implementation

Unbelievably complex and *almost entirely non-functional*, don't forget.

I do wonder how much of this is because the 'new' API is almost entirely undocumented. How can you expect anyone to use your API right if you don't document it or its proper use at all? "grep for driver uses" only works if all the drivers are exploiting every corner-case of the API, or at the very least if none are using it wrong. This requires every future driver implementor to produce perfect bug-free code: one mistake and people will copy the buggy driver and soon you have this mess. But of course 100% bug-free code is a reasonable minimum requirement, right? It's much easier than spending five minutes writing a few comments describing the interface in more detail, and means that Neil gets to spend countless hours trying to read your mind across time, which is no doubt much more fun for him than reading docs would be!

(Really, I *cannot* understand what goes through kernel developers' heads when they decide to design an API for general use, like this one, and then not document it at all. Surely it's not "I can't be bothered"? Maybe they're all non-native-English speakers and the language barrier is just too high?)