ath 11n update: (better) handling of TX sessions after an interface reset

[Adrian Chadd]

Hi all,

I've begun work on fixing up the TX code so it doesn't drop (lots) of
frames when an interface is reset (eg stuck beacon.)
There's a few lock order issues which need resolving (surprise
surprise!) but by and large it seems to at least work in my own tests.

I'd appreciate some testing. What's left before I throw this into -HEAD:

* more testing! Lots, lots more testing! On all the atheros NICs that I have!
* BAR TX;
* reserving TX ath_buf's for TX'ing management frames (BAR, ADDBA, etc.);
* Fixing LORs;
* Fixing the TX buffer recycling order so TDMA works again;
* Figure out what the heck to do about locking in general, as right
now I'm doing some very evil things at the moment to try and get this
to work.

Thanks,


Adrian