Please test - collapsing the TX locking down to ATH_TX_LOCK()

[Adrian Chadd]

Hi,

I've committed the patch, with an extra bugfix (recursive locking, eek.)

I'm now going to go through and begin un-doing the couple of LORs I
know will occur when ieee80211_free_node() is called whilst the TX
lock is held.

Please update to -HEAD and test thoroughly with WITNESS enabled.

I'm going to make a big push now to simplify a lot of the locking in
the TX path so I can squeeze some more TX performance out of the code.

Thanks,



Adrian