K 10
svn:author
V 6
adrian
K 8
svn:date
V 27
2011-10-15T16:18:35.341230Z
K 7
svn:log
V 1018
Hide ath_rx_proc() behind the ATH_LOCK, and start using it before resetting
the hardware.

When resetting the NIC (eg when a stuck beacon occurs), there may be received
frames in the RX queue. This should happen:

* disable interrupts;
* stop pcu receive;
* handle frames in the RX queue;
* reset the hardware as needed;
* start the PCU RX (which purges the queue.)

This prevents RX frames from disappearing after an interface reset (eg
ANI config change, forced beacon stuck) and thus the RX AMPDU tracking
getting out of whack with the TX'ing side. Since PCU receive has stopped,

This doesn't entirely fix things, as the TX side can and will drop frames
on an interface reset or node flush. I'll worry about this in a subsequent
commit.

TODO:

* see whether the pcu rx should be stopped before disabling interrupts,
  that way the MAC definitely stops ACKing frames. Instead, right now
  with the above order, frames coming after intr disable but before pcu
  stop may overflow the RX fifo and cause some issues.

END