[ath9k-devel] Auth Packet TX Delay

[Christian Lamparter <chunkeey@googlemail.com>]

(Added Marco and me in the CC - please keep it)

On Sunday, March 24, 2013 11:40:27 PM Robert Shade wrote:
> > Hm, so it's doing some fast channel changes?
> 
> Yes, the fastcc does seem to work, it's just that sometimes the chip
> can get in a bad state when it's not cold reset.
> > Just disable fast channel change entirely and re-test? And why not
> > just force a cold reset always? Why bother checking for the queue to
> > be stopped?
> 
> Just disabling fastcc was not enough, the cold resets are what seemed
> to have made the difference.  I was actually thinking about
> re-enabling fastcc and testing again.
> 
> The TXE/RXE checking path is from felix's "ath9k_hw: improve reset
> reliability after errors" patch.  I've just got the exception in there
> for 9160, since I don't have other hardware to test with.  What do the
> hardware engineers say about warm vs. cold reset?  I did notice that
> your latest ar9300 HAL has a note stating that "Warm reset is
> optimistic."

Marco reported in "carl9170: monitor mode hangs due to channel changes"
<http://www.spinics.net/lists/linux-wireless/msg105100.html> that
carl9170 would stop receiving frames after some time.
I looked a bit closer look and it seems that the fast channel change
for AR9170 also seems to be the culprit in this case.

(It seems that the AGC_CONTROL register is suddenly (re-)set by
something to "0x20" (default is 0x0004dd20) and then everything
stops.)

Regards,
	Christian