[rft] please test -HEAD ath; lots of TX changes

[Lev Serebryakov]

Hello, Adrian.
You wrote 27 мая 2013 г., 11:29:29:

AC> Would you mind re-testing with what's now in -HEAD?
 Ok, now I can not "force" connection NOT TO PICK UP n-rates (without
disabling N on sever or client).
 Performance is better, than usual (150-180Mbit/s UDP), and SOMETIMES
here are such messages:

May 28 00:40:51 gateway kernel: ath0: ath_tx_tid_bar_suspend: tid=0, bar_wait=0, bar_tx=0, called
May 28 00:40:51 gateway kernel: ath0: ath_tx_tid_bar_tx_ready: c4:85:08:3f:9e:c2: TID=0, bar ready
May 28 00:40:51 gateway kernel: ath0: ath_tx_tid_bar_tx: c4:85:08:3f:9e:c2: TID=0, called
May 28 00:40:51 gateway kernel: ath0: ath_tx_tid_bar_tx: c4:85:08:3f:9e:c2: TID=0, new BAW left edge=186
May 28 00:40:51 gateway kernel: ath0: ath_bar_response: c4:85:08:3f:9e:c2: called; txa_tid=0, atid->tid=0, status=0, attempts=1
May 28 00:40:51 gateway kernel: ath0: ath_tx_tid_bar_unsuspend: c4:85:08:3f:9e:c2: TID=0, called
May 28 00:40:52 gateway kernel: ath0: ath_tx_tid_bar_suspend: tid=0, bar_wait=0, bar_tx=0, called
May 28 00:40:52 gateway kernel: ath0: ath_tx_tid_bar_tx_ready: c4:85:08:3f:9e:c2: TID=0, bar ready
May 28 00:40:52 gateway kernel: ath0: ath_tx_tid_bar_tx: c4:85:08:3f:9e:c2: TID=0, called
May 28 00:40:52 gateway kernel: ath0: ath_tx_tid_bar_tx: c4:85:08:3f:9e:c2: TID=0, new BAW left edge=3517
May 28 00:40:52 gateway kernel: ath0: ath_bar_response: c4:85:08:3f:9e:c2: called; txa_tid=0, atid->tid=0, status=0, attempts=1
May 28 00:40:52 gateway kernel: ath0: ath_tx_tid_bar_unsuspend: c4:85:08:3f:9e:c2: TID=0, called

 Which correlates with dropping of bandwidth to 90Mbit/s for one
second. But no re-association problems, no "20-30Mbit/s" problem. And
I've tried to power-cycle client and AP, it still pick up N rates from
first packet now!

  When I disable N on client, it shows stable 30Mbit/s (no 54m though
:)), but still no de/re-association problems for 600 seconds.

  But, I stress this out: there was NO way to get non-N rates when N
 was enabled on both ends, there was NO auth/association problems for
 several 600-seconds runs. And, yes, in my previous experiments,
 auth/association problems were only in situation when two N-enabled
 parties used non-N rates.

-- 
// Black Lion AKA Lev Serebryakov <lev at FreeBSD.org>