ath(4) hostap problems (still)

Hi,

when running ath(4) on -CURRENT/-RELENG_6 as client
and hostap on FreeBSD-5 on another box, you will notice
that the box locks up (it's stuck in a busy loop
inside the kernel somewhere, if I'm lucky I can get enough
cpu-time to press the reboot-kombo). I've described this
problem already in my earlier mails.

Today I tried to run the hostap box on FreeBSD-6. The
problem still exists, but the client seems to recover
from the lock after a few seconds and I can get some
information out of it. The connection cannot be established
once after the short lock has happened and any try to
ifconfig down/up, will lock up the client for another few
seconds.

I could start athstats after the client recovered from the
lock-situation:

3 beacon miss interrupts
22 mib overflow interrupts
62 tx management frames
180 tx frames discarded prior to association
6 tx failed 'cuz too many retries
69 long on-chip tx retries
55 tx frames with no ack marked
96 tx frames with short preamble
3 periodic calibrations
avg recv rssi: 30
1 switched default/rx antenna
Antenna profile:
[1] tx      101 rx        0
[2] tx        0 rx       70

The value of "tx frames discarded prior to association" several
times a minute. The antenna profiles change sometimes, but
everything else does not change, if you run athstats multiple
times.

I hope it helps to investigate the problem.

Client: Thinkpad R40 with Netgear WG511T PCMCIA.
HostAP: Sempron 2200+, Asrock K7VT4A+, Netgear WG311T PCI.

Martin