LOR - COM LOCK

[Adrian Chadd]

This is the 900mhz NICs, right?

On 8 October 2011 05:02, Edgar Martinez <emartinez at kbcnetworks.com> wrote:
> Jan  1 00:00:36 wlanmon: (RTM_IEEE80211_ASSOC) [wlan3] [00:80:81:8d:9d:83]
> ath1: stuck beacon; resetting (bmiss count 4)

.. I wonder why that is happening on those NICs. This is in normal
hostap+sta WDS setup, right? Rather than adhoc or mesh?
(That changes how the MAC handles beacon TX.)

> Jan  1 00:01:32 wlanmon: (RTM_IEEE80211_JOIN) [wlan1] [00:c0:c1:cb:6c:c2]

> Expensive timeout(9) function: 0xc05aa150(0xc58c3b80) 0.013488188 s

What was this caused by again? Which part of ath_calibrate() was
causing the issue?
Is it the noise floor save/load? IQ calibration? ANI? Anything else?

> Jan  1 00:16:45 wlanmon: (RTM_IEEE80211_ASSOC) [wlan2] [00:c0:c1:cb:6c:c3]
> lock order reversal:
>  1st 0xc59126c0 ath1_node_lock (ath1_node_lock) @ /usr/src/sys/net80211/ieee80211_ioctl.c:1326
>  2nd 0xc5911014 ath1_com_lock (ath1_com_lock) @ /usr/src/sys/net80211/ieee80211_node.c:2510

Blah. So the above is in setmlme_dropsta() -> domime() ->
ieee80211_node_leave() (MLME_DISASSOC or MLME_DEAUTH in hostap mode,
on a broadcast address).

What's the second path?  How the heck does ieee80211_parse_beacon()
call ieee80211_node_leave() ?



Adrian