I fix the issue with this patch
--- a/net/mac80211/sta_info.c 2013-04-17 16:47:16.000000000 +0200
+++ b/net/mac80211/sta_info.c 2013-04-17 16:50:02.697623684 +0200
@@ -1036,7 +1036,8 @@ void ieee80211_sta_ps_deliver_wakeup(str
BUILD_BUG_ON(BITS_TO_LONGS(IEEE80211_NUM_TIDS) > 1);
sta->driver_buffered_tids = 0;
- if (!(local->hw.flags & IEEE80211_HW_AP_LINK_PS))
+ if (!(local->hw.flags & IEEE80211_HW_AP_LINK_PS) &&
+ sta->sta_state >= IEEE80211_STA_ASSOC)
drv_sta_notify(local, sdata, STA_NOTIFY_AWAKE, &sta->sta);
skb_queue_head_init(&pending);
Is it correct ?
Cedric Voncken
-----Message d'origine-----
De?: [email protected]
[mailto:[email protected]] De la part de voncken
Envoy??: mercredi 17 avril 2013 15:57
??: [email protected]; [email protected]
Objet?: RE: ath9k null pointer with mesh network
I understand why the ac pointer is null in ath_tx_aggr_wakeup
function.
My wireless card receive packet, the ieee80211_rx function is
called. This function call indirectly ath_tx_aggr_wakeup function and my
kernel crash because the ac pointer is null.
The Ac pointer is null because the sta_state is set to
IEEE80211_STA_NONE, and on this state the function ath_tx_node_init (in
net/wireless/ath/ath9k/xmit.c) is not already called (call only when the
sta_state go to IEEE80211_STA_ASSOC from IEEE80211_STA_AUTH )
Have you any idea to fix it ?
The kernel panic call stack
ath_tx_aggr_wakeup+0x64/0xe4 [ath9k] (unreliable) [ 128.340486]
[c7ffbd10] [c92b0b34] ath9k_deinit_device+0x8d0/0xb28 [ath9k] [ 128.347324]
[c7ffbd20] [c9173cf4]
ieee80211_sta_ps_deliver_wakeup+0xa4/0x1e4 [mac80211] [ 128.355365]
[c7ffbd50] [c91a3f74]
ieee80211_mps_sta_status_update+0xe4/0x158 [mac80211] [ 128.363403]
[c7ffbd70] [c9186f28]
ieee80211_sta_ps_transition+0xff0/0x34e4 [mac80211] [ 128.371268]
[c7ffbe00] [c91893b0]
ieee80211_sta_ps_transition+0x3478/0x34e4 [mac80211] [ 128.379219]
[c7ffbe40] [c9189ae8] ieee80211_rx+0x6cc/0x798 [mac80211] [ 128.385678]
[c7ffbea0] [c92b43c4] ath_rx_tasklet+0xe80/0xf74 [ath9k] [ 128.392039]
[c7ffbf70] [c92b1fdc] ath9k_tasklet+0xd0/0x114 [ath9k] [ 128.398219]
[c7ffbf90] [c0020b04] tasklet_action+0x88/0xe0 [ 128.403694] [c7ffbfb0]
[c0021228] __do_softirq+0xa4/0x12c [ 128.409088] [c7ffbff0] [c000cec8]
call_do_softirq+0x14/0x24 [ 128.414655] [c73b7ee0] [c000550c]
do_softirq+0x64/0xa4 [ 128.419783] [c73b7f00] [c0021464] irq_exit+0x48/0x74
[ 128.424739] [c73b7f10] [c0005424] do_IRQ+0x108/0x120 [ 128.429697]
[c73b7f40] [c000ec5c] ret_from_except+0x0/0x14
Cedric Voncken
-----Message d'origine-----
De?: [email protected]
[mailto:[email protected]] De la part de Cedric VONCKEN
Envoy??: mardi 16 avril 2013 18:26 ??: [email protected]
Objet?: ath9k null ppointer
I tried to use a mesh network with security (meshd/authsae).
I am using :
- openwrt with the latest mac80211 package from openwrt
(2013-02-22).
- the Thomas Pedersen patch to add the authsae package in my
openwrt source tree
(http://permalink.gmane.org/gmane.linux.drivers.mesh.open80211s/2530)
(Use the latest version, because I sent a patch in meshd this morning,
without it the HT mode should not correctly enable)
- AR9xx wirless card
- Kernel 3.3.8 on power PC (mpc 8314)
When I configure the meshd daemon to use the HT mode, I have a null pointer
in ath9k (That work correctly without HT mode).
In the file, drivers/net/wireless/ath/ath9k in the function
ath_tx_aggr_wakeup the ac pointer is null.
Do you know why this pointer can be null?
Cedric Voncken