On Wed, 2013-09-25 at 11:46 -0500, Larry Finger wrote: > On 09/25/2013 07:47 AM, Dan Williams wrote: > > On Tue, 2013-09-24 at 17:43 -0500, Larry Finger wrote: > >> On 09/09/2013 05:15 PM, Dan Williams wrote: > >>> On Mon, 2013-09-09 at 16:52 -0500, Larry Finger wrote: > >> > >>>> I have been running rtl8192cu for the past 24 hours without a permanent > >>>> disconnect. Under NetworkManager, I see some reason 7 deauthentications, but > >>> > >>> Running wpa_supplicant with debugging on might shed some light on these; > >>> basically: > >>> > >>> mv /usr/sbin/wpa_supplicant / > >>> killall -TERM wpa_supplicant > >>> /wpa_supplicant -dddtu <piped to your favorite log file> > >>> > >>> and NM should automatically reconnect, and then we can figure out what's > >>> going on in the supplicant. > >> > >> Dan, > >> > >> The log of wpa_supplicant associated with the reason 7 disconnects are as follows: > > > > So reason 7 is "Incorrect frame type or subtype received from > > unassociated station" which seems like the AP thinks we got > > disconnected, and would seem to be a driver/mac80211 issue still, right? > > Yes. These only happen with rtl8192ce and rtl8192cu. They are a bit more common > when running NetworkManager than with ifup. In my current run, they have been at > intervals of 1000 to 30,000 seconds apart. Capturing them with wireshark may not > be practical. > > --snip-- > > > And got reconnected after a bit more than one second. So at least it > > recovers quickly, but the question is more about why the reason 7 > > happened, and what frames caused it, I think. > > I agree. The sequence seems to start with an MLME Event 39: > > .908249: nl80211: MLME event 39 > .908252: nl80211: MLME event frame - hexdump(len=26): c0 00 3a 01 1c 65 9d 5a c3 > 9d 20 e5 2a 01 f7 ea 20 e5 2a 01 f7 ea a0 f6 07 00 > .908269: wlan3: Event DEAUTH (12) received > .908273: wlan3: Deauthentication notification > > All that happens within 25 usec, but I have no clue what triggers that. In > addition, I have been unable to find any documentation on MLME events. Any > suggestions regarding a source would be appreciated. Periodic scanning maybe and some mis-management of nullfunc frames in the driver when switching channels for a scan? That used to be the most common cause of issues like this, where the device would go scan a channel but forget the nullfunc, so while it was off-channel the AP wouldn't hear anything from it and disassociate it. Dan -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html