Re: Wireless connection (WPA-EAP) stops working after a while

Sam Leffler wrote:
> Daniel Thiele wrote:
>> Sam Leffler wrote:
>>> Daniel Thiele wrote:
>>>> Hi,
>>>>
>>>> I have a problem with my wireless connection. I am running FreeBSD
>>>> 8.0-CURRENT (from Mar 22) with wpa_supplicant v0.6.8 using an Atheros
>>>> based ExpressCard (D-Link DWA-643 via ath(4)) and alternatively a 
>>>> Ralink
>>>> based USB adapter (Linksys WUSB54GC-EU via rum(4)).
>>>>
>>>> At home using the Atheros card together with a FreeBSD (7.1) based
>>>> access point (using rum(4) in hostap mode) and the wpa_supplicant.conf
>>>> (attached at the end of this email) settings for SSID="home" I don't
>>>> have any problems. With a Linksys WRT54GL-DE access point
>>>> running the OpenWRT White Russian 0.9 firmware and OpenVPN over an open
>>>> wireless LAN also works flawlessly.
>>>>
>>>> At the university, however, (SSID="IDA" in the wpa_supplicant.conf at
>>>> the end of this email) the wireless connection only works for about an
>>>> hour. The vague term "wireless connection" in this case means, that the
>>>> WPA connection is opened and associated, then I get an IP address via
>>>> dhclient. There is a message about "OpenSSL: tls_connection_handshake -
>>>> Failed to read possible Application Data
>>>> error:00000000:lib(0):func(0):reason(0)" and "TLS: Unsupported Phase2
>>>> EAP method 'MSCHAPv2'" but the authentication seems to succeed:
>>>>
>>>
>>> Do you know if this is a regression w/ the 0.6.8 supplicant I recently
>>> imported (i.e. relative to 0.5.11+HEAD)?  
>>
>> Unfortunately I cannot answer that, since I have not been running
>> CURRENT for that long. Is there a way I can help to test that? I still
>> have my old notebook (7.1-STABLE running wpa_supplicant 0.5.10 with ipw
>> and rum wireless adapters). I can get the logs from that machine, too,
>> if that would help.
> 
> You could rollback wpa_supplicant using svn/cvs but it's a bit painful 
> as the changes were extensive due to the switch from the 0.5 series 
> releases to 0.6.  Alternatively you could rollback your entire tree to 
> r189262 (I think) and you'd have the previous code.
> 
>>
>>> Jouni just released a 0.6.9
>>> version and it appeared to have a fix that might be appropriate.  I need
>>> to digest your logs to understand but you can try importing 0.6.9 
>>> yourself.
>>
>> I will give it a try. Just now I merged the diffs between 0.6.8 and
>> 0.6.9 to my /usr/src. Except for
>> wpa_supplicant-0.6.8/src/common/nl80211_copy.h,
>> wpa_supplicant-0.6.9/src/l2_packet/l2_packet_linux.c and,
>> wpa_supplicant-0.6.9/wpa_supplicant/Makefile
>> I was able to apply the diffs.
> 
> We do not use any of these files.  You might be better off looking at 
> diff's between 0.6.8 and 0.6.9 and selectively applying changes to 
> usr/src/contrib/wpa.  Some of these may be ignored as Jouni took back 
> some changes I provided him.

That's what I did. Sorry, if I was too vague about that in my reply
above. One thing I also noticed was that OID_802_3_MULTICAST_LIST which
is used in 0.6.9's src/drivers/driver_ndis.c by ndis_add_multicast()
does not seem to be defined anywhere in the 0.6.9 sources, which caused
make in usr.sbin/wpa to fail. Since I am not using the NDIS driver I
made ndis_add_multicast() return -1 to make make work again.

> Note that usr.sbin/wpa/wpa_supplicant has the driver for freebsd and a 
> reach-over Makefile; we do not use Jouni's bsd driver or his build 
> structure.
> 
>>
>> I will try this at university on Thursday or Friday and report any
>> success or failure here.
> 
> I'll try to look at merging 0.6.9 quickly.
> 

I am using the patched wpa_supplicant right now from home, so at least
didn't break anything obvious. I will try either this setup or
(preferred) your merged version at university and see how far I can get.


Thanks,

Daniel