B43 not seeing MadWifi AP

B43 not seeing MadWifi AP

[Andrew Lunn]

Hi Folks

I have a number of Linux laptops running kernel 2.6.23.9, and
compat-wireless-2.6 with what git-describe says is
v2.6.24-rc7-716-g62ec51e and i am using the b43 driver with a BCM4312
device.

I also have some laptop with Atheros PCCard devices and the Madwifi
driver from SVN, version r3318.

The Madwifi device is running as an AP. The problem i have is that the
b43 does not see it. iwlist wlan0 scan does not see this madwifi
AP. It does see other APs, but not my madwifi driver. Other madwifi
devices see the AP. Also an ipw2200 device sees the AP. So i know the
AP is running, sending beacons, etc. It just seems that the b43 does
not like it. 

iwlist ath0 scan lists the "missing" AP as:

   Cell 04 - Address: 06:03:7F:03:42:0B
             ESSID:"TOWN-SSID"
             Mode:Master
             Frequency:2.442 GHz (Channel 7)
             Quality=49/70  Signal level=-46 dBm  Noise level=-95 dBm
             Encryption key:off
             Bit Rates:1 Mb/s; 2 Mb/s; 5.5 Mb/s; 6 Mb/s; 9 Mb/s
                       11 Mb/s; 12 Mb/s; 18 Mb/s; 24 Mb/s; 36 Mb/s
                       48 Mb/s; 54 Mb/s
             Extra:bcn_int=100
             Extra:wme_ie=dd180050f2020101830002a4400027a4000042435e0062322f00
             Extra:ath_ie=dd0900037f01010034ff7f

Is the b43 picky about Extra things in the beacon? 

At the beginning of ieee80211_rx_bss_info() there is some commented
out printk for all received beacons and probe responses. I enabled
this and don't see anything for the missing device. So it looks like
the problem is lower down the stack. However i don't see any obvious
place were it would be discarding beacons.

Anybody got any ideas?

        Thanks
                Andrew

Re: B43 not seeing MadWifi AP

[Pavel Roskin]

On Wed, 2008-02-06 at 10:59 +0100, Andrew Lunn wrote:
> iwlist ath0 scan lists the "missing" AP as:
> 
>    Cell 04 - Address: 06:03:7F:03:42:0B

This MAC address is not reserved for any vendor, but MAC addresses
starting with 00:03:7f belong to Atheros.  I think it indicates the
Atheros card uses several virtual interfaces, one of which is what you
are seeing.

I suggest that you check in the monitor mode that there are indeed
beacons sent with this MAC address.  Maybe they are not.  It's quite
possible.  MadWifi has an option to disable beacons for some virtual
APs.

Maybe mac80211 decided to do a passive scan, perhaps due to your country
regulations?

It's quite possible that b43 will associate if you set the ESSID
correctly.  Maybe you'll need to set "ap" with iwconfig as well.

-- 
Regards,
Pavel Roskin

Re: B43 not seeing MadWifi AP

[Andrew Lunn]

On Wed, Feb 06, 2008 at 01:17:50PM -0500, Pavel Roskin wrote:
> On Wed, 2008-02-06 at 10:59 +0100, Andrew Lunn wrote:
> > iwlist ath0 scan lists the "missing" AP as:
> > 
> >    Cell 04 - Address: 06:03:7F:03:42:0B
> 
> This MAC address is not reserved for any vendor, but MAC addresses
> starting with 00:03:7f belong to Atheros.  I think it indicates the
> Atheros card uses several virtual interfaces, one of which is what you
> are seeing.

Correct. 

> I suggest that you check in the monitor mode that there are indeed
> beacons sent with this MAC address.  Maybe they are not.  It's quite
> possible.  MadWifi has an option to disable beacons for some virtual
> APs.

Well, other devices are seeing it, eg other Atheros devices and intel
IPW2200. So somehow i doubt it is not sending beacons.

 
> Maybe mac80211 decided to do a passive scan, perhaps due to your country
> regulations?

CH has no such regulations... 
 
> It's quite possible that b43 will associate if you set the ESSID
> correctly.  Maybe you'll need to set "ap" with iwconfig as well.

I hacked the madwifi driver a little, to disable the wme_ie and the
ath_ie IEs in the beacon and probe response messages. Only then did
the b43 device list the Atheros device in its scan list. However i
could not get it to associate. Authentication always failed. The b43
driver gives error messages and the MadWifi driver never received the
Authentication request messages.

In the end i gave up. I have another machine with an ipw2100 card
which worked first time and i don't see any point using the B43 device
when it causes me to much pain.

     Andrew

Re: B43 not seeing MadWifi AP

[Pavel Roskin]

Quoting Andrew Lunn <andrew.lunn@ascom.ch>:

>> I suggest that you check in the monitor mode that there are indeed
>> beacons sent with this MAC address.  Maybe they are not.  It's quite
>> possible.  MadWifi has an option to disable beacons for some virtual
>> APs.
>
> Well, other devices are seeing it, eg other Atheros devices and intel
> IPW2200. So somehow i doubt it is not sending beacons.

What I have found is that there are beacons for every ESSID, but  
scanning finds only one ESSID.  I tested that with orinoco driver,  
which doesn't depend on mac80211.

>> Maybe mac80211 decided to do a passive scan, perhaps due to your country
>> regulations?
>
> CH has no such regulations...

You are right, it's more like the opposite is happening.  It looks  
like mac80211 relies only on probe requests and not on beacons.

>> It's quite possible that b43 will associate if you set the ESSID
>> correctly.  Maybe you'll need to set "ap" with iwconfig as well.
>
> I hacked the madwifi driver a little, to disable the wme_ie and the
> ath_ie IEs in the beacon and probe response messages. Only then did
> the b43 device list the Atheros device in its scan list. However i
> could not get it to associate. Authentication always failed. The b43
> driver gives error messages and the MadWifi driver never received the
> Authentication request messages.

It seems to me that the problem is on the MadWifi side.  Changing the  
code may be irrelevant.  MadWifi replied to probe requests only on one  
virtual interface,  and this time it was the interface you needed (for  
whatever reason).  Maybe it switched back when you tried to associate.

> In the end i gave up. I have another machine with an ipw2100 card
> which worked first time and i don't see any point using the B43 device
> when it causes me to much pain.

Maybe b43 does something different, so that the trick with virtual  
devices doesn't work.  But it looks like some problems don't even need  
b43 on the client side.

-- 
Regards,
Pavel Roskin