Problems and [possibly] bugs with WDS

[Jouni Malinen]

On Sat, Dec 14, 2002 at 11:43:16PM +0100, Ricardo Galli wrote:

> 1. When there is traffic, which is very slow, among the two AP I get the following 
> errors in both APs:

> Dec 14 23:16:50 ponti kernel: wlan0: STA 00:05:5d:d6:4b:da TX rate raised to 55
> Dec 14 23:16:50 ponti kernel: wlan0: STA 00:05:5d:d6:4b:da TX rate raised to 110
> Dec 14 23:16:51 ponti kernel: wlan0: STA 00:05:5d:d6:4b:da TX rate lowered to 55

I would not call these errors.. Automatic TX rate controlling is
lowering the rate whenever a packet is not ACKed and then raising it if
packets are not lost for some time.

> 2. Sometimes it seems they try to "bypass" wds: 
> Dec 14 23:22:07 raquel kernel: wlan0: deauthentication: 00:50:c2:01:96:14 len=2, reason_code=3

That is odd.. It looks like one of the APs sent deauthenticatin frame
even though they should not be authenticated/associated at all in this
kind of configuration.. I haven't seen this happening before. It might
be possible that station firmware is sending deauthentication frame even
in Host AP mode under some circumstances (e.g., when port0 is disabled &
enabled for configuration change).

> 3. Sometimes gives the following error:
> Dec 14 22:51:27 ponti kernel: AP: packet to non-associated STA 00:40:43:05:66:00
> But 00:40:43:05:66:00 is an ethernet! (Nokia ethernet router).

I have also seen this many times. I would assume that Linux bridge code
did not know where to find that hwaddr and decided to send the frame to
all interfaces just in case, or something like this. Host AP driver
would print this if such frame is sent using wlan0 interface.

> 4. If a wireless station C is connected to AP-B:
> C ~~~~~ AP-B ~~~wds~~~~~ AP-A
> AP-A cannot connect to C for a long period (more than 5 minutes) if 
> before was connected to AP-A. I'm not sure if it's a problem of the STP or WDS...

If you mean that C was first associated with AP-A and then reassociated
with AP-B, then I would assume this problem occurs due to AP-A still
having the association up for C. At the moment, old AP does not get
any information when an associated station moves to another AP. IAPP
should eventually remove this problem.

> And one not related to WDS: if hostap is compiled as module against 2.4.20 and the 
> last pcmcia modules version, hostap_crypt_wep is not loaded automatically.

It should be loaded when you are setting up WEP keys, assuming your
module loader is working properly. At least hostap.o should request that
module automatically.

-- 
Jouni Malinen                                            PGP id EFC895FA