Bug#403727: [pkg-wpa-devel] Bug#403727: Disconnects every few seconds on madwifi

[Kel Modderman]

Hi Maik,

On Tuesday 19 December 2006 21:45, Maik Zumstrull wrote:
> Package: wpasupplicant
> Version: 0.5.5-4
>
> When using wpa_supplicant in roaming mode on current Debian madwifi
> (see attachment: package-versions), the connection drops every thirty
> seconds or so.

MadWifi sucks a bit in certain situations. Numerous people have reported 
scanning/association/mode changing issues.

> This behavior is completely reproducible and occurs on any networks I
> have tested it with, among them:
> - My university WLAN with WPA-EAP
> - My home network with WPA-PSK
> - My university WLAN without any encryption
> - The network of a local cafe using WEP-40
>
> Each test on the university network has been performed in different
> buildings, which use compatible configurations, but completely
> different hardware (LANCOM vs. Trapeze Networks; DHCP and Radius
> servers are shared).
>
> Where this is an option (i.e. the above networks without actual WPA), I
> also tested using only the driver and ifupdown, without starting
> wpa_supplicant. In this scenario, the dropouts do not occur.

Dropouts do not occur? Or dropouts are not "noticed" because ifup/ifdown is 
not called when the driver borks? The roaming setup you use is very sensitive 
to driver stability, or lack of. Using static ifupdown stanzas on the other 
hand is quite insensitive to transient dropouts.

Is it your opinion that these transient lapses in operation _only_ occur in 
conjunction with wpa_supplicant? Do you experience any momentary pauses in 
operation when not using wpa_supplcant? 

>
> Note: I am aware that madwifi should theoretically be usable with both
> -Dmadwifi and -Dwext, and that the latter is actually the recommended
> configuration. On my system, -Dwext doesn't work at all, any
> association attempt fails. With -Dmadwifi, the network is usable (for
> seconds at a time) with the dropouts this report is about.
>
> Further attached information:
> - The wpa_supplicant and ifupdown configuration files (shortened to
>   relevant sections, with private information removed; access to full
>   files can be granted via private mail if this is deemed necessary.)
> - The output of a session of wpa_cli with debug level 0, time-stamped.
>   I waited for several dropouts to occur before ending the session.
>   The time-stamps appear to be pretty useless since wpa_cli seems to
>   dump its output in big chunks at a time.
>
> In the hope that I didn't just send this report and it turns out to be
> a small dumb misconfiguration on my part,

The attached files show no obvious configuration errors, so no, you have done 
nothing dumb in my opinion.

Thanks, Kel.