Wpa_supplicant & netbsd-current & wi-driver

[Risto Sainio]

On Sunday 06 April 2008 20:33, Sam Leffler wrote:
> Jouni Malinen wrote:
> > On Fri, Apr 04, 2008 at 08:27:29PM +0200, Risto Sainio wrote:
> >> As I happen to have the documents for the chipset I started to modify
> >> the driver and after a while I got it so far fixed that it scans,
> >> associates and authorizes. This all is done by the 802.11 layer written
> >> for netbsd, that means that I configured the card to run in hostap-mode
> >> and the 802.11-layer sends/receives all necessary frames and runs its
> >> own state-machine.
> >
> > This sounds somewhat complex way of using Prism2/2.5/3 cards in client
> > mode.. In theory, you might be able to make this work, but Host AP mode
> > is designed for AP, not client functionality.. I've seen number of
> > firmware version specific issues in doing this.
> >
> > If you're main goal is to just make the client mode work with WPA-PSK, I
> > would suggest using the standard client mode for this and fixing the
> > driver to support this..
> >
> >> Now I am stuck with the 4-way handshake: namely the hostapd sends the
> >> first EAPOL-message and it seems to me that when my card receives that
> >> dataframe it replies with deauth reason:3.
> >
> > That sounds like one of the main problems I had in using Host AP mode
> > with certain firmware versions.. The firmware has some sort of
> > bug/feature that ends up sending out deauth when trying to trick it to
> > act as a client in Host AP mode. This does not happen with all firmware
> > versions, though. I don't remember all the details, but I think this was
> > something that worked with old firmware versions, but not new..
> >
> >> Now my question is: How can I tell the firmware that it is already
> >> associated and authorized OR do I have to issue a JOIN request ??
> >
> > I don't know whether there is such a mechanism in Host AP mode. You are
> > trying to do something that the firmware was not designed to do.. If you
> > want to avoid random tests and experiments to figure out what exactly
> > the buggy firmware does, you would probably be better off using the
> > standard client mode instead.
>
> Look here:
>
> http://perforce.freebsd.org/depotTreeBrowser.cgi?FSPC=//depot/projects/vap/
>sys/dev/wi
>
> That's a working wi driver that does WPA (both sta+ap).  You want to use
> monitor mode to scan and then operate as a sta using the normal firmware
> support.
>
>     Sam
Thank you both

I will check the freebsd-version - this was originally my plan B as well, but 
the 802.11-layer looked first like the lazy-man way of doing it as the layer 
takes care of the frames.

risto