wpa_supplicant: isn't "802.1x Start" required for WPA?
Tue Apr 19 19:45:26 PDT 2005
On Tue, Apr 19, 2005 at 08:07:50PM +0200, Arjan van Bentem wrote:
> Trying to figger out why my Linksys card works on Windows, but not on
> Linux I used Ethereal to see what's actually sent. I noticed some
> differences. Any comment?
Getting a raw Ethereal capture file would be helpful..
> As a response to the initial key as sent by the access point, the
> Windows XP driver first seems to send a "Start", shown by Ethereal like:
> 802.1x Authentication
> Version: 1
> Type: Start (1)
> Length: 0
> Does anyone know if this "Start" might be required for some access
> points (in my case: a Speedtouch 580 access point)? wpa_supplicant /
> ndiswrapper do not send such "Start" but sends the EAPOL WPA key right
EAPOL-Start frame is not part of WPA-PSK and I'm not aware of any AP
requiring it. Yes, I know that Microsoft supplicant sent this, but it is
not correct behavior as far as I can tell.
> The first EAPOL WPA key message as sent by wpa_supplicant is almost the
> same as the Windows message. However: the value for "Key Length" is zero
> on Windows, but not in wpa_supplicant.
This field is not really used in WPA and as such, I would not expect
authenticator to verify that it is zero. Anyway, I did actually change
wpa_supplicant to use zero in this field. This change is currently only
in the development branch, so you would need to download a development
snapshot in order to test this.
> And the Windows driver sends a
> value for "WPA Key" that is 2 bytes longer and ends in two zero bytes:
This is fine. The difference is in whether the optional WPA capabilities
field is included in WPA IE or not.
> whereas the response when using wpa_supplicant does not have the
> "Install flag"and "Key Ack flag" set:
This packets looks odd.. I would need to see full headers to see what it
is, but it looks more like a reply from the client (i.e., msg 2/4) than
anything from the AP..
Jouni Malinen PGP id EFC895FA
More information about the Hostap