WPA2 interoperability problems between hostapd 0.4.10 and wpa_supplicant 0.5.9?

Clem Taylor clem.taylor at gmail.com
Mon Mar 24 14:26:17 EDT 2008


On Mon, Mar 24, 2008 at 1:14 PM, Jouni Malinen <j at w1.fi> wrote:
> On Mon, Mar 24, 2008 at 12:32:28PM -0400, Clem Taylor wrote:
>
>  > I'm using wpa_supplicant v0.5.9 and I've been having problems keeping
>  > connected to a wireless router that is running hostapd 0.4.10. I'm
>  > using WPA2-PSK (RSN) and sometimes when the client looses its
>  > connection (AP reboot and random connection loss) it fails to
>  > re-authenticate with the AP.
>
>  Which driver are you using in the client and in the AP?

The client side is using madwifi 0.9.4.1 with AR5212s (one node has an
A/B/G radio the other is A only).

The AP is also AR5212 based, but with older drivers:
wlan: 0.8.4.2 (rev  2007-03-31 18:23:46)
ath_hal: 0.9.18.0
ath_pci: 0.9.4.5 (rev  2007-03-31 18:23:46)

>  > But other times it doesn't produce an error message. If I terminate
>  > the wpa_supplicant and restart, the connection will sometimes recover,
>  > but other times it will continue to fail to authenticate. If I
>  > terminate wpa_supplicant, wait several minutes and then restart the
>  > connection always seems to recover.
>
>  This could indicate that the driver and wpa_supplicant in the client get
>  out-of-sync on PMK entries (PMKID is included in the RSN IE which has to
>  match between (Re)Association Request and msg 1/2 and these are from the
>  driver and wpa_supplicant, respectively).

What I really don't understand is why this out-of-sync condition
repeats continuously. Running 'disconnect' and 'reconnect' with
wpa_cli doesn't seem to help. Likewise, killing off wpa_supplicant and
restarting doesn't help. I have to kill off wpa_supplicant, wait
several minutes and then restart wpa_supplicant and then it is happy
again.

I have also managed to trigger this problem just rebooting the AP (but
this has been difficult to reproduce). Actually, this problem in
general is hard to reproduce, I ran a test over the weekend without
any problem (~.6TB transferred), but it seems to fail nightly during
the week.

                               Thanks,
                               Clem Taylor


More information about the HostAP mailing list