I just removed it from the kernel because hostapd doesn't really use it.
Kill all references.
Signed-off-by: Johannes Berg <[email protected]>
--- hostap.orig/hostapd/driver_devicescape.c 2007-08-10 23:27:07.000000000 +0200
+++ hostap/hostapd/driver_devicescape.c 2007-08-10 23:28:21.000000000 +0200
@@ -1413,9 +1413,15 @@ static void handle_frame(struct hostapd_
case ieee80211_msg_tx_callback_fail:
handle_tx_callback(hapd, buf, data_len, 0);
return;
+/*
+ * TODO
+ * the kernel never sends this any more, add new nl80211
+ * notification if you need this.
+
case ieee80211_msg_wep_frame_unknown_key:
- /* TODO: ieee802_11_rx_unknown_key(hapd, buf, data_len); */
+ ieee802_11_rx_unknown_key(hapd, buf, data_len);
return;
+ */
case ieee80211_msg_michael_mic_failure:
hostapd_michael_mic_failure(hapd, buf, data_len);
return;
On Wed, Aug 15, 2007 at 12:45:59PM +0200, Johannes Berg wrote:
> On Tue, 2007-08-14 at 20:35 -0700, Jouni Malinen wrote:
> > However, this was causing
> > interoperability issues with some client implementations and the code
> > was eventually commented out and this was only used to update a dot11
> > MIB counters. Consequently, I did not see much point in integrating it
> > into hostapd and we might as well just remove the TODO item completely
> > from hostapd.
> I'm not sure whether it'd be worth it going forward anyway, with WEP
> deployments hopefully going down.
I don't think we can enable the original behavior of disassociating a
station based on use of incorrect key index since it can cause problems
with commonly used clients (well, clients that were commonly used four
years ago ;-) and there are also potential issues with opening yet
another DoS problem. If needed, we could have a MIB counter for the
errors updated, but I'm not too concerned about not having this
available.
--
Jouni Malinen PGP id EFC895FA
On Tue, 2007-08-14 at 20:35 -0700, Jouni Malinen wrote:
> This feature was originally used to allow a station to be disassociated
> if it seemed to be using incorrect keys.
Makes sense, this is a bit icky.
> However, this was causing
> interoperability issues with some client implementations and the code
> was eventually commented out and this was only used to update a dot11
> MIB counters. Consequently, I did not see much point in integrating it
> into hostapd and we might as well just remove the TODO item completely
> from hostapd.
I'm not sure whether it'd be worth it going forward anyway, with WEP
deployments hopefully going down.
johannes
On Fri, Aug 10, 2007 at 11:29:26PM +0200, Johannes Berg wrote:
> I just removed it from the kernel because hostapd doesn't really use it.
> Kill all references.
Thanks, applied. The current wireless-dev.git seems to still have code
for this, but since it was not currently used in hostapd, I'm fine with
this patch.
This feature was originally used to allow a station to be disassociated
if it seemed to be using incorrect keys. However, this was causing
interoperability issues with some client implementations and the code
was eventually commented out and this was only used to update a dot11
MIB counters. Consequently, I did not see much point in integrating it
into hostapd and we might as well just remove the TODO item completely
from hostapd.
--
Jouni Malinen PGP id EFC895FA