Jouni Malinen [Sun, 30 Oct 2011 20:37:12 +0000 (22:37 +0200)]
Process RSN pre-authentication candidates when skipping roam
wpa_supplicant_rsn_preauth_scan_results() needs to be called to
update RSN pre-authentication candidates. This cannot be done before
the wpa_supplicant_connect() call on the first association, but when
trying to figure out whether to roam, it is fine to do so for the
case when roaming is skipped.
Yoni Divinsky [Sun, 30 Oct 2011 20:19:49 +0000 (22:19 +0200)]
Configure new GTK/IGTK to driver in case of TKIP countermeasures
The GTK is renewed in the hostapd after a MIC attack dissassociation
without informing the driver, causing decryption failures. This patch
sends the new GTK/IGTK to the driver after it is updated by the hostapd.
Olivier Sobrie [Sun, 30 Oct 2011 20:10:40 +0000 (22:10 +0200)]
WPS: Send the credential when learning AP params in registrar role
When the supplicant acts as a registrar to learn the access point
parameters send the credentials to the wpa_cli interface after
receiving the 7th message. This is needed for proper behavior with
wps_cred_processing set to 1 or 2.
Without this patch, after the 7th message you got the WPS-CRED-RECEIVED
notification without the credentials. This was because the cred_attr and
cred_attr_len were not filled in in the wps structure.
Gary Morain [Fri, 5 Aug 2011 23:23:12 +0000 (16:23 -0700)]
Prefer 5 GHz networks over 2.4 GHz networks
In scan.c, merge a channel's noise value into the scan results. When
comparing scan results, compute the signal-to-noise ratio and use it
when available. Prefer a 5 GHz network if its SNR is really big (> 30)
or if its SNR is relatively close to the other network's.
Jouni Malinen [Sun, 30 Oct 2011 10:49:26 +0000 (12:49 +0200)]
Ignore Michael MIC failure reports if cipher is not TKIP
Some stations have been reported to send EAPOL-Key Error Reports
indicating Michael MIC failures even when the cipher is not TKIP
(e.g., when the network is using only CCMP). Ignore such reports
to avoid starting TKIP countermeasures unnecessarily. This can
prevent certaint types of denial of service attacks by insiders,
but mostly this is to work around invalid station implementations.
Dan Williams [Mon, 24 Oct 2011 16:03:04 +0000 (11:03 -0500)]
dbus: Implement EAP SM control request signals
Add a D-Bus signal for EAP SM requests. This signal is emitted on the
Interface object so that clients only have to listen to one object for
requests rather than to all network objects. This signal is analogous
to the socket control interface's CTRL-REQ- request.
Dan Williams [Mon, 24 Oct 2011 16:00:19 +0000 (11:00 -0500)]
Use an enum for EAP SM requests
Control requests will be extended for non-EAP uses later, so it makes
sense to have them be generic. Furthermore, having them defined as an
enum is easier for processing internally, and more generic for control
interfaces that may not use field names. The public ctrl_req_type /
field_name conversion function will be used later by the D-Bus control
interface too.
Jouni Malinen [Sun, 30 Oct 2011 10:02:20 +0000 (12:02 +0200)]
EAP-TTLS peer: Fix user input during implicit identity request
Special processing is needed to handle EAP user request for
identity or password at the beginning of Phase 2 when the implicit
identity request is used. data->pending_phase2_req needs to be set
to an empty buffer in that case to avoid re-processing the previous
part of TLS negotiation when the user enters the needed information.
Jouni Malinen [Sun, 30 Oct 2011 09:41:22 +0000 (11:41 +0200)]
Skip rate set matching if BSS frequency is not known
At least with driver_test.c, the BSS table may not include frequency
information. In such a case, we need to skip rate set matching during
BSS selection to avoid rejecting valid AP.
Jouni Malinen [Sat, 29 Oct 2011 20:02:47 +0000 (23:02 +0300)]
P2P: Indicate scan during group formation as P2P probe
The interface may not yet have been set to P2P client mode immediately
after GO Negotiation when doing the scan before association request.
Consequently, the scan request in this state has to specify that the
specific P2P probe mode is used to disable IEEE 802.11b rates.
Jouni Malinen [Sat, 29 Oct 2011 19:46:58 +0000 (22:46 +0300)]
nl80211: Disable IEEE 802.11b rates only for P2P iftypes
Instead of hardcoding IEEE 802.11b rates to be disabled whenever
P2P support is built in and supported by the driver, do this only
when an interface is set to P2P mode (both when adding a new
interface and when changing the interface mode). This re-enables
use of IEEE 802.11b APs with CONFIG_P2P=y builds.
Jouni Malinen [Sat, 29 Oct 2011 19:25:28 +0000 (22:25 +0300)]
Remove disable_11b_rates() driver_ops
This function was used unconditionally if wpa_supplicant build
includes CONFIG_P2P=y. Adding a separate driver_ops for such use
is not really useful since the driver wrappers can do the same
internally. Remove this driver_ops and move matching functionality
into driver_nl80211.c which was the only driver wrapper using
this driver_ops callback.
Jouni Malinen [Sat, 29 Oct 2011 19:12:26 +0000 (22:12 +0300)]
nl80211: Fix re-enabling of 802.11b rates
nl80211_disable_11b_rates() function was supposed to use the 'disabled'
parameter to figure out whether to disable or re-enable 802.11b rates.
In addition, the driver deinit path was now ending up re-disabling the
enabled rates at the end of the deinit operations when the interface
type was forced back to station. Both of these issues are now fixed
and the 2.4 GHz band TX rate index is cleared properly when
wpa_supplicant is stopped.
Jouni Malinen [Sat, 29 Oct 2011 18:49:46 +0000 (21:49 +0300)]
Add no_cck parameter for send_action() driver_ops
This can be used to apply the no-CCK rule conditionally depending on
which frame is being sent. The no-CCK rule applies only for P2P
management frames while SA Query and FT use cases do not have similar
restrictions.
nl80211: Add support to disable CCK rate for P2P frames
Add a new attribute to specify whether to use CCK rate or not during
scan and sending management frames. This helps to send P2P probes
at non-CCK rate. This ensures that P2P probe request/response/action
frames are always sent at non-CCK rates.
Daniel Drake [Sat, 29 Oct 2011 16:23:42 +0000 (19:23 +0300)]
dbus: Raise PropertiesChanged on org.freedesktop.DBus.Properties
wpa_supplicant dbus objects are currently difficult to use with
GDBusProxy. This is because they do not follow the dbus standard in
emitting the PropertiesChanged signal on the standard D-Bus properties
interface, so GDBusProxy stores stale property values.
Fix this by raising PropertiesChanged on the standard interface.
For backwards compatibility, the existing custom PropertiesChanged
signals are not removed; I anticipate them being removed in a future
major release.
See
http://lists.freedesktop.org/archives/dbus/2011-August/014593.html
for more info.
Jouni Malinen [Sat, 29 Oct 2011 15:27:10 +0000 (18:27 +0300)]
mac80211_hwsim: Remove forgotten files
These live in the Linux kernel tree with mac80211_hwsim.c that was
already removed from hostap.git. The notes and example configuration
files are in Documentation/networking/mac80211_hwsim directory, so
remove the unnecessarily duplicated files from hostap.git.
Jouni Malinen [Sat, 29 Oct 2011 08:24:53 +0000 (11:24 +0300)]
Remove unconfigurable cts_protection_type
This was included in the AP configuration parameters, but there has
been no way to set it to anything else than the default automatic
mode. Remove this parameter and just hardcode ERP determination to
follow the automatic mode.
Evan Broder [Fri, 28 Oct 2011 21:23:54 +0000 (00:23 +0300)]
For MS-CHAP, convert the password from UTF-8 to UCS-2
The MS-CHAPv1 and MS-CHAPv2 RFCs specify that the password is a string
of "Unicode characters", which for Windows means UCS-2; thus the
password could be any even-length string of up to 512 bytes.
Instead of making the incompatible change of requiring the incoming
password to be UCS-2 encoded, assume the password is UTF-8 encoded and
convert it before using it in NtPasswordHash and
EncryptPwBlockWithPasswordHash.
Jithu Jance [Fri, 28 Oct 2011 19:26:30 +0000 (22:26 +0300)]
P2P: Fix frequency in the P2P_EVENT_GROUP_STARTED event
P2P Client did not show correct frequency in the control interface
event P2P_EVENT_GROUP_STARTED. Fix that by using the frequency from
the BSS table or association event.
Jouni Malinen [Fri, 28 Oct 2011 18:08:15 +0000 (21:08 +0300)]
nl80211: Ignore netlink interface down if interface is up
The netlink event processing is delayed since they are only returned
after control returns to eloop. This can result in netlink down events
being processed at a point when the interface has is actually still
there (the first event on new interface is down) and that can result in
odd behavior especially when the resulting interface-disabled event is
delivered to AP mode interface with wpa_supplicant.
Work around this by filtering netlink down events if the matching
interface is up at the time the netlink event is processed. This fixes
an issue brought up by commit 36d84860bbe09641f782fcc21b09e5a6952b4629.
Jouni Malinen [Fri, 28 Oct 2011 09:03:45 +0000 (12:03 +0300)]
Fix WPA authenticator configuration to not leave uninitialized fields
hostapd_wpa_auth_conf() is called on uninitialized memory and the
conditional blocks in this function may leave some fields into
uninitialized state. This can result in unexpected behavior elsewhere
since some of the variables may be used without matching #ifdef
blocks. Fix this by zeroing the memory.
Jouni Malinen [Thu, 27 Oct 2011 20:15:46 +0000 (23:15 +0300)]
TDLS: Do not clear peer data too early for teardown
TPK will be needed to be able to generate FTIE MIC for the
teardown message, so maintain peer data for a bit longer in
case the teardown FTIE request comes back from the driver.
Jithu Jance [Tue, 25 Oct 2011 22:28:51 +0000 (01:28 +0300)]
P2P: Stop any on-going "p2p_find" on creating an Autonomous GO
This patch stops any on-going "p2p_find" on creating an Autonomous GO.
This is already taken care in case of invoking a persistent GO. GO as
such will be advertising via the beacons for other devices to discover.
So normally the GO doesn't need to do a p2p_find until and unless it
wants to invite a P2P Client. In case of Invite scenario, logically it
is better to do a explicit discover [via user intervention] after the GO
is created. This patch will help to reduce battery wastage [due to
p2p_find operations] in scenarios where user creates a GO and doesn't
initiate a connection.
Jouni Malinen [Tue, 25 Oct 2011 21:06:06 +0000 (00:06 +0300)]
wlantest: Add STA counters for disconnect reason 6/7 RX
These can be useful in tests involving association state mismatch
between the AP and the STA (i.e., STA assumes it is still associated
but the AP does not have association state). In such a case, the AP
would be sending out unprotected Deauthentication or Disassociation
frames with reason code 6 or 7 depending on what frame is triggering
this.
Jay Katabathuni [Tue, 25 Oct 2011 08:31:21 +0000 (11:31 +0300)]
Interworking: Fix Advertisement Protocol element length limit
The Query Response Length Limit is not allowed to be zero when
this is sent by the AP. Use 0x7F to indicate that the limit is
based on maximum number of GAS fragments.
Jithu Jance [Mon, 24 Oct 2011 21:13:03 +0000 (00:13 +0300)]
P2P: Append P2P Device Address to AP-STA-CONNECTED event
For P2P, the p2p_connect takes in device address argument to make a
connection. However the connected event AP-STA-CONNECTED comes with
interface address. The application listening on events would find it
difficult to map interface address to the p2p device address which is
provided for connection.
Append P2P Device Address to AP-STA-CONNECTED event for P2P Client
connection. This will help applications to easily map the P2P Interface
Address to P2P Device Address on CONNECTED event. For non-P2P case, it
will just print the usual STA MAC address alone.
Jithu Jance [Mon, 24 Oct 2011 20:37:39 +0000 (23:37 +0300)]
P2P: Fix wpa_supplicant crash on P2P WPS PBC overlap case
Once PBC overlap detected when using dynamic group interfaces, the wpa_s
corresponding to P2P group interface is freed. This patch avoids
accessing the wpa_s data structure after it is freed. This fixes a
possible crash in P2P client role in such a case.
Jouni Malinen [Mon, 24 Oct 2011 14:29:37 +0000 (17:29 +0300)]
P2P: Advertise Persistent Reconnect group capability
The persistent_reconnect configuration parameter was used to decide
whether to accept invitation to re-establish a persistent group.
However, this was not being advertised in the Group Capability bitmap.
Add the Persistent Reconnect bit based on this configuration to GO
Negotiation frames and Beacon/Probe Response frames from the GO.
Arik Nemtsov [Sun, 23 Oct 2011 11:02:57 +0000 (14:02 +0300)]
TDLS: Add peer as a STA during link setup
Before commencing setup, add a new STA entry to the driver representing
the peer. Later during setup, update the STA entry using information
received from the peer.
Extend sta_add() callback for adding/modifying a TDLS peer entry and
connect it to the TDLS state machine. Implement this callback for the
nl80211 driver and send peer information to kernel.
Mark TDLS peer entries with a new flag and translate it to a
corresponding nl80211 flag in the nl80211 driver.
In addition, correct TDLS related documentation in the wpa_driver_ops
structure.
Signed-off-by: Arik Nemtsov <arik@wizery.com> Cc: Kalyan C Gaddam <chakkal@iit.edu>
TDLS: Collect peer capabilities and supp-rates during link setup
Record the capabilities and supported rates of the TDLS peer during
link setup. These are given in the IEs passed in Setup Request and
Setup Response frames.
Signed-off-by: Arik Nemtsov <arik@wizery.com> Cc: Kalyan C Gaddam <chakkal@iit.edu>
TDLS/nl80211: Support receiving TDLS discovery response frames
Register for the TDLS discovery response public action frame in nl80211.
Print out a debug message when a Discovery Resp frame is received and
validated.
Signed-off-by: Arik Nemtsov <arik@wizery.com> Cc: Kalyan C Gaddam <chakkal@iit.edu>
TDLS: Process discovery requests and send discovery responses
When a discovery request is received, add the peer to the TDLS peer
cache and send a response containing minimal data. Mandatory IEs in
the discovery response frame will be filled out by the driver.
Signed-off-by: Arik Nemtsov <arik@wizery.com> Cc: Kalyan C Gaddam <chakkal@iit.edu>
Jouni Malinen [Sun, 23 Oct 2011 17:09:40 +0000 (20:09 +0300)]
Remove incorrect statement about get_hw_feature_data() driver_ops
This function is actually needed for quite a bit more than just
user space MLME implementation. The old comment here has not been
accurate for a long time, so get rid of it.
Jouni Malinen [Sun, 23 Oct 2011 17:07:57 +0000 (20:07 +0300)]
Remove unused driver_ops client MLME functions
These driver_ops functions set_channel, set_ssid, set_bssid,
mlme_add_sta, and mlme_remove_sta were used with the user space
MLME implementation in wpa_supplicant. That was only for testing
purposes and was removed, but these driver_ops were forgotten.
Remove them now to remove confusing driver_ops definitions.
The hardware feature data is required in several different places
throughout the code. Previously, the data was acquired and freed on
demand, but with this patch wpa_supplicant will keep a single copy
around at runtime for everyone to use.
Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Jouni Malinen [Sun, 23 Oct 2011 10:20:52 +0000 (13:20 +0300)]
Postpone global_init() call until first driver instance is initialized
This avoids allocating global driver state for driver wrappers that
are built in but not used. This can save some resources and avoids
failures with driver_nl80211.c that is now initializing netlink
connections for nl80211 in global_init().
Jouni Malinen [Sat, 22 Oct 2011 19:45:38 +0000 (22:45 +0300)]
Remove user space client MLME
This code was used only with driver_test.c to allow MLME operations
in hostapd to be tested without having to use a real radio. There
are no plans on extending this to any other use than testing and
mac80211_hwsim has now obsoled the need for this type of testing.
As such, we can drop this code from wpa_supplicant to clean up the
implementation of unnecessary complexity.
Ben Greear [Thu, 20 Oct 2011 23:03:25 +0000 (02:03 +0300)]
nl80211: Use global netlink rtm event object
Netlink sockets can be shared among all driver instances, saving lots
of sockets, spurious log messages, memory, and CPU usage when using
multiple interfaces in a single process.