mech_eap.git
9 years agoInteworking: Add support to update the ANQP Capability List into the BSS
ASHUTOSH NARAYAN [Wed, 18 Feb 2015 02:35:10 +0000 (21:35 -0500)]
Inteworking: Add support to update the ANQP Capability List into the BSS

In addition, add support for returning the capability list through
the BSS control interface command.

Signed-off-by: ASHUTOSH NARAYAN <ashutoshx.narayan@intel.com>
9 years agoHS 2.0: Add support to update the HS20 Capability List into the BSS
ASHUTOSH NARAYAN [Wed, 18 Feb 2015 02:35:09 +0000 (21:35 -0500)]
HS 2.0: Add support to update the HS20 Capability List into the BSS

In addition, add support for returning the capability list through the
BSS control interface command.

Signed-off-by: ASHUTOSH NARAYAN <ashutoshx.narayan@intel.com>
9 years agowpa_gui: Debug enhancement
Arkadiusz (Arkq) Bokowy [Sat, 14 Feb 2015 16:32:02 +0000 (17:32 +0100)]
wpa_gui: Debug enhancement

Instead of calling a dummy printf function use preprocessor to determine
if debugging mode is enabled. Also use native Qt debug function.

Signed-off-by: Arkadiusz Bokowy <arkadiusz.bokowy@gmail.com>
9 years agoAdd forgotten network profile parameters to config file writing
Jouni Malinen [Sat, 21 Feb 2015 10:48:16 +0000 (12:48 +0200)]
Add forgotten network profile parameters to config file writing

Number of network profile parameters were not written to the
configuration file.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoIBSS: Add support for VHT80 configuration
Janusz Dziedzic [Thu, 19 Feb 2015 06:15:48 +0000 (07:15 +0100)]
IBSS: Add support for VHT80 configuration

Configure VHT80 based on driver capabilities.

Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
9 years agoMake hostapd_set_freq_params() common
Janusz Dziedzic [Thu, 19 Feb 2015 06:15:47 +0000 (07:15 +0100)]
Make hostapd_set_freq_params() common

Now this function can also be used from wpa_supplicant.

Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
9 years agoIBSS: Update operating frequency if joining an existing IBSS
Janusz Dziedzic [Thu, 19 Feb 2015 06:15:46 +0000 (07:15 +0100)]
IBSS: Update operating frequency if joining an existing IBSS

If a matching IBSS is found in scan results, change requested frequency
to match and disable OBSS scan.

Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
9 years agoIBSS: Add fixed_freq network parameter
Janusz Dziedzic [Thu, 19 Feb 2015 06:15:45 +0000 (07:15 +0100)]
IBSS: Add fixed_freq network parameter

Add fixed_freq=<0/1> network block parameter and pass it to the driver
when starting or joining an IBSS. If this flag is set, IBSS should not
try to look for other IBSS networks to merge with on different channels.

Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
9 years agoUse priority list instead of global for PNO
Dmitry Shmidt [Thu, 12 Feb 2015 18:25:15 +0000 (10:25 -0800)]
Use priority list instead of global for PNO

Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
9 years agotests: Add test to check disconnect in powersave
Johannes Berg [Fri, 9 Jan 2015 18:55:45 +0000 (19:55 +0100)]
tests: Add test to check disconnect in powersave

The kernel had two bugs (one in hwsim and one more important one in
mac80211) in this area, add a test to make sure we can disconnect
without any kernel issues while in powersave.

Also make sure that the TIM bit gets set and cleared again (by checking
with tshark.)

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
9 years agoAllow libnl-3.0 include path be specified
Romain Naour [Mon, 29 Dec 2014 20:54:02 +0000 (21:54 +0100)]
Allow libnl-3.0 include path be specified

The '/usr/include/libnl3' header path is unsafe for cross-compilation.
Use pkg-config to find libnl-3.0 headers by default and alternatively,
allow LIBNL_INC=<path> to be used in .config to override.

Signed-off-by: Romain Naour <romain.naour@openwide.fr>
9 years agoP2PS: Add P2PS interface info
Rajiv Ranjan [Tue, 13 Jan 2015 05:18:17 +0000 (10:48 +0530)]
P2PS: Add P2PS interface info

This adds documentation on P2PS related interface commands and events.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Re-association to same BSS to toggle PMF status
Jouni Malinen [Thu, 19 Feb 2015 14:37:12 +0000 (16:37 +0200)]
tests: Re-association to same BSS to toggle PMF status

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoAdd optional reassoc-to-same-BSS optimization
Jouni Malinen [Thu, 19 Feb 2015 14:35:39 +0000 (16:35 +0200)]
Add optional reassoc-to-same-BSS optimization

The new reassoc_same_bss_optim=1 configuration parameter can now be used
to request wpa_supplicant to bypass the unnecessary Authentication frame
exchange when reassociating back to the same BSS with which the device
is already associated. This functionality is disabled by default since
it may cause undesired interoperability issues with some APs.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoFix passive_scan config parameter writing
Jouni Malinen [Thu, 19 Feb 2015 14:22:47 +0000 (16:22 +0200)]
Fix passive_scan config parameter writing

Commit c35e35ed8133bb6d7b96765c5f0b1d41efe09fc5 ('Add passive_scan
configuration parameter') used incorrect parameter name when writing the
passive_scan parameter into a configuration file.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Re-sign expired test certificates
Jouni Malinen [Thu, 19 Feb 2015 12:18:57 +0000 (14:18 +0200)]
tests: Re-sign expired test certificates

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agohs20-osu-client: Ensure NULL checks are done before dereferencing
Subhani Shaik [Wed, 18 Feb 2015 00:06:35 +0000 (16:06 -0800)]
hs20-osu-client: Ensure NULL checks are done before dereferencing

In some error cases, pointers were dereferenced before NULL check is
done. Fix this by adding checks before the dereference.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoFix OCSP debug messages
Jouni Malinen [Thu, 19 Feb 2015 11:33:33 +0000 (13:33 +0200)]
Fix OCSP debug messages

These were not supposed to include a newline at the end of the message
text since such formatting gets handled by tls_show_errors(). In
addition, change the message about the issuer's issuer to be more
accurate.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoOpenSSL: Fix OCSP error path
Jouni Malinen [Thu, 19 Feb 2015 11:32:05 +0000 (13:32 +0200)]
OpenSSL: Fix OCSP error path

If addition of a peer issuer certificate fails, the certs pointer would
be NULL when being passed to sk_X509_push() for peer issuer's issuer.
Fix this by skipping addition of issuer's issue if issuer addition
fails.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoHTTP: Fix OCSP error path
Subhani Shaik [Tue, 17 Feb 2015 19:29:06 +0000 (11:29 -0800)]
HTTP: Fix OCSP error path

If addition of a peer issuer certificate fails, the certs pointer would
be NULL when being passed to sk_X509_push() for peer issuer's issuer.
Fix this by skipping addition of issuer's issue if issuer addition
fails.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoAndroid: Remove commented out non-Android build parameters
Jouni Malinen [Thu, 19 Feb 2015 10:55:19 +0000 (12:55 +0200)]
Android: Remove commented out non-Android build parameters

These hs20-osu-client parameters were never applicable for Android
builds and were just copied from the non-Android Makefile as a reminder,
but not removed once rest of the Android build was fixed.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoAndroid: Remove libxml2 config defines
Jouni Malinen [Thu, 19 Feb 2015 10:53:03 +0000 (12:53 +0200)]
Android: Remove libxml2 config defines

These need to be done in the libxml2 build, not in hs20-osu-client. This
workaround was previously used to allow parts of the build to go
through, but that was not a complete fix and resulted in warnings now
that external/libxml2 in Android 5.0 is defining the same parameters.
Remove these from hs20-osu-client Android.mk to avoid that warning.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoAndroid: Silence unused function parameter warnings
Jouni Malinen [Thu, 19 Feb 2015 10:48:51 +0000 (12:48 +0200)]
Android: Silence unused function parameter warnings

Numbers of hs20-osu-client functions do not use all of the parameters
currently. This makes the compiler output difficult to read due to
undesired warning messages. Get rid of those specific warnings for now.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoAndroid: Fix hs20-osu-client build on Android 5.0
Jouni Malinen [Thu, 19 Feb 2015 10:48:00 +0000 (12:48 +0200)]
Android: Fix hs20-osu-client build on Android 5.0

The LOCAL_EXPORT_C_INCLUDE_DIRS from ICU did not seem to fully resolve
the build (e.g., "mm -B" failed to build, but following that with "mm"
allowed the build to complete). For now, add the include directory
manually here for Android 5.0.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoHS 2.0R2: Fix permissions for SP/<fqdn> directory on Android
Neelansh Mittal [Mon, 16 Feb 2015 14:29:46 +0000 (19:59 +0530)]
HS 2.0R2: Fix permissions for SP/<fqdn> directory on Android

As part of OSU, the AAA TrustRoot cert is downloaded into SP/<fqdn>
directory. On Android, wpa_supplicant runs with Wifi uid privileges, and
hence might not have read access to the AAA TrustRoot present SP/<fqdn>
directory. Hence, make AID_WIFI as the group owner of SP/<fqdn>
directory and allow the members of AID_WIFI group to read files present
in this directory.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Make wpas_mesh_password_mismatch more robust
Jouni Malinen [Thu, 12 Feb 2015 13:50:52 +0000 (15:50 +0200)]
tests: Make wpas_mesh_password_mismatch more robust

It seems to be possible for dev2 (the one with incorrect password) to
stop retries before either dev0 or dev1 reports the authentication
failure event. For now, allow the test case pass if either dev0 or dev1
reports the event rather than requiring both to report this. The
expected behavior can be fine-tuned in the future if the reporting
behavior is modified to be more consistent.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agonl80211: Allocate QCA vendor subcmds for DFS radar detected and CAC events
Ahmad Kholaif [Wed, 11 Feb 2015 02:48:39 +0000 (18:48 -0800)]
nl80211: Allocate QCA vendor subcmds for DFS radar detected and CAC events

When DFS offloading capability is supported by the driver, the driver
should use these events to indicate when a radar pattern has been
detected, channel availability check (CAC) has been completed, aborted
or finished after the non-occupancy period is over on a DFS channel.

Also, add a new driver.h event to be used by NL80211 to indicate CAC
Started event on a DFS channel.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Fix ap_{open,wpa2_psk}_wpas_in_bridge cleanup
Jouni Malinen [Wed, 11 Feb 2015 13:07:05 +0000 (15:07 +0200)]
tests: Fix ap_{open,wpa2_psk}_wpas_in_bridge cleanup

Due to a copy-paste error, these test cases left 4addr mode enabled on
wlan5. This resulted in number of connect_cmd_* test cases failing if
executed after the wpas_in_bridge tests.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Remove temporary config file in wpas_config_file
Jouni Malinen [Wed, 11 Feb 2015 12:57:30 +0000 (14:57 +0200)]
tests: Remove temporary config file in wpas_config_file

The new wpa_supplicant configuration file writing style leaves behind
the temporary file (<filename>.tmp) if renaming fails. Clean that up in
the test case execution.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: hostapd and external ifconfig down + up
Jouni Malinen [Tue, 10 Feb 2015 16:20:54 +0000 (18:20 +0200)]
tests: hostapd and external ifconfig down + up

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoDrop all hostapd STA entries on interface disabled event
Jouni Malinen [Tue, 10 Feb 2015 20:20:07 +0000 (22:20 +0200)]
Drop all hostapd STA entries on interface disabled event

If the driver indicates that the interface has been disabled, assume
that all associations have been lost and remove the hostapd STA entries.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agonl80211: Indicate interface-down event only for the main netdev
Jouni Malinen [Wed, 11 Feb 2015 08:26:26 +0000 (10:26 +0200)]
nl80211: Indicate interface-down event only for the main netdev

RTM_NEWLINK event without IFF_UP were processed for all related
interfaces (including VLANs and bridge). While these events may need to
be processed for other purposes, they should not end up claiming that
the main interface has been disabled, so indicate
EVENT_INTERFACE_DISABLED only if the ifname matches the first BSS ifname
for the interface. In addition, fix some of the ifup/down checks from
if_indextoname() cases to actually use the resolved ifname (namebuf)
rather than hardcoding the first configured ifname to be used.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoVLAN: Clean up RTM_NEW/DELLINK processing
Jouni Malinen [Wed, 11 Feb 2015 08:29:52 +0000 (10:29 +0200)]
VLAN: Clean up RTM_NEW/DELLINK processing

This uses couple of additional helper macros and prints more debug
information to make the VLAN events easier to analyze.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agohostapd: Avoid sending client probe on removed client
Peter Oh [Mon, 9 Feb 2015 21:23:53 +0000 (13:23 -0800)]
hostapd: Avoid sending client probe on removed client

Sending client probe on already removed client from kernel driver does
not have any benefit and may lead unintended behavior among variable
drivers (mac80211 has a WARN_ON() that could have been triggered after
ifconfig down+up earlier when hostapd did not re-enable beaconing on
ifup). Skip this step in discussion when the kernel driver reports that
client entry is removed.

Signed-off-by: Peter Oh <poh@qca.qualcomm.com>
9 years agoRe-configure WPA2 group keys on hostapd interface re-enable
Jouni Malinen [Tue, 10 Feb 2015 16:19:51 +0000 (18:19 +0200)]
Re-configure WPA2 group keys on hostapd interface re-enable

This allows WPA2 mode AP to be re-enabled automatically after external
ifconfig down + up on a netdev used by hostapd.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoRe-enable beaconing on interface disable+enable
Jouni Malinen [Tue, 10 Feb 2015 16:13:04 +0000 (18:13 +0200)]
Re-enable beaconing on interface disable+enable

This is a step towards enabling hostapd to restart AP mode functionality
if the interface is disabled and re-enabled, e.g., with ifconfig down
and up. This commit takes care of beaconining only which may be
sufficient for open mode connection, but not for WPA2 cases.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agonl80211: Print a debug log entry on NL80211_CMD_PROBE_CLIENT failures
Jouni Malinen [Tue, 10 Feb 2015 15:31:27 +0000 (17:31 +0200)]
nl80211: Print a debug log entry on NL80211_CMD_PROBE_CLIENT failures

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Make autogo_join_auto_go_neg more robust
Jouni Malinen [Tue, 10 Feb 2015 10:09:07 +0000 (12:09 +0200)]
tests: Make autogo_join_auto_go_neg more robust

This test cases used to fail if dev1 had seen dev0 as a GO in an earlier
test case, e.g., when running it after autogo_fail. Fix this by clearing
scan results on dev1 at the beginning of the test case.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoP2P: Fix send_action_in_progress clearing in corner cases
Jouni Malinen [Mon, 9 Feb 2015 22:41:27 +0000 (00:41 +0200)]
P2P: Fix send_action_in_progress clearing in corner cases

It is possible for an Action frame TX operation to be stopped in a way
that results in the TX status callback function not being called. This
could happen, e.g., when P2P_STOP_FIND was issued while waiting for PD
Response TX status. This specific case ended in leaving
p2p->send_action_in_progress set to 1 and that ending up stopping a
future TX operation when p2p_send_action_cb() gets called with
p2p->pending_action_state == P2P_NO_PENDING_ACTION.

This could result in reception of a fragmented service discovery
response failing due to the GAS sequence getting stopped when receiving
TX callback for the first GAS comeback request. That sequence could be
hit in mac80211_hwsim tests when p2p_listen_and_offchannel_tx was
followed by p2p_service_discovery_fragmentation (even after a long time
since this was on dev1 and there could be even 10 minutes between these
test cases).

Fix this issue by clearing send_action_in_progress whenever stopping
pending P2P operation with p2p_stop_find (or P2P_FLUSH for that matter).

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: hostapd and start_disabled=1
Jouni Malinen [Mon, 9 Feb 2015 17:57:46 +0000 (19:57 +0200)]
tests: hostapd and start_disabled=1

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoAdd hostapd UPDATE_BEACON ctrl_iface command
Jouni Malinen [Mon, 9 Feb 2015 17:56:32 +0000 (19:56 +0200)]
Add hostapd UPDATE_BEACON ctrl_iface command

This can be used to ask Beacon frames to be updated explicitly, e.g.,
after a dynamic configuration parameter change. This can also be used to
start beaconing on an interface that was started with start_disabled=1.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agonl80211: Allocate QCA vendor subcmd for DFS CAC Start event
Ahmad Kholaif [Thu, 5 Feb 2015 21:13:48 +0000 (13:13 -0800)]
nl80211: Allocate QCA vendor subcmd for DFS CAC Start event

When DFS offloading capability is supported by the driver, the driver
should use this event to indicate when channel availability check (CAC)
is started on a DFS channel.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agonl80211: Test vendor command and event
Jouni Malinen [Mon, 9 Feb 2015 15:26:54 +0000 (17:26 +0200)]
nl80211: Test vendor command and event

This adds testing code (for CONFIG_TESTING_OPTIONS=y builds only) to
send an nl80211 vendor command and report a test vendor event in case
the driver supports this.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: wpa_supplicant DUMP/GET global parameter
Jouni Malinen [Sun, 8 Feb 2015 20:06:07 +0000 (22:06 +0200)]
tests: wpa_supplicant DUMP/GET global parameter

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoAdd control interface commands for fetching wpa_config values
Ola Olsson [Tue, 27 Jan 2015 23:22:06 +0000 (00:22 +0100)]
Add control interface commands for fetching wpa_config values

The new "DUMP" and "SET <variable>" control interface commands can be
used to fetch global wpa_supplicant configuration parameters.

Signed-off-by: Ola Olsson <ola.olsson@sonymobile.com>
9 years agotests: INTERWORKING_ADD_NETWORK
Jouni Malinen [Sun, 8 Feb 2015 15:09:18 +0000 (17:09 +0200)]
tests: INTERWORKING_ADD_NETWORK

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoAdd INTERWORKING_ADD_NETWORK command
Jouni Malinen [Sun, 8 Feb 2015 14:56:04 +0000 (16:56 +0200)]
Add INTERWORKING_ADD_NETWORK command

This can be used to provide more control to upper layers on network
blocks generated as part of Interworking network selection.
INTERWORKING_ADD_NETWORK behaves otherwise identically to
INTERWORKING_CONNECT, but it does not request a new connection after
having added the network block and it returns the network id of the
added network.

INTERWORKING_ADD_NETWORK followed by REASSOCIATE would behave more or
less identically to INTERWORKING_CONNECT, but this allows the created
network profile to be modified, if desired, and/or stored externally.
SELECT_NETWORK can also be used with the network id returned from
INTERWORKING_ADD_NETWORK to enforce that specific network profile to be
used for the next connection (though, it should be noted that this
behavior may not meet all Hotspot 2.0 requirements if there were other
enabled networks that could have higher priority).

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoAP: Do not reply to Probe Request frames with DS Params mismatch
Ilan Peer [Fri, 6 Feb 2015 01:37:04 +0000 (20:37 -0500)]
AP: Do not reply to Probe Request frames with DS Params mismatch

Do not reply to a Probe Request frame with a DSSS Parameter Set element
in which the channel is different than the operating channel of the AP,
as the sending station is not found on the AP's operating channel.

IEEE Std 802.11-2012 describes this as a requirement for an AP with
dot11RadioMeasurementActivated set to true, but strictly speaking does
not allow such ignoring of Probe Request frames if
dot11RadioMeasurementActivated is false. Anyway, this can help reduce
number of unnecessary Probe Response frames for cases where the STA is
less likely to see them (Probe Request frame sent on a neighboring, but
partially overlapping, channel).

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Successful P2P_CONNECT-auto cases
Jouni Malinen [Sun, 8 Feb 2015 10:41:14 +0000 (12:41 +0200)]
tests: Successful P2P_CONNECT-auto cases

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoP2P: Document P2P_CONNECT-auto
Jouni Malinen [Sun, 8 Feb 2015 10:24:37 +0000 (12:24 +0200)]
P2P: Document P2P_CONNECT-auto

Commit b31be3a0fd58b38e669d804aa082b1039408566f ('P2P: Add automatic GO
Negotiation vs. join-a-group selection') added this P2P_CONNECT 'auto'
parameter, but did not update any documentation on it.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agotests: P2P_CONNECT-auto not finding GO
Jouni Malinen [Sun, 8 Feb 2015 10:18:03 +0000 (12:18 +0200)]
tests: P2P_CONNECT-auto not finding GO

This is a regression test for an earlier bug that resulted in using
freed memory after a P2P group interface was removed as part of
fallback-to-GO-Negotiation in P2P_CONNECT-auto.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoAdd STOP_AP control interface command
Jouni Malinen [Sun, 8 Feb 2015 10:16:13 +0000 (12:16 +0200)]
Add STOP_AP control interface command

This is mainly for testing purposes to allow beaconing to be stopped
without clearing AP state in hostapd.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoP2P: Add event messages for P2P_CONNECT-fallback-to-GO-Neg
Jouni Malinen [Sun, 8 Feb 2015 10:04:15 +0000 (12:04 +0200)]
P2P: Add event messages for P2P_CONNECT-fallback-to-GO-Neg

This makes it easier for upper layer programs to follow progress of
P2P_CONNECT-auto operation.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoP2P: Fix P2P_CONNECT-auto fallback to GO Neg with group interface
Jouni Malinen [Sun, 8 Feb 2015 09:38:56 +0000 (11:38 +0200)]
P2P: Fix P2P_CONNECT-auto fallback to GO Neg with group interface

If a separate P2P group interface was used, P2P_CONNECT-auto fallback to
GO Negotiation could result in use of freed memory and segmentation
fault. This happened in cases where the peer GO was found in some old
scans, but not in the first scan triggered by the P2P_CONNECT-auto
command ("P2P: Peer was found running GO in older scan -> try to join
the group" shows up in the debug log). In addition, the GO would still
need to reply to PD Request to allow this code path to be triggered.

When five scans for the GO were completed in this sequence, the P2P
group interface was removed as part of falling back to GO Negotiation.
However, that ended up dereferencing the freed wpa_s instance at the end
of scan event processing. Fix this by reordering code a bit and breaking
out from EVENT_SCAN_RESULTS processing if the interface could have been
removed.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agotests: Make wpas_ctrl_many_networks more robust under valgrind
Jouni Malinen [Sat, 7 Feb 2015 20:38:21 +0000 (22:38 +0200)]
tests: Make wpas_ctrl_many_networks more robust under valgrind

It is possible for a low powered CPU to take excessively long time to
delete 1000 network blocks when running under valgrind. This would have
resulted in the test case failing and the following reset operation
timing out which would then stop the test sequence completely.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agotests: Mesh authentication failure events
Jouni Malinen [Sun, 8 Feb 2015 11:08:32 +0000 (13:08 +0200)]
tests: Mesh authentication failure events

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agomesh: Fix remaining BLOCKED state after SAE auth failure
Masashi Honma [Thu, 5 Feb 2015 14:00:01 +0000 (23:00 +0900)]
mesh: Fix remaining BLOCKED state after SAE auth failure

When SAE authentication fails, wpa_supplicant retries four times. If all
the retries result in failure, SAE state machine enters BLOCKED state.
Once it enters this state, wpa_supplicant doesn't retry connection. This
commit allow connection retries even if the state machine entered
BLOCKED state.

There could be an opinion "Is this patch needed? User could know the SAE
state machine is in the BLOCKED mode by MESH-SAE-AUTH-BLOCKED event.
Then user can retry connection. By user action, SAE state machine can
change the state from BLOCKED to another.". Yes, this is a true at the
joining mesh STA. However, a STA that is already a member of existing
mesh BSS should not retry connection because if the joining mesh STA
used wrong password, all the existing STA should do something from UI to
retry connection.

Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
9 years agomesh: Add a monitor event on SAE authentication getting blocked
Masashi Honma [Thu, 5 Feb 2015 14:00:00 +0000 (23:00 +0900)]
mesh: Add a monitor event on SAE authentication getting blocked

Send MESH-SAE-AUTH-BLOCKED event if SAE authentication is blocked. The
BLOCK state will finish when a new peer notification event is sent for
the same MAC address.

Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
9 years agomesh: Add a monitor event for SAE authentication failure
Masashi Honma [Thu, 5 Feb 2015 13:59:59 +0000 (22:59 +0900)]
mesh: Add a monitor event for SAE authentication failure

SAE authentication fails likely with wrong password. This commit adds a
notification of the failure to the upper application (UI) so that the
application can notify suspection of a wrong password to the user. The
control interface monitor even for this is "MESH-SAE-AUTH-FAILURE
addr=<peer>".

Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
9 years agomesh: Fix inactivity timer for 32 bit system
Masashi Honma [Wed, 4 Feb 2015 02:26:51 +0000 (11:26 +0900)]
mesh: Fix inactivity timer for 32 bit system

Commit 5a2a6de6a5fec58dcfdb4320e4ec2b69d183a4c1 ('mesh: Make inactivity
timer configurable') has a problem on 32 bit systems. Setting
NL80211_MESHCONF_PLINK_TIMEOUT to 0xffffffff causes expiration of STA in
a minute by NL80211_CMD_DEL_STATION event. this is the kernel rule for
STA expiration:

(current jiffies) > (frame Rx jiffies + NL80211_MESHCONF_PLINK_TIMEOUT * 250)

On a 32 bit system, the right side could overflow and be unexpected
small value if NL80211_MESHCONF_PLINK_TIMEOUT is sufficiently large. STA
expiration occurs by this reason.

This patch solves the problem by disabling the STA expiration
functionality in mac80211. However, old kernel does not support
disabling it. If so, this patch sets mac80211 inactivity timer 60
seconds into future from the wpa_supplicant inactivity timer.

And I mis-understood that mesh_max_inactivity=0 disables inactivity
timer in wpa_supplicant. This commit fixes it also.

Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
9 years agotests: Remove forgottend debug print from p2ps_connect_adv_go_persistent
Jouni Malinen [Sat, 7 Feb 2015 14:29:22 +0000 (16:29 +0200)]
tests: Remove forgottend debug print from p2ps_connect_adv_go_persistent

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agomesh: Remove duplicated no_auto_peer update
Jouni Malinen [Sat, 7 Feb 2015 14:19:53 +0000 (16:19 +0200)]
mesh: Remove duplicated no_auto_peer update

Commit 07cb45ccb2baa6e9627e3d21a285b95d744a6c77 ('mesh: Add no_auto_peer
config option') added a new struct wpa_ssid argument and added an
unnecessary parsing and setting of the value in
wpa_supplicant_ctrl_iface_update_network(). This is not needed since
wpa_config_set() takes care of parsing the parameters.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoD-Bus: Fix network block type change
Jouni Malinen [Sat, 7 Feb 2015 14:17:11 +0000 (16:17 +0200)]
D-Bus: Fix network block type change

It is possible for a network profile to change its type from P2P
persistent group to a normal network and back. The D-Bus interface uses
different types of objects for those, so the object needs to
re-registered in case of type change. This fixes issues in leaving
behind an incorrect type of object and leaking memory when freeing such
a network block that has had its disabled parameter changed.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agotests: Remove forgotten debug print from grpform_pbc_overlap
Jouni Malinen [Sat, 7 Feb 2015 13:38:29 +0000 (15:38 +0200)]
tests: Remove forgotten debug print from grpform_pbc_overlap

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoFix HT40 co-ex scanning issue on hostapd error path
Jouni Malinen [Sat, 7 Feb 2015 11:02:19 +0000 (13:02 +0200)]
Fix HT40 co-ex scanning issue on hostapd error path

If HT40 co-ex scan fails due to the driver rejecting scan triggers
multiple times, it was possible for the ap_ht40_scan_retry() timeout
being left behind and it getting run after hapd->drv_priv has been
cleared. This would result in NULL pointer dereference in
driver_nl80211_scan.c. Fix this by canceling the timeout when disabling
the interface.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agoFix Linux packat socket regression work around
Jouni Malinen [Fri, 6 Feb 2015 23:13:34 +0000 (01:13 +0200)]
Fix Linux packat socket regression work around

Commit e6dd8196e5daf39e4204ef8ecd26dd50fdca6040 ('Work around Linux
packet socket regression') added a mechanism to close the workaround
bridge socket in l2_packet_receive(). However, it did not take into
account the possibility of the l2->rx_callback() closing the l2_packet
socket altogether. This could result in use of freed memory when usin
RSN pre-authentication. Fix this by reordering the calls to clear the
workaround socket before calling the rx_callback.

Signed-off-by: Jouni Malinen <j@w1.fi>
9 years agotests: Use the global control interface in test_nfc_p2p.py
Ilan Peer [Fri, 6 Feb 2015 01:31:01 +0000 (20:31 -0500)]
tests: Use the global control interface in test_nfc_p2p.py

Signed-off-by: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Modify autogo_bridge test to use group interface
David Spinadel [Fri, 6 Feb 2015 01:31:00 +0000 (20:31 -0500)]
tests: Modify autogo_bridge test to use group interface

1. Add get_group_ifname() to wpasupplicant.py
2. Use the function to get the interface name for the bridge.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
9 years agotests: Modify test_p2p_discovery to use global and group interfaces
Ilan Peer [Fri, 6 Feb 2015 01:30:59 +0000 (20:30 -0500)]
tests: Modify test_p2p_discovery to use global and group interfaces

Modify the tests in test_p2p_discovery to use the global
control interface and group interface.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Use the group control interface in p2p_channel_avoid
Ilan Peer [Fri, 6 Feb 2015 01:30:58 +0000 (20:30 -0500)]
tests: Use the group control interface in p2p_channel_avoid

Use the group control interface to wait for P2P-REMOVE-AND-REFORM-GROUP
as the event is a group event.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Use group control interface in autogo
Ilan Peer [Fri, 6 Feb 2015 01:30:57 +0000 (20:30 -0500)]
tests: Use group control interface in autogo

P2P-PRESENCE-RESPONSE is a group event, and thus need
to use the group control interface.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Use global control interface in wpas_ctrl_global
Ilan Peer [Fri, 6 Feb 2015 01:30:56 +0000 (20:30 -0500)]
tests: Use global control interface in wpas_ctrl_global

Use the global control interface to set/get the status of the
P2P Device.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Remove unnecessary use of sudo from test cases
Jouni Malinen [Fri, 6 Feb 2015 22:22:32 +0000 (00:22 +0200)]
tests: Remove unnecessary use of sudo from test cases

run-tests.py is running as root, so sudo does not need to be used
anymore from within each test case.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Replace subprocess.call for rm with os.remove()
Jouni Malinen [Fri, 6 Feb 2015 22:15:01 +0000 (00:15 +0200)]
tests: Replace subprocess.call for rm with os.remove()

There is no need to use sudo and external rm to remove files now that
run-tests.py is required to run as root.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Fix wpas_config_file after implementation change
Jouni Malinen [Fri, 6 Feb 2015 22:10:04 +0000 (00:10 +0200)]
tests: Fix wpas_config_file after implementation change

The new wpa_supplicant configuration writing design (rename instead of
write to original file) did not fail with the symlink-to-self case, so
replace this with the config file being replaced with a directory. In
addition, get rid of unnecessary use of subprocess since run-tests.py is
running as root nowadays.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoDon't write to wpa_supplicant.conf directly
Vinit Deshpande [Thu, 5 Feb 2015 20:48:02 +0000 (12:48 -0800)]
Don't write to wpa_supplicant.conf directly

There is a chance that wpa_supplicant may get killed during
the time it is writing config file. If this happens, user
information like SSIDs and passwords can be lost forever.

This change works around that by writing config to a
temporary file and then renaming the file to the correct name.

Signed-off-by: Vinit Deshpande <vinitd@google.com>
9 years agotests: IBSS with WEP and TKIP
Jouni Malinen [Fri, 6 Feb 2015 19:43:50 +0000 (21:43 +0200)]
tests: IBSS with WEP and TKIP

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoIBSS: Do not enable HT with WEP or TKIP
Janusz Dziedzic [Fri, 6 Feb 2015 12:11:24 +0000 (13:11 +0100)]
IBSS: Do not enable HT with WEP or TKIP

We should not enable HT if WEP or TKIP is configured.
Without the patch and WEP configuration we will get message:
Association request to the driver failed

Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
9 years agoACS: Accept channel if any (rather than all) survey results are valid
Jouni Malinen [Fri, 6 Feb 2015 16:36:13 +0000 (18:36 +0200)]
ACS: Accept channel if any (rather than all) survey results are valid

Previously, a channel with even a single scan/survey result missing
information was skipped in ACS. This may not be desirable in cases when
multiple scan iterations are used (which is the case by default in
hostapd). Instead, use all channels that provided at least one complete
set of results. Calculate the average interference factor as an average
of the iterations that did provide complete values.

This seems to help with some cases, e.g., when ath9k may not be able to
report the noise floor for all channels from the first scan iteration
immediately after the driver has been loaded, but then returns it for
all other scan iterations.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: ACS with acs_chan_bias parameter
Jouni Malinen [Fri, 6 Feb 2015 16:02:33 +0000 (18:02 +0200)]
tests: ACS with acs_chan_bias parameter

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoACS: Allow specific channels to be preferred
Jouni Malinen [Fri, 6 Feb 2015 15:59:57 +0000 (17:59 +0200)]
ACS: Allow specific channels to be preferred

The new acs_chan_bias configuration parameter is a space-separated list
of <channel>:<bias> pairs. It can be used to increase (or decrease) the
likelihood of a specific channel to be selected by the ACS algorithm.
The total interference factor for each channel gets multiplied by the
specified bias value before finding the channel with the lowest value.
In other words, values between 0.0 and 1.0 can be used to make a channel
more likely to be picked while values larger than 1.0 make the specified
channel less likely to be picked. This can be used, e.g., to prefer the
commonly used 2.4 GHz band channels 1, 6, and 11 (which is the default
behavior on 2.4 GHz band if no acs_chan_bias parameter is specified).

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agoACS: Use weighted average for 2.4 GHz adjacent channel interference
Jouni Malinen [Thu, 5 Feb 2015 15:57:06 +0000 (17:57 +0200)]
ACS: Use weighted average for 2.4 GHz adjacent channel interference

The interference factors for adjacent 2.4 GHz channels were summed
together without doing any kind of weighted average on them. This
resulted in the channels at the band edges getting undue preference due
to only including interference factors from three channels vs. five for
the channels in the middle of the band.

While it is somewhat unclear whether the design here was supposed to
count overlapping channels together in this way or whether that is
already covered in channel survey results, it is clear that this summing
of three to five values together and then comparing the sum rather than
average of some kind gives too much preference to the channels at the
edges of the band by assuming that there is no interference whatsoever
outside the band.

Use weighted average of the interference factors rather than a sum from
different number of values. For now, the adjacent 2.4 GHz channels get
weight of 0.85 (1.0 for the main channel itself) and the neighboring
channels to those adjacent ones get 0.55 weight. Band-edge channels are
handled in a way that takes average over the channels that were actually
considered instead of assuming zero interference from neighboring bands.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
9 years agotests: Use global control interface in test_p2p_concurrency
Ilan Peer [Wed, 4 Feb 2015 09:30:36 +0000 (04:30 -0500)]
tests: Use global control interface in test_p2p_concurrency

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Use global control interface in test_p2p_service
Ilan Peer [Wed, 4 Feb 2015 09:30:35 +0000 (04:30 -0500)]
tests: Use global control interface in test_p2p_service

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Modify test_p2p_discovery to use global control interface
Ilan Peer [Wed, 4 Feb 2015 09:30:34 +0000 (04:30 -0500)]
tests: Modify test_p2p_discovery to use global control interface

1. Modify discovery_stop to use global control interface when calling
   P2P_FLUSH.
2. Modify p2p_listen_and_offchannel_tx to use the global control
   interface when waiting for P2P PD event.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Add printk tracer to trace-cmd
Eliad Peller [Wed, 4 Feb 2015 09:30:33 +0000 (04:30 -0500)]
tests: Add printk tracer to trace-cmd

Signed-off-by: Eliad Peller <eliadx.peller@intel.com>
9 years agotests: Modify tests in test_ap_pmf to use own_addr()
Ben [Wed, 4 Feb 2015 09:30:32 +0000 (04:30 -0500)]
tests: Modify tests in test_ap_pmf to use own_addr()

Some of the tests used p2p_dev_addr() that can be different from
own_addr() if a dedicated P2P Device interface is used.

Signed-off-by: Ben Rosenfeld <ben.rosenfeld@intel.com>
9 years agotests: Modify tests in test_hapd_ctrl to use own_addr()
Ben [Wed, 4 Feb 2015 09:30:31 +0000 (04:30 -0500)]
tests: Modify tests in test_hapd_ctrl to use own_addr()

Some of the tests used p2p_dev_addr() that can be different from
own_addr() if a dedicated P2P Device interface is used.

Signed-off-by: Ben Rosenfeld <ben.rosenfeld@intel.com>
9 years agotests: Fix ap_wps_per_station_psk to use own_addr()
Ben [Wed, 4 Feb 2015 09:30:30 +0000 (04:30 -0500)]
tests: Fix ap_wps_per_station_psk to use own_addr()

The test used p2p_dev_addr() that can be different from own_addr()
if a dedicated P2P Device interface is used.

Signed-off-by: Ben Rosenfeld <ben.rosenfeld@intel.com>
9 years agotests: Fix ap_wpa2_psk_file to use own_addr
Ben [Wed, 4 Feb 2015 09:30:27 +0000 (04:30 -0500)]
tests: Fix ap_wpa2_psk_file to use own_addr

The test used p2p_dev_addr() that can be different from own_addr()
if a dedicated P2P Device interface is used.

Signed-off-by: Ben Rosenfeld <ben.rosenfeld@intel.com>
9 years agotests: Fix destination address in ap_hs20_remediation_required_ctrl
Matti Gottlieb [Wed, 4 Feb 2015 09:30:26 +0000 (04:30 -0500)]
tests: Fix destination address in ap_hs20_remediation_required_ctrl

Fix the destination address that is sent in the WNM-Notification to be
the BSS address opposed to the P2P address.

Signed-off-by: Matti Gottlieb <matti.gottlieb@intel.com>
9 years agotests: Use get_bss() with ifname in autogo
Ilan Peer [Wed, 4 Feb 2015 09:30:25 +0000 (04:30 -0500)]
tests: Use get_bss() with ifname in autogo

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Add option to specify ifname to get_bss()
Ilan Peer [Wed, 4 Feb 2015 09:30:24 +0000 (04:30 -0500)]
tests: Add option to specify ifname to get_bss()

This is needed for cases that the group interface differs from the main
interface, i.e., when a dedicated P2P Device interface is used.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Use the global control interface in test_p2p_grpform
Ilan Peer [Wed, 4 Feb 2015 09:30:23 +0000 (04:30 -0500)]
tests: Use the global control interface in test_p2p_grpform

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Use global_request for SET commands in test_p2p_grpform
Ilan Peer [Wed, 4 Feb 2015 09:30:22 +0000 (04:30 -0500)]
tests: Use global_request for SET commands in test_p2p_grpform

When the 'SET' command is used to configure parameters related to P2P
operation use the global control interface and not the per interface one
as otherwise the setting will only have effect on the interface and will
work if a dedicated P2P_DEVICE is used.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Update group ifname in p2p_go_invite_auth
Ilan Peer [Wed, 4 Feb 2015 09:30:21 +0000 (04:30 -0500)]
tests: Update group ifname in p2p_go_invite_auth

Once the connection is established need to call group_form_result() on
the invited device, as otherwise the group interface name is not updated
and the connectivity test is done with the main interface instead of the
group interface.

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Use global control interface in discovery_dev_id
Ilan Peer [Wed, 4 Feb 2015 09:30:20 +0000 (04:30 -0500)]
tests: Use global control interface in discovery_dev_id

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Modify autogo_legacy to use the group interface
Ilan Peer [Wed, 4 Feb 2015 09:30:17 +0000 (04:30 -0500)]
tests: Modify autogo_legacy to use the group interface

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Use global control interface in wait_go_ending_session()
Ilan Peer [Wed, 4 Feb 2015 09:30:13 +0000 (04:30 -0500)]
tests: Use global control interface in wait_go_ending_session()

P2P_EVENT_GROUP_REMOVED is a global event, so use
wait_global_event() instead of wait_event().

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
9 years agotests: Verify SD TX callback processing after P2P find is stopped
Jouni Malinen [Wed, 4 Feb 2015 18:47:14 +0000 (20:47 +0200)]
tests: Verify SD TX callback processing after P2P find is stopped

These are far from perfect since timing is quite difficult to match for
the case that behaved incorrectly. Anyway, it looks loke
p2p_service_discovery_peer_not_listening was able to hit the error now
and then, so this should be sufficient as a regression test case.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>