mech_eap.git
8 years agoRADIUS: Update full message for interim accounting updates
Jouni Malinen [Mon, 29 Feb 2016 09:44:43 +0000 (11:44 +0200)]
RADIUS: Update full message for interim accounting updates

Instead of using the RADIUS client retransmission design with the old
RADIUS message contents for each retry, trigger a completely new interim
accounting update instance more quickly (using the same schedule as
RADIUS message retransmissions) to improve accounting updates in cases
where RADIUS message delivery fails. This allows the server to get up to
date information from the time the "retry" message was sent instead of
the old information from the time the first failed attempt was sent.

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoDocument nas_identifier requirements for RADIUS accounting
Jouni Malinen [Sun, 28 Feb 2016 19:19:14 +0000 (21:19 +0200)]
Document nas_identifier requirements for RADIUS accounting

nas_identifier needs to be set to a unique value for RADIUS accounting
to work properly. This needs to be unique for each BSS.

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoReplace hostapd_mac_comp_empty() with is_zero_ether_addr()
Jouni Malinen [Sun, 28 Feb 2016 16:00:48 +0000 (18:00 +0200)]
Replace hostapd_mac_comp_empty() with is_zero_ether_addr()

There is no need to maintain two implementations of the functionality.
is_zero_ether_addr() is easier to understand, so use it.

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoVLAN: Avoid use of libnl cache
Michael Braun [Wed, 24 Feb 2016 11:53:50 +0000 (12:53 +0100)]
VLAN: Avoid use of libnl cache

Using rtnl_link_alloc_cache() is expensive as it fills in all configured
links. Using rtnl_link_get_kernel() is much more lightweight.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agonl80211: Clean up ifidx properly if interface in a bridge is removed
Michael Braun [Wed, 24 Feb 2016 11:53:44 +0000 (12:53 +0100)]
nl80211: Clean up ifidx properly if interface in a bridge is removed

When an interface is removed while it was in a bridge, the bridge was
not removed from the local ifidx list.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agoFT: Check destination MAC address on RRB receive
Michael Braun [Wed, 24 Feb 2016 11:53:41 +0000 (12:53 +0100)]
FT: Check destination MAC address on RRB receive

As the Linux variant of l2_packet_init() does not use its own_addr
argument and l2_packet_receive() does not filter on destination MAC
address, this needs to be checked in the callback.

If there are multiple BSSes listening for FT RRB packets, all their
BSSIDs need to be local to the bridge interface. As l2_packet_init() is
going to receive all of them going for any local address, those RRB
messages started turning up on BSSes that were not destinated for and
cluttering logs.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agoRADIUS: Allow RADIUS server to provide PSK instead of passphrase
Michael Braun [Wed, 24 Feb 2016 11:53:38 +0000 (12:53 +0100)]
RADIUS: Allow RADIUS server to provide PSK instead of passphrase

If the AP is slow, passphrase hashing takes too long to serve the client
before timeout. Extend the Tunnel-Password design to allow a 64
character value to be interpreted as a PSK and send SSID to RADIUS
server. This allows the RADIUS server to either take care of passphrase
hashing or to use raw PSK without such hashing.

This is especially important for FT-PSK with FT-over-air, where hashing
cannot be deferred.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agoCache hashed passphrase in RADIUS-based PSK delivery
Michael Braun [Wed, 24 Feb 2016 11:53:37 +0000 (12:53 +0100)]
Cache hashed passphrase in RADIUS-based PSK delivery

Instead of copying the full struct hostapd_sta_wpa_psk_short, share the
existing entry and use reference counting to check when it needs to be
freed. This allows caching of PSKs derived from passphrases to avoid
having to perform the heavy hashing operation multiple times.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agoDefer passphrase-to-PSK hashing out of 802.11 authentication ACL check
Michael Braun [Wed, 24 Feb 2016 11:53:36 +0000 (12:53 +0100)]
Defer passphrase-to-PSK hashing out of 802.11 authentication ACL check

Hashing takes quite some time (can be about one second on a low-power
CPU for each passphrase provided), so hostapd can easily hit the 900 ms
Wi-Fi client authentication deadline (mac80211 uses 3x 300 ms). This can
be fixed by storing the passphrase instead of PSK with the STA and defer
the hashing into the WPA/RSN 4-way handshake, when enumerating all PSKs.

This applies for the case where a RADIUS server is used to store the
per-STA passphrases and this passphrase is delivered as part of the MAC
ACL check during IEEE 802.11 Authentication frame processing.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agoVLAN: Use stack instead of heap allocation for new interface name
Michael Braun [Wed, 24 Feb 2016 11:53:35 +0000 (12:53 +0100)]
VLAN: Use stack instead of heap allocation for new interface name

The VLAN ifname is limited to the maximum length of IFNAMSIZ, so there
is no need to use heap allocation for it.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agoFT: Use BSSID as r1_key_holder if no value is configured
Michael Braun [Wed, 24 Feb 2016 11:53:34 +0000 (12:53 +0100)]
FT: Use BSSID as r1_key_holder if no value is configured

r1_key_holder is an identifier that was always set to zero if unless
configured before.

See 11.6.1.7.4 of IEEE Std 802.11-2012 which reads
 "R1KH-ID is a MAC address of the holder of the PMK-R1 in the
  Authenticator of the AP"
See 12.2.2 of IEEE Std 802.11-2012 which reads
 "Each R0KH-ID and R1KH-ID is assumed to be expressed as a unique
  identifier within the mobility domain."
 "The R1KH-ID shall be set to a MAC address of the physical entity
  that stores the PMK-R1 ..."

Defaulting this to BSSID is a more reasonable value since we have not
rejected the missing r1_key_holder as invalid configuration.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agotests: FT RRB internal delivery to non-WPA BSS
Michael Braun [Wed, 24 Feb 2016 11:53:24 +0000 (12:53 +0100)]
tests: FT RRB internal delivery to non-WPA BSS

A malicious station could try to do FT-over-DS with a non WPA-enabled
BSS. When this BSS is located in the same hostapd instance, internal RRB
delivery will be used and thus the FT Action Frame will be processed by
a non-WPA enabled BSS. This processing used to crash hostapd as
hapd->wpa_auth is NULL.

This test implements such a malicious request for regression testing.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agoFT: Check hapd->wpa_auth before RRB internal delivery
Michael Braun [Wed, 24 Feb 2016 11:53:25 +0000 (12:53 +0100)]
FT: Check hapd->wpa_auth before RRB internal delivery

A malicious station could try to do FT-over-DS with a non WPA-enabled
BSS. When this BSS is located in the same hostapd instance, internal RRB
delivery will be used and thus the FT Action Frame will be processed by
a non-WPA enabled BSS. This processing used to crash hostapd as
hapd->wpa_auth is NULL. If the target BSS is on a different hostapd
instance, it will not listen for these packets and thus not crash.

Fix this by checking hapd->wpa_auth before delivery.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agoFT: Fix R0KH-R1KH protocol data length values
Michael Braun [Wed, 24 Feb 2016 11:53:11 +0000 (12:53 +0100)]
FT: Fix R0KH-R1KH protocol data length values

The FT RRB hostapd packets have a length field. For PULL frames, it
counted the bytes starting with nonce and up to the last before pad. For
RESP frames, it counted the bytes starting with nonce and up to the last
before pad except for 2 bytes. For PUSH frames, it counted the bytes
starting with nonce and up to including pad.

As rounding is done with AES encryption, including pad does not make
sense. Not including the last field before pad does not make sense
either. These were broken in the earlier addition of the 2 octet
pairwise field in commit 1b484d60e55a01d02bf01f790a7c6fb9ea606098 ('FT:
Include pairwise cipher suite in PMK-R0 SA and PMK-R1 SA').

AES encryption is not affected, as rounding hides the differences. The
packets data_length field is not used, so the differences have no effect
there.

This patch changes the constants to match the bytes used, thus excluding
the pad. To validate the changes, look at remainder modulo 8 of the sum
of the size constants and the padding sizes.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agotests: RSN pre-auth and PMKSA caching with per_sta_vif
Michael Braun [Wed, 24 Feb 2016 11:53:10 +0000 (12:53 +0100)]
tests: RSN pre-auth and PMKSA caching with per_sta_vif

This adds new tests to verify pmksa_cache_preauth when
used with per_sta_vif and possibly vlans.

While at it it refactors the code such that the tests
  pmksa_cache_preauth
  pmksa_cache_preauth_vlan_enabled
  pmksa_cache_preauth_vlan_used
  pmksa_cache_preauth_per_sta_vif
  pmksa_cache_preauth_vlan_enabled_per_sta_vif
  pmksa_cache_preauth_vlan_used_per_sta_vif
share code where possible.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agotests: Document tshark and python-netifaces as dependencies
Michael Braun [Wed, 24 Feb 2016 11:53:09 +0000 (12:53 +0100)]
tests: Document tshark and python-netifaces as dependencies

tshark and python-netifaces are used by some tests.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agotests: Verify correct VLAN operation after reconnect
Michael Braun [Wed, 24 Feb 2016 11:53:08 +0000 (12:53 +0100)]
tests: Verify correct VLAN operation after reconnect

This adds a test case ap_vlan_reconnect. It connects, disconnects, and
reconnects a station in a VLAN. This tests for a regression with
wpa_group entering the FATAL_FAILURE state as the AP_VLAN interface is
removed before the group was stopped.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agotests: Tagged-VLAN only change on reauthentication
Michael Braun [Wed, 24 Feb 2016 11:53:07 +0000 (12:53 +0100)]
tests: Tagged-VLAN only change on reauthentication

Check VLAN ID change during reauthentication when using tagged-only
configuration.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agotests: P2PS connection with cfg80211 P2P Device
Jouni Malinen [Sat, 27 Feb 2016 17:44:11 +0000 (19:44 +0200)]
tests: P2PS connection with cfg80211 P2P Device

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agotests: p2p_device_nfc_invite with no separate group interface
Jouni Malinen [Sat, 27 Feb 2016 17:36:30 +0000 (19:36 +0200)]
tests: p2p_device_nfc_invite with no separate group interface

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agotests: P2P persistent group with cfg80211 P2P Device
Jouni Malinen [Fri, 26 Feb 2016 20:21:15 +0000 (22:21 +0200)]
tests: P2P persistent group with cfg80211 P2P Device

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agotests: P2P join-group using cfg80211 P2P Device
Jouni Malinen [Fri, 26 Feb 2016 16:40:19 +0000 (18:40 +0200)]
tests: P2P join-group using cfg80211 P2P Device

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agotests: P2P autogo using cfg80211 P2P Device
Jouni Malinen [Fri, 26 Feb 2016 16:36:44 +0000 (18:36 +0200)]
tests: P2P autogo using cfg80211 P2P Device

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agotests: P2P group formation with driver using cfg80211 P2P Device
Jouni Malinen [Fri, 26 Feb 2016 16:21:51 +0000 (18:21 +0200)]
tests: P2P group formation with driver using cfg80211 P2P Device

This extends the P2P Device testing coverage to include the newly
enabled option of using the primary interface (e.g., wlan0) for P2P
group operation instead of always forcing a separate group interface to
be created dynamically.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoP2P: Support dedicated P2P_DEVICE without separate group interface
Lior David [Sun, 24 Jan 2016 15:36:49 +0000 (17:36 +0200)]
P2P: Support dedicated P2P_DEVICE without separate group interface

Add support for drivers with dedicated P2P_DEVICE interface, but without
group interface concurrency (only a single netdev is used). With such
devices, wpa_supplicant tried to use the p2p_dev interface instead of
the group interface and most P2P operations failed. Extend
wpa_supplicant to use the primary interface instead of a separate group
interface in such cases.

Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
8 years agoP2P: Add a separate pointer to the P2P Device instance
Lior David [Sun, 24 Jan 2016 15:36:49 +0000 (17:36 +0200)]
P2P: Add a separate pointer to the P2P Device instance

In many places in the code there was a reference to wpa_s->parent to get
from group interface to p2p_dev interface. These places can break if
P2P_DEVICE interface would need to be used with the primary interface as
the group interface, since the parent of the primary interface points to
itself and not the p2p_dev interface.

Fix this by adding a separate "p2pdev" pointer to wpa_supplicant,
it will be the same as parent pointer in most cases but whenever
the primary interface is used as a group interface, change it to
point to the correct p2p_dev interface.

Signed-off-by: Lior David <qca_liord@qca.qualcomm.com>
8 years agoGAS client: Make PMF check on RX more consistent
Jouni Malinen [Sat, 27 Feb 2016 16:49:31 +0000 (18:49 +0200)]
GAS client: Make PMF check on RX more consistent

Use the SA field instead of BSSID in the received Action frame to
determine whether PMF has been negotiated with the transmitter. While
these fields are supposed to be same for Public Action frames from an
AP, it would be possible that a frame is received with different values.
The following operations in gas_query_rx() use SA, so do the same for
the PMF check.

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agotests: GAS query with another AP while associated and using PMF
Jouni Malinen [Sat, 27 Feb 2016 09:47:58 +0000 (11:47 +0200)]
tests: GAS query with another AP while associated and using PMF

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agotests: Multi-icon fetch with REQ_HS20_ICON
Jouni Malinen [Sat, 27 Feb 2016 09:14:02 +0000 (11:14 +0200)]
tests: Multi-icon fetch with REQ_HS20_ICON

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agotests: WNM BSS Transition Management and scan behavior
Jouni Malinen [Fri, 26 Feb 2016 15:18:39 +0000 (17:18 +0200)]
tests: WNM BSS Transition Management and scan behavior

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoWNM: Optimize a single BSS transition management candidate scan
Jouni Malinen [Fri, 26 Feb 2016 15:16:35 +0000 (17:16 +0200)]
WNM: Optimize a single BSS transition management candidate scan

If the BSS Transition Management Request frame includes only a single
candidate and we need to scan for the BSS to get up-to-date information,
use a scan for the known BSSID instead of wildcard BSSID. In addition,
set the SSID in the scan if it is known based on old scan results in the
BSS table. This removes unnecessary Probe Response frames when we are
interested in results from only a single BSS.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agonl80211: Add an option to specify the BSSID to scan for
Jouni Malinen [Fri, 26 Feb 2016 14:31:33 +0000 (16:31 +0200)]
nl80211: Add an option to specify the BSSID to scan for

This allows scans to be optimized when a response is needed only from a
single, known BSS.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agotests: MBO STA supported operating classes
Jouni Malinen [Wed, 24 Feb 2016 10:31:09 +0000 (12:31 +0200)]
tests: MBO STA supported operating classes

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoAP: Store STA supported operating classes information
Jouni Malinen [Wed, 24 Feb 2016 10:20:31 +0000 (12:20 +0200)]
AP: Store STA supported operating classes information

This makes hostapd track Supported Operating Classes information from
the associated STAs. The stored information is available through the STA
control interface command (supp_op_classes row) as a hexdump of the
Supported Operating Classes element starting from the Length field. This
information can be used as input to BSS transition management and
channel switching decisions.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoAP: Debug print management frame TX result
Dedy Lansky [Tue, 9 Feb 2016 07:16:56 +0000 (09:16 +0200)]
AP: Debug print management frame TX result

Inside management frame TX status callback, print the TX result where it
was missing. This is useful for debugging management frame drops.

Signed-off-by: Dedy Lansky <qca_dlansky@qca.qualcomm.com>
8 years agotests: MBO non-preferred channel list
Jouni Malinen [Mon, 22 Feb 2016 17:51:36 +0000 (19:51 +0200)]
tests: MBO non-preferred channel list

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoMBO: Parse non-preferred channel list on the AP
Jouni Malinen [Mon, 22 Feb 2016 18:37:21 +0000 (20:37 +0200)]
MBO: Parse non-preferred channel list on the AP

This adds parsing of non-preferred channel list on an MBO AP. The
information in (Re)Association Request and WNM Notification Request
frames is parsed to get the initial value and updates from each
associated MBO STA. The parsed information is available through the STA
control interface command non_pref_chan[i] rows.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agotests: WNM Sleep Mode - AP side OOM
Jouni Malinen [Mon, 22 Feb 2016 17:35:39 +0000 (19:35 +0200)]
tests: WNM Sleep Mode - AP side OOM

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoWNM: Fix a memory leak on AP error path
Jouni Malinen [Mon, 22 Feb 2016 17:34:06 +0000 (19:34 +0200)]
WNM: Fix a memory leak on AP error path

If the second memory allocation in ieee802_11_send_wnmsleep_resp() were
to fail and ieee80211_11_get_tfs_ie() succeed, the wnmtfs_ie allocation
would not have been freed on the error path.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoVLAN: Fix vlan_compare() for tagged VLANs
Michael Braun [Sun, 21 Feb 2016 11:01:39 +0000 (12:01 +0100)]
VLAN: Fix vlan_compare() for tagged VLANs

While refactoring VLAN comparison into vlan_compare(), it was overlooked
that modifications are needed for tagged VLAN support.

Signed-off-by: Michael Braun <michael-dev@fami-braun.de>
8 years agohostapd_cli: Add support for RAW command
Dedy Lansky [Sun, 21 Feb 2016 12:49:44 +0000 (14:49 +0200)]
hostapd_cli: Add support for RAW command

Same as for wpa_cli, RAW command is a passthrough to hostapd control
interface.

Signed-off-by: Dedy Lansky <qca_dlansky@qca.qualcomm.com>
8 years agoMBO: Mandate use of PMF for WPA2+MBO association (STA)
Jouni Malinen [Mon, 22 Feb 2016 14:46:29 +0000 (16:46 +0200)]
MBO: Mandate use of PMF for WPA2+MBO association (STA)

If WPA2 is used, MBO AP must enable PMF. Refuse to select a BSS that has
MBO and WPA2 enabled without PMF.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoMBO: Mandate use of PMF for WPA2+MBO association (AP)
Jouni Malinen [Mon, 22 Feb 2016 14:39:24 +0000 (16:39 +0200)]
MBO: Mandate use of PMF for WPA2+MBO association (AP)

If WPA2 and MBO are enabled, PMF needs to be enabled in hostapd
configuration. If PMF is optional in the configuration, an MBO STA is
required to negotiate use of PMF.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agotests: MBO cellular data capability update with PMF required
Jouni Malinen [Mon, 22 Feb 2016 14:26:29 +0000 (16:26 +0200)]
tests: MBO cellular data capability update with PMF required

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agotests: MBO cellular data capability update
Jouni Malinen [Mon, 22 Feb 2016 10:06:33 +0000 (12:06 +0200)]
tests: MBO cellular data capability update

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoMBO: Update STA cellular data capability based on WNM Notification
Jouni Malinen [Mon, 22 Feb 2016 11:24:21 +0000 (13:24 +0200)]
MBO: Update STA cellular data capability based on WNM Notification

This makes hostapd parse a received WNM Notification Request frame
subelements and if a WFA MBO cellular data capability subelement is
seen, update the cellular data capability for the STA.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoMBO: Track STA cellular data capability from association request
Jouni Malinen [Mon, 22 Feb 2016 11:03:28 +0000 (13:03 +0200)]
MBO: Track STA cellular data capability from association request

This makes hostapd parse the MBO attribute in (Re)Association Request
frame and track the cellular data capability (mbo_cell_capa=<val> in STA
control interface command).

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoWNM: Minimal processing for WNM Notification Request frames on AP
Jouni Malinen [Mon, 22 Feb 2016 10:41:48 +0000 (12:41 +0200)]
WNM: Minimal processing for WNM Notification Request frames on AP

Write debug log entries on receiving WNM Notification Request frames.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoMBO: Indicate WNM-Notification support on AP when MBO is enabled
Jouni Malinen [Mon, 22 Feb 2016 10:41:00 +0000 (12:41 +0200)]
MBO: Indicate WNM-Notification support on AP when MBO is enabled

This is needed to allow MBO STAs to send WNM Notification Request
frames.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoSimplify hostapd_build_ap_extra_ies() with helper functions
Jouni Malinen [Mon, 22 Feb 2016 09:37:03 +0000 (11:37 +0200)]
Simplify hostapd_build_ap_extra_ies() with helper functions

This removes multiple copies of wpabuf_resize() following by
wpabuf_put_{buf,data}() with the help of two simple helper functions.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agotests: WNM BSS Transition Management query
Jouni Malinen [Mon, 22 Feb 2016 08:57:37 +0000 (10:57 +0200)]
tests: WNM BSS Transition Management query

This tests BSS Transition Management Query frame generation with
candidate list and transmission of the following request and response
frames.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoMBO: Expire non-matching bss_tmp_disallowed entries as part of check
Jouni Malinen [Sun, 21 Feb 2016 21:01:45 +0000 (23:01 +0200)]
MBO: Expire non-matching bss_tmp_disallowed entries as part of check

This makes wpa_is_bss_tmp_disallowed() expire old entries from the
bss_tmp_disallowed list even if they do not match the BSSID that is
being searched for. This allows the list to be kept at shorter length to
speed up operations and minimize memory use in cases where the
previously disabled BSS is not in radio range anymore.

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoMBO: Parse MBO IE in ieee802_11_parse_elems()
Jouni Malinen [Sun, 21 Feb 2016 18:54:23 +0000 (20:54 +0200)]
MBO: Parse MBO IE in ieee802_11_parse_elems()

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoMBO: Send WNM-Notification when cellular capabilities change
Avraham Stern [Mon, 15 Feb 2016 14:54:02 +0000 (16:54 +0200)]
MBO: Send WNM-Notification when cellular capabilities change

Send a WNM-Notification to the associated AP to indicate changes in
cellular data capabilities.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agotests: MBO association disallowed indication
Avraham Stern [Mon, 15 Feb 2016 14:53:55 +0000 (16:53 +0200)]
tests: MBO association disallowed indication

Add a test that verifies that no Association Request frame is sent to
APs that include the MBO IE with association disallowed attribute in
Beacon and Probe Response frames.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agotests: WNM BSS Transition Management with MBO IE
Avraham Stern [Mon, 15 Feb 2016 14:53:54 +0000 (16:53 +0200)]
tests: WNM BSS Transition Management with MBO IE

Add tests to verify that MBO IE in BSS Transition Management Request
frame is parsed correctly:

1. The MBO transition reason code is received by the MBO station.
2. The MBO cellular data connection preference is received by the
   MBO station.
3. The MBO station does not try to connect to the AP until the retry
   delay is over.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agotests: Enable CONFIG_MBO in hwsim configs
Ilan Peer [Mon, 15 Feb 2016 14:53:58 +0000 (16:53 +0200)]
tests: Enable CONFIG_MBO in hwsim configs

Signed-off-by: Ilan Peer <ilan.peer@intel.com>
8 years agohostapd: Add MBO IE to BSS Transition Management Request frame
Avraham Stern [Mon, 15 Feb 2016 14:53:49 +0000 (16:53 +0200)]
hostapd: Add MBO IE to BSS Transition Management Request frame

Add an option to add MBO IE to BSS Transition Management Request frame.
The MBO IE includes the transition reason code, cellular data connection
preference, and, if the disassoc imminent bit is set, it may also
include re-association retry delay. Otherwise, the re-association retry
delay should be set to zero.

The additional BSS_TM_REQ argument uses the following format:
mbo=<reason>:<reassoc delay>:<cell pref>
reason: 0-9
reassoc delay: 0-65535 (seconds; 0 = disabled)
cell pref: 0, 1, 255

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agohostapd: Add MBO IE to Beacon, Probe Response, Association Response
Avraham Stern [Mon, 15 Feb 2016 14:53:52 +0000 (16:53 +0200)]
hostapd: Add MBO IE to Beacon, Probe Response, Association Response

Add MBO IE with AP capability attribute to Beacon, Probe Response, and
(Re)Association Response frames to indicate the AP supports MBO.

Add option to add Association Disallowed attribute to Beacon, Probe
Response, and (Re)Association Response frames. Usage:

SET mbo_assoc_disallow <reason code>

Valid reason code values are between 1-5. Setting the reason code to
0 will remove the Association Disallowed attribute from the MBO IE
and will allow new associations.

MBO functionality is enabled by setting "mbo=1" in the config file.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoMove Hotspot 2.0 element in (Re)Association Request frames
Avraham Stern [Mon, 15 Feb 2016 14:53:48 +0000 (16:53 +0200)]
Move Hotspot 2.0 element in (Re)Association Request frames

According to IEEE Std 802.11-2012, Table 8-22, vendor specific elements
must follow all other elements, so Hotspot 2.0 element which is actually
a vendor specific element must come after all other elements.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoRe-order elements in (Re)Association Request frames
Avraham Stern [Mon, 15 Feb 2016 14:53:48 +0000 (16:53 +0200)]
Re-order elements in (Re)Association Request frames

According to IEEE Std 802.11-2012, Table 8-22, RM Enabled Capabilities
element must come before the Extended Capabilities element.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoWNM: Add candidate list to BSS transition query
Avraham Stern [Mon, 15 Feb 2016 14:53:43 +0000 (16:53 +0200)]
WNM: Add candidate list to BSS transition query

Add an option to configure a candidate list to BSS transition query
("list" as the second argument to WNM_BSS_QUERY). The candidate list is
built from the available scan results. If no updated scan results (< 10
sec) are available, the command fails.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoWNM: Add candidate list to BSS transition response
Avraham Stern [Mon, 15 Feb 2016 14:53:41 +0000 (16:53 +0200)]
WNM: Add candidate list to BSS transition response

Add the transition candidate list to BSS Transition Management Response
frame. The candidates preference is set using the regular wpa_supplicant
BSS selection logic. If the BSS transition request is rejected and
updated scan results are not available, the list is not added.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoutils: Derive phy type by frequency and bandwidth
David Spinadel [Mon, 15 Feb 2016 14:53:40 +0000 (16:53 +0200)]
utils: Derive phy type by frequency and bandwidth

Add a function to derive phy type from frequency and bandwidth
as defined in IEEE Std 802.11ac-2013 Annex C (dot11PHYType).

Signed-off-by: David Spinadel <david.spinadel@intel.com>
8 years agoMBO: Add MBO IE to BSS Transition Management Response frame
Avraham Stern [Mon, 15 Feb 2016 14:53:37 +0000 (16:53 +0200)]
MBO: Add MBO IE to BSS Transition Management Response frame

When rejecting a BSS Transition Management Request frame, add MBO IE to
the BSS Transition Management Response frame to specify the transition
rejection reason.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoMBO: Parse MBO IE in BSS Transition Management Request frames
Avraham Stern [Mon, 15 Feb 2016 14:53:36 +0000 (16:53 +0200)]
MBO: Parse MBO IE in BSS Transition Management Request frames

Add parsing of MBO IE in BSS Transition Management Request frames. If
the MBO IE includes the association retry delay attribute, do not try to
reconnect to the current BSS until the delay time is over.

If the MBO IE includes the cellular data connection preference attribute
or the transition rejection reason attribute, send a message to upper
layers with the data.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoMBO: Add Supported Operating Classes element to Association Request
Avraham Stern [Mon, 15 Feb 2016 14:53:34 +0000 (16:53 +0200)]
MBO: Add Supported Operating Classes element to Association Request

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoMBO: Add global operating class definitions
Avraham Stern [Mon, 15 Feb 2016 14:53:32 +0000 (16:53 +0200)]
MBO: Add global operating class definitions

Add definitions for global operating classes. These definitions will be
used to construct supported operating classes information element.

The operating classes definitions used locally for P2P module will be
removed and included in the general operating classes definitions.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoMBO: Prevent association to APs that explicitly disallow this
David Spinadel [Mon, 15 Feb 2016 14:53:29 +0000 (16:53 +0200)]
MBO: Prevent association to APs that explicitly disallow this

Prevent association to MBO APs that have association disallowed
attribute in MBO IE in Beacon or Probe Response frames.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
8 years agoMBO: Add cellular capability to MBO IE
David Spinadel [Mon, 15 Feb 2016 14:53:28 +0000 (16:53 +0200)]
MBO: Add cellular capability to MBO IE

Add cellular capability attribute to MBO IE and add MBO IE with cellular
capabilities to Probe Request frames. By default, cellular capability
value is set to Not Cellular capable (3).

Signed-off-by: David Spinadel <david.spinadel@intel.com>
8 years agoMBO: Send MBO WNM-Notification Request frames to notify changes
David Spinadel [Mon, 15 Feb 2016 14:53:25 +0000 (16:53 +0200)]
MBO: Send MBO WNM-Notification Request frames to notify changes

Send a WNM-Notification Request frame with Non-preferred Channel Report
subelement if the non-preferred channels list changes during an
association.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
8 years agoMBO: Implement MBO non-preferred channel report in Association Request
David Spinadel [Mon, 15 Feb 2016 14:53:24 +0000 (16:53 +0200)]
MBO: Implement MBO non-preferred channel report in Association Request

Add MBO IE with non-preferred channels to (Re)Association Request
frames.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
8 years agoMBO: Add non-preferred channel configuration in wpa_supplicant
David Spinadel [Mon, 15 Feb 2016 14:53:22 +0000 (16:53 +0200)]
MBO: Add non-preferred channel configuration in wpa_supplicant

Add non-preferred channel configuration to wpa_config for MBO.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
8 years agoMBO: Add Multi Band Operation definitions
David Spinadel [Mon, 15 Feb 2016 14:53:19 +0000 (16:53 +0200)]
MBO: Add Multi Band Operation definitions

These are based on the specification draft WFA_MBO_TechSpec_v0.0_r19.

Signed-off-by: David Spinadel <david.spinadel@intel.com>
8 years agondis: Use the new get_ie() helper to avoid duplicated code
Jouni Malinen [Sun, 21 Feb 2016 15:07:47 +0000 (17:07 +0200)]
ndis: Use the new get_ie() helper to avoid duplicated code

This removes more duplicated implementation of finding the first IE
based on the id.

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoutils: Share a single helper function to get IE by ID
Avraham Stern [Mon, 15 Feb 2016 14:53:17 +0000 (16:53 +0200)]
utils: Share a single helper function to get IE by ID

Add a helper function to find a certain IE inside IEs buffer by ID and
use this function in several places that implemented similar
functionality locally.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agowpa_supplicant: Share a single get_mode() implementation
Avraham Stern [Mon, 15 Feb 2016 14:53:13 +0000 (16:53 +0200)]
wpa_supplicant: Share a single get_mode() implementation

There is no need to duplicate this helper function in multiple files.

Signed-off-by: Avraham Stern <avraham.stern@intel.com>
8 years agoVLAN: Check vlan_desc validity in a failure debug print
Jouni Malinen [Sat, 20 Feb 2016 17:22:43 +0000 (19:22 +0200)]
VLAN: Check vlan_desc validity in a failure debug print

The recent VLAN changes added an explicit code path that sets vlan_desc
= NULL within ap_sta_set_vlan(). This makes some code analyzers warn
about the debug print that could potentially dereference this pointer.
Silence that warning by verifying the pointer more consistently within
this function.

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoUse 64-bit TX/RX byte counters for statistics
Nick Lowe [Fri, 19 Feb 2016 15:22:25 +0000 (15:22 +0000)]
Use 64-bit TX/RX byte counters for statistics

If the driver supports 64-bit TX/RX byte counters, use them directly.
The old 32-bit counter extension is maintained for backwards
compatibility with older drivers.

For nl80211 driver interface, the newer NL80211_STA_INFO_RX_BYTES64 and
NL80211_STA_INFO_TX_BYTES64 attributes are used when available. This
resolves the race vulnerable 32-bit value wrap/overflow. Rework RADIUS
accounting to use these for Acct-Input-Octets, Acct-Input-Gigawords,
Acct-Output-Octets, and Acct-Output-Gigawords, these values are often
used for billing purposes.

Signed-off-by: Nick Lowe <nick.lowe@lugatech.com>
8 years agoAP: Set STA assoc flag in the driver before sending Assoc Resp frame
Andrei Otcheretianski [Tue, 16 Feb 2016 09:54:33 +0000 (11:54 +0200)]
AP: Set STA assoc flag in the driver before sending Assoc Resp frame

Previously, stations were added to the driver only after the
(Re)Association Response frame was acked. In the time period between the
station has acked the (Re)Association Response frame and the time the
station was added to the kernel, the station can already start sending
Data frames, which will be dropped by the hardware/driver. In addition
to the data loss, the driver may ignore NDPs with PM bit set from this
STA.

Fix this by setting/adding the STA with associated flag set to the
driver before the AP sends the (Re)Association Response frame with
status success. If the (Re)Association Response frame wasn't acked,
remove the station from the driver.

Note that setting a station to associated state before the non-AP
station acknowledges the (Re)Association Response frame is not compliant
with the IEEE 802.11 standard that specifically states that a non-AP
station should transition to authenticated/associated state only after
it acknowledged the (Re)Association Response frame. However, this is a
justifiable simplification to work around the issue described above since

1. The station will be removed in case it does not acknowledge the
   (Re)Association Response frame.
2. All Data frames would be dropped until the station is set to
   authorized state and there are no known issues with processing the
   other Class 3 frames during the short window before the
   acknowledgement is seen.

Signed-off-by: Andrei Otcheretianski <andrei.otcheretianski@intel.com>
8 years agoAP: Add support for full station state
Ayala Beker [Tue, 16 Feb 2016 09:54:32 +0000 (11:54 +0200)]
AP: Add support for full station state

Add support for drivers that support full AP client state, i.e., can
handle adding stations that are not associated yet. For such drivers,
add a station after processing the authentication request, instead of
adding it in the association response callback.

Doing so is beneficial in cases where the driver cannot handle the add
station request, in which case it is useless to perform the complete
connection establishment.

Signed-off-by: Ayala Beker <ayala.beker@intel.com>
8 years agonl80211: Add support for full station state operations
Ayala Beker [Tue, 16 Feb 2016 09:54:31 +0000 (11:54 +0200)]
nl80211: Add support for full station state operations

This provides means for determining whether the driver supports full AP
station state and setting the needed STA flags for using this
functionality.

Signed-off-by: Ayala Beker <ayala.beker@intel.com>
8 years agotests: EAP-FAST protocol testing
Jouni Malinen [Sat, 20 Feb 2016 10:05:48 +0000 (12:05 +0200)]
tests: EAP-FAST protocol testing

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoEAP-FAST peer: Remove fixed return value from eap_fast_parse_phase1()
Jouni Malinen [Sat, 20 Feb 2016 08:06:48 +0000 (10:06 +0200)]
EAP-FAST peer: Remove fixed return value from eap_fast_parse_phase1()

This function was always returning 0, so the error path was unreachable.

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agotests: Verify fast_max_pac_list_len=0 special case
Jouni Malinen [Sat, 20 Feb 2016 08:06:02 +0000 (10:06 +0200)]
tests: Verify fast_max_pac_list_len=0 special case

Signed-off-by: Jouni Malinen <j@w1.fi>
8 years agoEAP-pwd server: Use os_get_random() for unpredictable token
Nick Lowe [Tue, 9 Feb 2016 16:02:32 +0000 (16:02 +0000)]
EAP-pwd server: Use os_get_random() for unpredictable token

Do not use os_random() that uses a low quality PRNG to generate the
anti-clogging token. The construction can be improved upon by replacing
it with a call to os_get_random(), which uses a high quality PRNG. While
the RFC 5931 explictly recommends not to do this ("SHOULD NOT be from a
source of random entropy"), it does still mandate unpredicability ("MUST
be unpredictable"). The anti-clogging token is most unpredictable when
it is taken from a high quality PRNG.

Signed-off-by: Nick Lowe <nick.lowe@lugatech.com>
8 years agoDFS: Remove the os_random() fallback
Nick Lowe [Tue, 9 Feb 2016 14:49:35 +0000 (14:49 +0000)]
DFS: Remove the os_random() fallback

Remove the fallback dependency on os_random() from the code that gets a
valid DFS channel. This is exceptionally unlikely to ever be called as
the call to os_get_random() is unlikely to fail. The intention is to
facilitate future removal of os_random() as it uses a low quality PRNG.

Signed-off-by: Nick Lowe <nick.lowe@lugatech.com>
8 years agoWPS: Use only os_get_random() for PIN generation
Nick Lowe [Tue, 9 Feb 2016 14:47:47 +0000 (14:47 +0000)]
WPS: Use only os_get_random() for PIN generation

Remove the fallback dependency on os_random() when generating a WPS pin.
This is exceptionally unlikely to ever be called as the call to
os_get_random() is unlikely to fail. The intention is to facilitate
future removal of os_random() as it uses a low quality PRNG.

Signed-off-by: Nick Lowe <nick.lowe@lugatech.com>
8 years agoUse os_get_random() for Shared Key authentication challenge
Nick Lowe [Tue, 9 Feb 2016 14:43:21 +0000 (14:43 +0000)]
Use os_get_random() for Shared Key authentication challenge

Do not use the system clock or os_random() that uses a low quality PRNG
as part of the pseudo-random challenge in auth_shared_key(). The
construction can be improved upon by replacing it with a call to
os_get_random(), which uses a high quality PRNG.

Signed-off-by: Nick Lowe <nick.lowe@lugatech.com>
8 years agotests: Allow PIN generation failure during OOM in ap_wps_random_ap_pin
Jouni Malinen [Fri, 19 Feb 2016 16:43:45 +0000 (18:43 +0200)]
tests: Allow PIN generation failure during OOM in ap_wps_random_ap_pin

This is needed to avoid reporting failures after a change to remove the
fallback path in PIN generation.

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agoAdd RADIUS Service-Type attribute with a value of Framed
Nick Lowe [Sat, 13 Feb 2016 12:20:12 +0000 (12:20 +0000)]
Add RADIUS Service-Type attribute with a value of Framed

This seems to be the common value used by APs and also mentioned in RFC
3580.

Signed-off-by: Nick Lowe <nick.lowe@lugatech.com>
8 years agotests: Renew expired certificates
Jouni Malinen [Fri, 19 Feb 2016 16:41:23 +0000 (18:41 +0200)]
tests: Renew expired certificates

Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
8 years agomesh: Drop Authentication frames from BLOCKED STA
Masashi Honma [Wed, 17 Feb 2016 10:48:53 +0000 (19:48 +0900)]
mesh: Drop Authentication frames from BLOCKED STA

Previously, only mesh Action frames from BLOCKED STA were dropped.
Extend that to drop Authentication frames as well.

Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
8 years agoSAE: Fix PMKID calculation for PMKSA cache
Masashi Honma [Mon, 15 Feb 2016 02:23:37 +0000 (11:23 +0900)]
SAE: Fix PMKID calculation for PMKSA cache

The SAE PMKID is calculated with IEEE Std 802.11-2012 11.3.5.4, but the
PMKID was re-calculated with 11.6.1.3 and saved into PMKSA cache. Fix
this to save the PMKID calculated with 11.3.5.4 into the PMKSA cache.

Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
8 years agoPrint Acct-Session-Id and Acct-Multi-Session-Id 64-bit values
Nick Lowe [Sat, 13 Feb 2016 20:29:47 +0000 (20:29 +0000)]
Print Acct-Session-Id and Acct-Multi-Session-Id 64-bit values

These are now 64-bit variables and the printf formats and type casts
need to be updated to match.

Signed-off-by: Nick Lowe <nick.lowe@lugatech.com>
8 years agokqueue: Use 0 instead of NULL for udata
Roy Marples [Wed, 10 Feb 2016 20:39:14 +0000 (20:39 +0000)]
kqueue: Use 0 instead of NULL for udata

Use 0 for udata instead of NULL for portability.
NetBSD uses uintptr_t, others a pointer.

Signed-off-by: Roy Marples <roy@marples.name>
8 years agoctype functions require an unsigned char
Roy Marples [Wed, 10 Feb 2016 19:27:42 +0000 (19:27 +0000)]
ctype functions require an unsigned char

Ensure that characters are represented as unsigned char when using
isblank() and isspace(). These function take in a "int c" argument, but
it needs to be unsigned for the cases where EOF is not indicated.

Signed-off-by: Roy Marples <roy@marples.name>
8 years agoFix compile on NetBSD for vlan
Roy Marples [Wed, 10 Feb 2016 11:53:41 +0000 (11:53 +0000)]
Fix compile on NetBSD for vlan

Shuffle includes above system ones so to fix a compile issue
on NetBSD where the if_type #define from <net/if.h>
conflicts with the wpa_driver_if_type enum.

Signed-off-by: Roy Marples <roy@marples.name>
8 years agowired: Fix compile on NetBSD for wired driver
Roy Marples [Wed, 10 Feb 2016 11:22:11 +0000 (11:22 +0000)]
wired: Fix compile on NetBSD for wired driver

Shuffle wpa_supplicant includes above system ones so that
to fix a compile problem on NetBSD where if_type #define
conflicts with the wpa_driver_if_type enum.

Signed-off-by: Roy Marples <roy@marples.name>
8 years agoAdd CONFIG_ELOOP_KQUEUE to defconfig
Roy Marples [Wed, 10 Feb 2016 10:48:01 +0000 (10:48 +0000)]
Add CONFIG_ELOOP_KQUEUE to defconfig

Signed-off-by: Roy Marples <roy@marples.name>