Jouni Malinen [Sun, 19 Aug 2012 15:25:25 +0000 (18:25 +0300)]
wpabuf: Get rid of separate ext_data pointer
Use an explicit pointer to the beginning of the buffer and a flag
to indicate whether that is to external data or not. This avoids
a branch whenever accessing the buffer and helps some static
analyzers to understand the wpabuf memory uses better.
Jouni Malinen [Sun, 19 Aug 2012 14:23:00 +0000 (17:23 +0300)]
Fix memory allocation failure handling in EAP-TTLS/MSCHAPv2 server
If the os_malloc() call for the User-Name value fails in EAP-TTLS
server, the inner MSCHAPv2 processing could have tried to dereference a
NULL pointer. Avoid this by handling this cleanly as an internal error
and reject the authentication attempt.
Jouni Malinen [Sun, 19 Aug 2012 13:53:15 +0000 (16:53 +0300)]
FIPS: Remove md5-non-fips.c
Commit c9e08af24fd7dda3f21674cdc744579b8c38fa28 removed the only user of
the special case MD5 use that would be allowed in FIPS mode in
tls_prf_sha1_md5(). Commit 271dbf1594bea461ea2ea7946a773a30bfa254aa
removed the file from the build, but left the implementation into the
repository. To clean things up even further, remove this functionality
completely since it is not expected to be needed for FIPS mode anymore.
Michael Braun [Sun, 19 Aug 2012 11:27:30 +0000 (14:27 +0300)]
Add User-Name/CUI from RADIUS ACL in Accounting messages
This allows User-Name and Chargeable-User-Identity attributes to be
passed from Access-Accept into Accounting messages even when IEEE 802.1X
is not used.
Signed-hostap: Michael Braun <michael-dev@fami-braun.de>
Jouni Malinen [Sun, 19 Aug 2012 09:47:20 +0000 (12:47 +0300)]
Interworking: Do not match credentials without WPA2-Enterprise
Since we currently support only HS 2.0 networks with Interworking
network selection, do not indicate credential match unless the
network uses WPA2-Enterprise.
Jouni Malinen [Sun, 19 Aug 2012 09:39:39 +0000 (12:39 +0300)]
Interworking: Enforce WPA2-Enterprise/CCMP
Since we currently support only HS 2.0 networks with Interworking
network selection, enforce that WPA2-Enterprise/CCMP is used on the
AP instead of allowing any WPA-Enterprise combination.
Jouni Malinen [Sun, 19 Aug 2012 09:34:14 +0000 (12:34 +0300)]
HS 2.0: Do not indicate HS 2.0 status without WPA2-Enterprise
Hotspot 2.0 allows only WPA2-Enterprise to be used, so other types of
networks must not be indicated as Hotspot 2.0 networks even if they
(incorrectly) advertise HS 2.0 support.
Jouni Malinen [Sun, 19 Aug 2012 09:26:21 +0000 (12:26 +0300)]
Interworking: Do not reconnect without "auto" parameter
When there was no credential match, but an enabled network block matched
with a scan result, wpa_supplicant reconnected at the end of
interworking_select command even if "auto" parameter was not used. Fix
this by running the reconnect only if requested to automatically select
a network.
Jouni Malinen [Fri, 17 Aug 2012 20:55:14 +0000 (23:55 +0300)]
Fix EAP-FAST with OpenSSL 1.0.1
The mechanism to figure out key block size based on ssl->read_hash
does not seem to work with OpenSSL 1.0.1, so add an alternative
mechanism to figure out the NAC key size that seems to work at
least with the current OpenSSL 1.0.1 releases.
Jouni Malinen [Fri, 17 Aug 2012 19:26:28 +0000 (22:26 +0300)]
Disable TLS Session Ticket extension by default for EAP-TLS/PEAP/TTLS
Some deployed authentication servers seem to be unable to handle the TLS
Session Ticket extension (they are supposed to ignore unrecognized TLS
extensions, but end up rejecting the ClientHello instead). As a
workaround, disable use of TLS Sesson Ticket extension for EAP-TLS,
EAP-PEAP, and EAP-TTLS (EAP-FAST uses session ticket, so any server that
supports EAP-FAST does not need this workaround).
Jouni Malinen [Fri, 17 Aug 2012 19:21:28 +0000 (22:21 +0300)]
Add phase1/phase2 parameter for disabling TLS Session Ticket extension
tls_disable_session_ticket=1 in phase1/phase2 can now be used to disable
use of TLS Session Ticket extension (which is enabled by default in
recent OpenSSL versions). This can be used to work around issues with
broken authentication servers that do not ignore unrecognized TLS
extensions properly.
Yoni Divinsky [Fri, 17 Aug 2012 17:53:44 +0000 (20:53 +0300)]
Supply the connection scan results to bgscan after init
If bgscan is enabled, then bgscan is initiated after the connection,
with no bss scan result information. Since a scan was performed prior to
the connection, the information exists and can be used to initialize the
bgscan's cache.
Avinash Patil [Fri, 17 Aug 2012 16:21:40 +0000 (19:21 +0300)]
hostapd: Disable WPS 2.0 when WPA-TKIP only configuration is set
WSC 2.0 deprecates use of WPA (TKIP) and does not allow WPA/TKIP only
configuration (i.e., WPA2/CCMP needs to be enabled in mixed mode for
WPA/TKIP to be acceptable). Enforce this by disabling WPS if
configuration enables WPA/TKIP without WPA2/CCMP.
Jouni Malinen [Thu, 16 Aug 2012 19:56:19 +0000 (22:56 +0300)]
OpenSSL: Silence a compiler warning with OpenSSL 0.9.7
The PKCS5_PBKDF2_HMAC_SHA1() function in OpenSSL 0.9.7 did not mark
the salt parameter const even though it was not modified. Hide the
compiler warning with a type cast when an old OpenSSL version is
used.
Jouni Malinen [Thu, 16 Aug 2012 16:38:15 +0000 (19:38 +0300)]
Do not export TLS keys in FIPS mode
Only allow the TLS library keying material exporter functionality to be
used for MSK derivation with TLS-based EAP methods to avoid exporting
internal TLS keys from the library.
Jouni Malinen [Thu, 16 Aug 2012 16:29:34 +0000 (19:29 +0300)]
OpenSSL: Use internal keying material exporter when possible
Use SSL_export_keying_material() if possible, i.e., if OpenSSL is
version 1.0.1 or newer and if client random value is used first. This
allows MSK derivation with TLS-based EAP methods (apart from EAP-FAST)
without exporting the master key from OpenSSL.
Jouni Malinen [Thu, 16 Aug 2012 16:02:44 +0000 (19:02 +0300)]
EAP-TTLS: Allow TTLS to be built without MSCHAPv2 support
Previously, CONFIG_EAP_MSCHAPV2=y was assumed to be set for
CONFIG_EAP_TTLS=y. Avoid this dependency by making including the
MSCHAPv2 parts in EAP-TTLS conditionally.
Jouni Malinen [Thu, 16 Aug 2012 15:38:11 +0000 (18:38 +0300)]
MSCHAPv2: Check challenge_hash() failures
In theory, the SHA1 operation may fail (e.g., if SHA1 becomes disallowed
in some security policies), so better check the return code from
challenge_hash().
Jouni Malinen [Thu, 16 Aug 2012 14:53:05 +0000 (17:53 +0300)]
Add "make fips" for wpa_supplicant builds with OpenSSL
This can be used for convenience to build wpa_supplicant with OpenSSL
FIPS object module. CONFIG_FIPS=y is expected to be used in
wpa_supplicant/.config when using this option.
Jouni Malinen [Thu, 16 Aug 2012 14:38:46 +0000 (17:38 +0300)]
OpenSSL: Fix memory leak on FIPS error paths
Do not leave the tls_global context allocated if the global OpenSSL
initialization fails. This was possible in case of FIPS builds if
the FIPS mode cannot be initialized.
P2P: Enable 40 MHz support for autonomous P2P group addition
Add optional "ht40" argument for p2p_group_add command to enable 40 MHz
in 5GHz band. This configures the secondary channel, when HT support is
enabled and if the HW supports 40 MHz channel width.
Jouni Malinen [Wed, 15 Aug 2012 16:34:04 +0000 (19:34 +0300)]
Allow non-WPA IEEE 802.1X to be select even if WPA is also enabled
If key_mgmt was set to allow both WPA and non-WPA IEEE 802.1X (i.e., to
IEEE8021X WPA-EAP), non-WPA IEEE 802.1X was rejected while preparing
association parameters. Allow this special case to be handled by
selecting non-WPA case if the scan results for the AP do not include
either WPA or RSN elements.
Eyal Shapira [Mon, 13 Aug 2012 18:48:55 +0000 (21:48 +0300)]
Fix adding extra IEs in sched scan
Refactoring done in commit 46ee0427b13d64dc4695fda24f4f228c99822c11
(IEEE 802.11u: Allow Interworking and HESSID to be configured)
broke adding extra IEs in sched scan. The IEs stopped being added
to the scan params within wpa_supplicant_extra_ies() but no code
was added to add them in wpa_supplicant_req_sched_scan().
Fix this and remove unused params arg in wpa_supplicant_extra_ies().
Jouni Malinen [Mon, 13 Aug 2012 18:33:00 +0000 (21:33 +0300)]
GAS: Fix reporting of GAS query timeouts
GAS_QUERY_TIMEOUT value was used for two different things - enum for
status callback and #define for internal eloop timeout). The latter
overwrites the former and as such, the timeout reported ended up going
out with value 5 which matches with GAS_QUERY_CANCELLED instead of
GAS_QUERY_TIMEOUT. This value was not used in existing code, so this
should not modify the current behavior. Anyway, the correct reason for
the failure should be reported. Rename the internal #define for the
timeout period to avoid the name conflict. [Bug 463]
Jouni Malinen [Mon, 13 Aug 2012 18:20:09 +0000 (21:20 +0300)]
Add os_realloc_array() wrapper
This function can be used as a wrapper for os_realloc(ptr, nmemb * size)
when a reallocation is used for an array. The main benefit over
os_realloc() is in having an extra check to catch integer overflows in
multiplication. This is similar to the os_zalloc() to os_calloc() move.
Jouni Malinen [Mon, 13 Aug 2012 17:42:24 +0000 (20:42 +0300)]
Add os_calloc() wrapper
This function can be used as a wrapper for os_zalloc(nmemb * size) when
an allocation is used for an array. The main benefit over os_zalloc() is
in having an extra check to catch integer overflows in multiplication.
Jouni Malinen [Sun, 12 Aug 2012 08:36:45 +0000 (11:36 +0300)]
Remove default ACM=1 from AC_VO and AC_VI
There is no need to mandate admission control (ACM=1) by default, so
clear that flag in the case the configuration file does not specify
wmm_ac_{vo,vi}_acm value. This gets closer to the values mentioned
in the sample hostapd.conf file and reduces the need to specify WMM
parameters in the configuration file for most common use cases.
In case of P2P GO and AP mode, wpa_supplicant uses the default hostapd
parameters for WMM. In the default parameters the ACM bit for video and
voice are set to 1, meaning, P2P devices and stations which are
connected to the GO cannot pass voice or video data packets. Allow this
to be changed through wpa_supplicant configuration file with wmm_ac_*
parameters.
Jouni Malinen [Sat, 11 Aug 2012 18:14:57 +0000 (21:14 +0300)]
P2P: Increase GO Negotiation timeouts
p2p_set_timeout() calls in GO Neg Req/Resp TX callbacks used timeout of
100 ms which is the value given in the P2P specification for GO
Negotiation, but this was actually shorter than the
wait-for-offchannel-TX value (200 ms) used for the driver call. In
addition, it looks like some devices (e.g., Galaxy Nexus with JB image)
can take longer time to reply to GO Negotiation Response (somewhere
between 200 and 250 ms has been observed).
Increase the wait-for-GO-Neg-Resp timeout from 100 ms to 200 ms if GO
Negotiation Request frame was acknowledged (this matches with the
offchannel wait timeout that used previously). The no-ack case is left
at 100 ms since we use GO Negotiation Request frame also to discover
whether the peer is on its Listen channel.
Increase the wait-for-GO-Neg-Conf timeout from 100 ms to 250 ms (and
increase the offchannel wait timeout to matching 250 ms) as a workaround
for devices that take over 200 ms to reply to GO Negotiation Response.
However, a device driver could return EINVAL for get_bssid() because it
recognizes it has already been disconnected. When the wpa_supplicant
received EINVAL, the bssid field could be used uninitialized in the
following flow:
Arik Nemtsov [Sat, 11 Aug 2012 14:08:54 +0000 (17:08 +0300)]
SME: Correctly check mode HT caps for enabling OBSS scan
Don't assume the 11g mode is always first in the list of mode (sometimes
it isn't). Traverse the array of modes and check the HT40 capability is
turned on for 11g.
Jouni Malinen [Sat, 11 Aug 2012 09:41:31 +0000 (12:41 +0300)]
P2P: Move group removal reason to a function parameter
wpa_s->removal_reason was set only when calling wpas_p2p_group_delete()
and while couple of call places did not set this, it should really be
set in each case. As such, it works better as a function parameter than
a variable in struct wpa_supplicant.
The commit b5fd8b18335f2a7784a1f0a5eb5dbe45976d9e21
"VHT: Do not allow use of TKIP with VHT" introduced
an off-by-one regression: a WPA/RSN 11n AP would no
longer accept stations to connect, instead it would
produce a bogus error message:
... Station tried to use TKIP with HT association.
Signed-hostapd: Christian Lamparter <chunkeey@googlemail.com>
Jouni Malinen [Fri, 10 Aug 2012 15:05:03 +0000 (18:05 +0300)]
PMKSA: Set cur_pmksa pointer during initial association
cur_pmksa was left to NULL during the initial association. This can
result in unexpected behavior, e.g., in expiring PMKSA cache entries
since the current entry is not locked in that case. Fix this by updated
cur_pmksa when adding the initial PMKSA entry during msg 1/4 processing.
Dan Williams [Fri, 10 Aug 2012 14:55:17 +0000 (17:55 +0300)]
PMKSA: Do not evict active cache entry when adding new ones
If the PMKSA cache is full (i.e., 32 candidates have been seen in scan
results and have not yet expired) then any additional entries can
potentially evict the current/active entry (if it is the oldest entry),
which triggers a pointless local deauthentication. The supplicant
shouldn't replace the current/active entry if it is still valid, but
instead the oldest entry that is *not* the current/active one.
Signed-hostap: Dan Williams <dcbw@redhat.com>
intended-for: hostap-1
Michael Braun [Fri, 10 Aug 2012 13:48:35 +0000 (16:48 +0300)]
Fix WPA GTK rekeying with multiple VLANs
When using multiple VLANs, GKeyDoneStations counter is not updated
properly since wpa_auth_for_each_sta() call in wpa_group_setkeys() ends
up iterating through all STAs and not just the STAs of a specific
wpa_group (VLAN). Consequently, GTK rekeying gets initialized multiple
times if more than a single group state machine exists. Fix this by
iterating only through the STAs in the specific wpa_group.
Signed-hostap: Michael Braun <michael-dev@fami-braun.de>
intended-for: hostap-1
Eyal Shapira [Fri, 10 Aug 2012 13:28:38 +0000 (16:28 +0300)]
Add p2p_go_max_inactivity config option
This enables setting a different max inactivity timeout for P2P GO.
This timeout is used to detect inactive clients. In some scenarios
it may be useful to have control over this and set a shorter timeout
than the default 300s. For example when running STA and P2P GO interfaces
concurrently, the STA interface may perform scans which may cause the
GO to miss a disassoc / deauth frames from a client and keep assuming
that the client is connected until the inactivity detection kicks in.
300 secs is a bit too long for such scenarios and creates a bad user
experience.
Michael Braun [Fri, 10 Aug 2012 09:55:33 +0000 (12:55 +0300)]
Add option to use netlink to create and remove VLAN interfaces
CONFIG_VLAN_NETLINK=y build option can now be used to replace the
ioctl()-based interface for creating and removing VLAN interfaces
with netlink-based interface.
Signed-hostap: M. Braun <michael-dev@fami-braun.de>
Jouni Malinen [Fri, 10 Aug 2012 08:48:57 +0000 (11:48 +0300)]
nl80211: Fix re-enabling of 802.11b rates after P2P use
Commit a11241fa114923b47892ad3279966839e9c2741d removed the 802.11b rate
enabling/disabling code from wpa_driver_nl80211_set_mode() and while
doing that, removed the only place where drv->disabled_11b_rates was
set. Fix this by updating the flag in nl80211_disable_11b_rates(). In
addition, re-enable the 802.11b rates when changing to non-P2P mode.
nl80211: Disable 11b rates on configuring P2P interface
Right now 11b rates are masked out while creating a P2P interface,
but this is always failing as the interface is down. Most drivers
allow to configure rates only when the interface is UP and running.
So let us disable 11b rates when interface type is changed into a
P2P type and it is UP and running.
Jouni Malinen [Tue, 7 Aug 2012 20:03:25 +0000 (23:03 +0300)]
Add extra validation of EAP header length field
These validation steps are already done in the EAP parsing code and in
the EAP methods, but the additional check is defensive programming and
can make the validation of received EAP messages more easier to
understand.
Jouni Malinen [Tue, 7 Aug 2012 18:59:15 +0000 (21:59 +0300)]
Add explicit EAPOL-Key length validation in processKey()
These checks would not really be needed since eapol_sm_rx_eapol()
validates the length fields. Anyway, these makes it more obvious to
anyone reviewing the code that there are no integer underflow issues in
processKey().
Jouni Malinen [Tue, 7 Aug 2012 18:11:04 +0000 (21:11 +0300)]
RADIUS: Add explicit attribute length validation checks in functions
These checks would not really be needed since radius_msg_parse()
validates the attribute header fields. Anyway, these makes it more
obvious to anyone reviewing the code that there are no integer underflow
issues in the functions processing RADIUS attributes.