The dh_p_len, dh_g_len, and dh_ys_len parameters were validated against
the received message structure, but that did not seem to be done in a
way that some static analyzers would understand this (CID 72699).
Signed-off-by: Jouni Malinen <j@w1.fi>
This makes the MACsec implementation more consistent with rest of
wpa_supplicant. (CID 72677, CID 72695, CID 72701, CID 72709, CID 72711)
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit e8c08c9a36 ('EAP-FAST server: Fix
potential read-after-buffer (by one byte)') changed the while loop
design in a way that does not require the pos variable to be updated
anymore. Remove that unneeded code to clean up static analyzer warnings
about unused assignments.
Signed-off-by: Jouni Malinen <j@w1.fi>
The hlen and len variables are identical here, but only the hlen was
used in the end. Change this to use the len variable to avoid
unnecessary static analyzer warnings about unused writes.
Signed-off-by: Jouni Malinen <j@w1.fi>
The driver_test.c driver wrapper (-Dtest in wpa_supplicant and
driver=test in hostapd) was previously used for testing without real
Wi-Fi hardware. mac80211_hwsim-based tests have practically replaced all
these needs and there has been no improvements or use for driver_test.c
in a long while. Because of this, there has not really been any effort
to maintain this older test tool and no justification to change this
either. Remove the obsoleted test mechanism to clean up the repository.
Signed-off-by: Jouni Malinen <j@w1.fi>
It looks like the use of sm->wpa == WPA_VERSION_WPA2 in two locations
within the function was a bit too much for clang static analyzer to
understand. Use a separate variable for storing the allocated memory so
that it can be freed unconditionally. The kde variable can point to
either stack memory or temporary allocation, but that is now const
pointer to make the design clearer.
Signed-off-by: Jouni Malinen <j@w1.fi>
dl_list_del() followed by dl_list_add() seemed to confuse clang static
analyzer somehow, so explicitly check for the prev pointer to be
non-NULL to avoid an incorrect warning.
Signed-off-by: Jouni Malinen <j@w1.fi>
The preceding dl_list_len() check guarantees that dl_list_first()
returns an entry and not NULL. However, that seems to be a bit too
difficult path to follow for static analyzers, so add an explicit check
for the dl_list_first() return value to be non-NULL to avoid warnings.
Signed-off-by: Jouni Malinen <j@w1.fi>
This allows drivers that implement AP SME internally to generate a
Deauthentication or Disassociation frame with the specified reason code.
This was already done with drivers that use hostapd/wpa_supplicant for
AP SME.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The new ext_eapol_frame_io parameter can be used to configure hostapd
and wpa_supplicant to use control interface for receiving and
transmitting EAPOL frames. This makes it easier to implement automated
test cases for protocol testing. This functionality is included only in
CONFIG_TESTING_OPTIONS=y builds.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
A P2P Client may be able to connect to the GO even if the WPS
provisioning step has not terminated cleanly (e.g., P2P Client does not
send WSC_Done). Such group formation attempt missed the event
notification about started group on the GO and also did not set the
internal state corresponding to the successful group formation.
This commit addresses the missing part by completing GO side group
formation on a successful first data connection if WPS does not complete
cleanly. Also, this commit reorders the STA authorization indications to
ensure that the group formation success notification is given prior to
the first STA connection to handle such scenarios.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Use os_exec() to run the external browser to avoid undesired command
line processing for control interface event strings. Previously, it
could have been possible for some of the event strings to include
unsanitized data which is not suitable for system() use.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Use os_exec() to run the external browser to avoid undesired command
line processing for control interface event strings. Previously, it
could have been possible for some of the event strings to include
unsanitized data which is not suitable for system() use.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Use os_exec() to run the external browser to avoid undesired command
line processing for control interface event strings. Previously, it
could have been possible for some of the event strings to include
unsanitized data which is not suitable for system() use.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This allows cases where neither 80 MHz segment requires DFS to be
configured. DFS CAC operation itself does not yet support 80+80, though,
so if either segment requires DFS, the AP cannot be brought up.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This extends the "XOR t" operation in aes_wrap() and aes_unwrap() to
handle up to four octets of the n*h+i value instead of just the least
significant octet. This allows the plaintext be longer than 336 octets
which was the previous limit.
Signed-off-by: Jouni Malinen <j@w1.fi>
This adds kek_len argument to aes_wrap() and aes_unwrap() functions and
allows AES to be initialized with 192 and 256 bit KEK in addition to
the previously supported 128 bit KEK.
The test vectors in test-aes.c are extended to cover all the test
vectors from RFC 3394.
Signed-off-by: Jouni Malinen <j@w1.fi>
The (int) typecast I used with sk_GENERAL_NAME_num() to complete the
BoringSSL compilation was not really the cleanest way of doing this.
Update that to use stack_index_t variable to avoid this just like the
other sk_*_num() calls.
Signed-off-by: Jouni Malinen <j@w1.fi>
BoringSSL is Google's cleanup of OpenSSL and an attempt to unify
Chromium, Android and internal codebases around a single OpenSSL.
As part of moving Android to BoringSSL, the wpa_supplicant maintainers
in Android requested that I upstream the change. I've worked to reduce
the size of the patch a lot but I'm afraid that it still contains a
number of #ifdefs.
[1] https://www.imperialviolet.org/2014/06/20/boringssl.html
Signed-off-by: Adam Langley <agl@chromium.org>
_pmksa_cache_free_entry() is a static function that is never called with
entry == NULL, so there is no need to check for that.
Signed-off-by: Jouni Malinen <j@w1.fi>
If the first entry in the PMKSA cache did not match the station's MAC
address, an infinite loop could be reached in pmksa_cache_get_okc() when
trying to find a PMKSA cache entry for opportunistic key caching cases.
This would only happen if OKC is enabled (okc=1 included in the
configuration file).
Signed-off-by: Jouni Malinen <j@w1.fi>
Previously, WPA/WPA2 case ended up using the hardcoded
dot11RSNAConfigPMKLifetime (43200 seconds) for PMKSA cache entries
instead of using the Session-Timeout value from the RADIUS server (if
included in Access-Accept). Store a copy of the Session-Timeout value
and use it instead of the default value so that WPA/WPA2 cases get the
proper timeout similarly to non-WPA/WPA2 cases.
Signed-off-by: Jouni Malinen <j@w1.fi>
Use hostapd_freq_params instead of simple frequency parameter for driver
commands. This is preparation for IBSS configuration to allow use of
HT/VHT in IBSS.
Signed-off-by: Janusz Dziedzic <janusz.dziedzic@tieto.com>
Commit 4fdc8def88 changed the wpa_ctrl
socket to be be non-blocking, so the comment about wpa_ctrl_recv()
blocking is not valid anymore.
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit 97279d8d (after hostap-2.0) dropped frame events from foreign
addresses. Unfortunately this commit did not handle the case where the
interface's MAC address might be changed externally, which other
wpa_supplicant code already handled. This causes the driver to reject
any MLME event because the address from the event doesn't match the
stale address in the driver data.
Changing an interface's MAC address requires that the interface be
down, the change made, and then the interface brought back up. This
triggers an RTM_NEWLINK event which driver_nl80211.c can use to
re-read the MAC address of the interface.
Signed-hostap: Dan Williams <dcbw@redhat.com>
Relying on qos qosinfo is not enough, as it can be 0 for WMM enabled
peers that don't support U-APSD. Further, some peers don't even contain
this IE (Google Nexus 5), but do contain the WMM IE during setup.
Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
A recent mac80211 patch ("8f02e6b mac80211: make sure TDLS peer STA
exists during setup") forces the TDLS STA to exist before sending any
mgmt packets. Add the STA before sending a concurrent-setup test packet.
Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Some drivers need to know the initiator of a TDLS connection in order
to generate a correct TDLS mgmt packet. It is used to determine
the link identifier IE. Pass this information to the driver.
Signed-off-by: Arik Nemtsov <arikx.nemtsov@intel.com>
Commit 97ed9a06df ('nl80211: Remove bridge
FDB entry upon sta_remove()') used nl_sock and nl_socket_* functions
which are not compatible with older versions of libnl. Fix this.
Signed-off-by: Ilan Peer <ilan.peer@intel.com>
IEEE 802.11 standard sends AID in a field that is defined in a bit
strange way to set two MSBs to ones. That is not the real AID and those
extra bits need to be filtered from the value before passing this to the
driver.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
random_mac_addr_keep_oui() is similar to random_mac_addr(), but it
maintains the OUI part of the source address.
Signed-off-by: Jouni Malinen <j@w1.fi>
The FDB entry removal ensures that the traffic destined for a
disassociated station's MAC address is no longer forwarded from the
bridge to the BSS.
Signed-off-by: Kyeyoon Park <kyeyoonp@qca.qualcomm.com>
Support 60 GHz band in P2P module by selecting random social channel
from all supported social channels in 2.4 GHz and 60 GHz bands.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
In case P2P is not enabled the if (dev_addr) is always ignored as
dev_addr will be NULL. As this code is relevant only to P2P, it can be
moved to be the ifdef to avoid static analyzer warnings. (CID 72907)
Signed-off-by: Philippe De Swert <philippe.deswert@jollamobile.com>
'\e' representing ESC (0x1b) is not C standard, it's an GNU extension.
https://gcc.gnu.org/onlinedocs/gcc/Character-Escapes.html#Character-Escapes
Since the code also compiles on Windows with Microsoft compiler, we
should use '\033' instead.
Note: I didn't try to build the whole wpa_supplicant on Windows, so I
don't know if it still builds (I have no Visual Studio 2005 for a quick
test). I just needed the string conversion routines for the P"" syntax
in both directions.
Signed-off-by: Bernhard Walle <bernhard@bwalle.de>
The sd_pending_bcast_queries variable should be decremented only
in case of success. This way, the supplicant can retry if a service
discovery request fails.
Signed-off-by: Constantin Musca <constantin.musca@intel.com>
It was possible for the p2p_go_neg_start timeout handler to get called
when there was a pending timeout from an earlier GO Negotiation start.
This could result in that old timeout expiring too early for the newly
started GO Negotiation to complete. Avoid such issues by setting a
sufficiently long timeout here just before triggering the new GO
Negotiation.
Signed-off-by: Jouni Malinen <j@w1.fi>
Previously, it was possible for the p2p->pending_listen_freq to be left
at non-zero value if Probe Request frame was received from a peer with
which we were waiting to start GO Negotiation/Invite process. That could
result in the following Listen operation getting blocked in some
operation sequences if the peer did not acknowledge the following P2P
Public Action frame.
Signed-off-by: Jouni Malinen <j@w1.fi>
Previously, extra connect/roam events were ignored when user space SME
was used to avoid confusing double events for association. However,
there was no matching code for ignoring auth/assoc events when using
driver SME. Such events would not normally show up since the driver SME
case would not generated them. However, when testing forced connect
command with force_connect_cmd=1 driver param, these events are
indicated.
Ignore the extra events in testing cases to make the hwsim test cases
match more closely the real code path for driver SME. In addition, this
resolves some test case failures where double association event could
end up causing a failure, e.g., when doing PMKSA caching in
pmksa_cache_oppurtunistic_connect.
Signed-off-by: Jouni Malinen <j@w1.fi>
This fixes couple of code paths where the WPA_DRIVER_FLAGS_DFS_OFFLOAD
flag was not checked properly and unexpected DFS operations were
initiated (and failed) in case the driver handles all these steps.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
In theory, this call could fail, so check the return value before using
the received data. These specific cases would not really care much about
the failures, but this keeps the code more consistent and keeps static
analyzer warnings more useful. (CID 72678, CID 72679, CID 72680,
CID 72683, CID 72689, CID 72698, CID 72703)
Signed-off-by: Jouni Malinen <j@w1.fi>
The driver-based SME case did not set STA flags properly to the kernel
in the way that hostapd-SME did in ieee802_11.c. This resulted in the FT
protocol case not marking the STA entry authorized. Fix that by handling
the special WLAN_AUTH_FT case in hostapd_notif_assoc() and also add the
forgotten hostapd_set_sta_flags() call to synchronize these flag to the
driver.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This allows updating roaming policy for drivers that select the BSS
internally so that wpa_supplicant (based on bssid parameter
configuration) and the driver remain in sync.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The network block bssid parameter can be used to force a specific BSS to
be used for a connection. It is also possible to modify this parameter
during an association. Previously, that did not result in any
notification to the driver which was somewhat problematic with drivers
that take care of BSS selection. Add a new mechanism to allow
wpa_supplicant to provide a driver update if the bssid parameter change
for the current connection modifies roaming policy (roaming
allowed/disallowed within ESS).
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
If a TX status event and RX event for a GO Negotiation frame gets
delayed long enough to miss the initial wait, it was possible for
reception of a GO Negotiation Response frame with status 1 to try to
initiate a new p2p-listen work item to wait for the peer to become ready
while a previous p2p-listen was already in progress due to that earlier
timeout while waiting for peer. This would result in the new
start_listen request getting rejected ("P2P: Reject start_listen since
p2p_listen_work already exists") and the negotiation not proceeding.
Work around this by using P2P_WAIT_PEER_CONNECT state instead of
P2P_WAIT_PEER_IDLE if P2P_CONNECT_LISTEN state has already been entered
when processing this special GO Negotiation Response status=1 case. This
can avoid double-scheduling of p2p-listen and as such, completion of the
GO negotiation even if the driver event or peer response are not
received in time (the response is supposed to be there within 100 ms per
spec, but there are number of deployed devices that do not really meet
this requirement).
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Previously, the shorter startWhen value was used based on build
parameters (i.e., if WPS was enabled). This is not really ideal and the
knowledge of WPS use can be provided to the EAPOL state machine to allow
this (and similar WPS workarounds) to be done only when the association
is for the purpose of WPS.
Reduce the default startWhen value from 3 to 2 seconds for non-WPS case
since WPS builds have likely received most testing for the past years
with the 1 second value and there is no strong justification for forcing
the longer 3 second wait should a frame be lost or something else
require the EAPOL-Start to initiate operation after a connection.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The internal entropy pool was previously used to prevent 4-way handshake
in AP mode from completing before sufficient entropy was available to
allow secure keys to be generated. This commit extends that workaround
for boards that do not provide secure OS level PRNG (e.g., /dev/urandom
does not get enough entropy) for the most critical WPS operations by
rejecting AP-as-enrollee case (use of AP PIN to learn/modify AP
configuration) and new PSK/passphrase generation. This does not have any
effect on devices that have an appropriately working OS level PRNG
(e.g., /dev/random and /dev/urandom on Linux).
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Even when hs20 action frame is unable to be registered,
for whatever reason, it should be possible to register
event handle for received driver messages. This patch also
avoids a segmentation fault, when p2p and hs20 were enabled
and GO NEG was unable to create p2p iface, the destroy eloop
was crashing by reading an invalid handle.
Signed-off-by: Eduardo Abinader <eduardo.abinader@openbossa.org>
In theory, this function could fail, so check the return value before
printing out the RADIUS local address debug message (CID 72700).
Signed-off-by: Jouni Malinen <j@w1.fi>
Due to a missing curly brackets, the OCSP status checking was not
working in the expected way. Instead of allowing optional-OCSP
configuration to accept connection when OCSP response was ready, all
such cases were marked as hard failures. In addition, the debug prints
were not necessarily accurate for the mandatory-OCSP-but-no-response
case (CID 72694, CID 72704).
Signed-off-by: Jouni Malinen <j@w1.fi>
Instead of using implicit limit based on 16-bit unsigned integer having
a maximum value of 65535, limit the maximum length of a CID explicitly
to 1500 bytes. This will hopefully help in reducing false warnings from
static analyzers (CID 72712).
Signed-off-by: Jouni Malinen <j@w1.fi>
TDLS Setup Request frame has to be rejected with status code 37 ("The
request has been declined"), if the BSSID in the received Link
Identifier does not match the current BSSID per IEEE Std 802.11-2012,
10.22.4 ('TDLS direct-link establishment') step (b). The previously used
status code 7 ('Not in same BSS') is described to used only when
processing TPK Handshake Message 2 in TDLS Setup Response frame.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
When a netdev is removed from a bridge, RTM_DELLINK message is received
with ifname (IFLA_IFNAME) pointing to the main netdev event though that
netdev is not deleted. This was causing issues with P2P GO interface
getting disabled when the netdev was removed from a bridge. Fix this by
filtering RTM_DELLINK events that are related to the bridge when
indicating interface status changes.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
While hostapd should not really receive the EVENT_ASSOC message for
IBSS, driver_nl80211.c could potentially generate that if something
external forces the interface into IBSS mode and the IBSS case does not
provide the struct assoc_info data. Avoid the potential NULL pointer
dereference by explicitly verifying for the event data to be present.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Disable Probe Request reporting for statically created AP interfaces
during de-initialization. We will enable it again while starting AP
operations.
Signed-off-by: Avinash Patil <avinashapatil@gmail.com>
Some devices have limitations which do not allow changing virtual
interface mode from AP to station or vice versa. To work around this,
check if such AP interface is not dynamic. If such an interface is
enumarated, just set ifmode to AP and avoid setting nlmode to default
station mode on deinit.
Signed-off-by: Avinash Patil <avinashapatil@gmail.com>
This adds hostapd support for the new WLAN-Pairwise-Cipher,
WLAN-Group-Cipher, WLAN-AKM-Suite, and WLAN-Group-Mgmt-Pairwise-Cipher
attributes defined in RFC 7268. These attributes are added to RADIUS
messages when the station negotiates use of WPA/RSN.
Signed-off-by: Jouni Malinen <j@w1.fi>
This adds hostapd support for the new WLAN-HESSID attribute defined in
RFC 7268. This attribute contains the HESSID and it is added whenever
Interworking is enabled and HESSID is configured.
Signed-off-by: Jouni Malinen <j@w1.fi>
This adds hostapd support for the new Mobility-Domain-Id attribute
defined in RFC 7268. This attribute contains the mobility domain id and
it is added whenever the station negotiates use of FT.
Signed-off-by: Jouni Malinen <j@w1.fi>
The special PAC_OPAQUE_TYPE_PAD case did not skip incrementing of the
pos pointer and could result in one octet read-after-buffer when parsing
the PAC-Opaque data.
Signed-off-by: Jouni Malinen <j@w1.fi>
This forces EAP peer implementation to drop any possible fast resumption
data if the network block for the current connection is not the same as
the one used for the previous one. This allows different network blocks
to be used with non-matching parameters to enforce different rules even
if the same authentication server is used. For example, this allows
different CA trust rules to be enforced with different ca_cert
parameters which can prevent EAP-TTLS Phase 2 from being used based on
TLS session resumption.
Signed-off-by: Jouni Malinen <j@w1.fi>
This is similar to wpa_msg_global() in the same way as wpa_msg_ctrl() is
to wpa_msg(). In other words, wpa_msg_global_ctrl() is used to send
global control interface events without printing them into the debug
log.
Signed-off-by: Jouni Malinen <j@w1.fi>
Use an explicit memset call to clear any configuration parameter and
dynamic data that contains private information like keys or identity.
This brings in an additional layer of protection by reducing the length
of time this type of private data is kept in memory.
Signed-off-by: Jouni Malinen <j@w1.fi>
This makes the EAP-pwd server and peer implementations more robust
should OpenSSL fail to derive random number for some reason. While this
is unlikely to happen in practice, the implementation better be prepared
for this should something unexpected ever happen. See
http://jbp.io/2014/01/16/openssl-rand-api/#review-of-randbytes-callers
for more details.
Signed-off-by: Florent Daigniere <nextgens@freenetproject.org>
This makes the implementation less likely to provide useful timing
information to potential attackers from comparisons of information
received from a remote device and private material known only by the
authorized devices.
Signed-off-by: Florent Daigniere <nextgens@freenetproject.org>
This changes OpenSSL calls to explicitly clear the EC_POINT memory
allocations when freeing them. This adds an extra layer of security by
avoiding leaving potentially private keys into local memory after they
are not needed anymore. While some of these variables are not really
private (e.g., they are sent in clear anyway), the extra cost of
clearing them is not significant and it is simpler to just clear these
explicitly rather than review each possible code path to confirm where
this does not help.
Signed-off-by: Florent Daigniere <nextgens@freenetproject.org>
This changes OpenSSL calls to explicitly clear the bignum memory
allocations when freeing them. This adds an extra layer of security by
avoiding leaving potentially private keys into local memory after they
are not needed anymore. While some of these variables are not really
private (e.g., they are sent in clear anyway), the extra cost of
clearing them is not significant and it is simpler to just clear these
explicitly rather than review each possible code path to confirm where
this does not help.
Signed-off-by: Florent Daigniere <nextgens@freenetproject.org>
FreeRADIUS releases before 1.1.4 did not send MS-CHAP2-Success in
EAP-TTLS/MSCHAPv2. A wpa_supplicant workaround for that was added in
2005 and it has been enabled by default to avoid interoperability
issues. This could be disabled with all other EAP workarounds
(eap_workaround=0). However, that will disable some workarounds that are
still needed with number of authentication servers.
Old FreeRADIUS versions should not be in use anymore, so it makes sense
to remove this EAP-TTLS/MSCHAPv2 workaround completely to get more
complete validation of server behavior. This allows MSCHAPv2 to verify
that the server knows the password instead of relying only on the TLS
certificate validation.
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
When driver interface is destroyed (via wpa_driver_nl80211_deinit) the
"preq" nl socket isn't always deleted but the callback struct associated
is. After the interface is destroyed we may still get event on the
socket but as the callback has been freed this will cause wpa_supplicant
to crash.
This patch ensures that the "preq" socket is destroyed when destroying
the interface.
Signed-off-by: Cedric IZOARD <cedricx.izoard@intel.com>
