The EVP_MAC context data needs to be freed on error paths.
Fixes: e31500adea ("OpenSSL: Implement HMAC using the EVP_MAC API")
Signed-off-by: Jouni Malinen <j@w1.fi>
The conversion to the new OpenSSL 3.0 API had forgotten to free the
context structure.
Fixes: bcd299b326 ("OpenSSL: Convert DH/DSA parameter loading to new API")
Signed-off-by: Jouni Malinen <j@w1.fi>
Extend IMSI privacy functionality to allow an attribute (in name=value
format) to be added using the new imsi_privacy_attr parameter.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This can be used to make sure wpa_supplicant does not process DPP
messages sent in Public Action frames when a test setup is targeting
DPP-over-TCP.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Use imsi_privacy_cert as the name of the configuration parameter for the
X.509v3 certificate that contains the RSA public key needed for IMSI
privacy. The only allowed format for this information is a PEM-encoded
X.509 certificate, so the previous name was somewhat confusing.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This is an allocated resource so it needs to be free on the error path.
Fixes: 42871a5d25 ("EAP-SIM/AKA peer: IMSI privacy")
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Disable a network profile that has set the imsi_privacy_key if a valid
key cannot be read from the specified file. Previously, this check was
done only after having associated, but there is no point in associating
just to see EAP authentication fail in such a case. This is needed for
avoiding connection attempts if the X.509 certificate for IMSI privacy
has expired.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Commit 9afb68b039 ("OpenSSL: Allow systemwide secpolicy overrides for
TLS version") with commit 58bbcfa31b ("OpenSSL: Update security level
drop for TLS 1.0/1.1 with OpenSSL 3.0") allow this workaround to be
enabled with an explicit network configuration parameter. However, the
default settings are still allowing TLS 1.0 and 1.1 to be negotiated
just to see them fail immediately when using OpenSSL 3.0. This is not
exactly helpful especially when the OpenSSL error message for this
particular case is "internal error" which does not really say anything
about the reason for the error.
It is is a bit inconvenient to update the security policy for this
particular issue based on the negotiated TLS version since that happens
in the middle of processing for the first message from the server.
However, this can be done by using the debug callback for printing out
the received TLS messages during processing.
Drop the OpenSSL security level to 0 if that is the only option to
continue the TLS negotiation, i.e., when TLS 1.0/1.1 are still allowed
in wpa_supplicant default configuration and OpenSSL 3.0 with the
constraint on MD5-SHA1 use.
Signed-off-by: Jouni Malinen <j@w1.fi>
Unfortunately, some objects (WlantestCapture, WpaSupplicant
and wpaspy.Ctrl) use __del__ and actually have some logic
there. This is more or less wrong, and we should be using
context managers for it. However, cleaning that up is a
pretty large task.
Unfortunately, __del__ can cause reentrant logging which is
wrong too, because it might be invoked while in the middle
of a logging call, and the __del__ of these objects closes
connections and logs while doing that.
Since we're (likely) using cpython, we can work around this
by explicitly calling gc.collect() in a context where the
logging and close is fine, not only ensuring that all the
connections are closed properly before the next test, but
also fixing the issue with reentrant logging.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Instead of printing a very long line of the failed tests, print the test
case names on separate lines up to the number of available lines at the
bottom of the screen. This avoids some issues with curses and overlong
lines. Furthermore, display the last failed test cases instead of
somewhat confusing sequence of test case names from the VMs.
Signed-off-by: Jouni Malinen <j@w1.fi>
tcp_addr=from-uri can now be used as a special case for initiating
DPP-over-TCP to the destination indicated in the peer bootstrapping URI.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Parse the host information, if present, in bootstrapping URI and allow
such information to be added when generating the URI.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Only use the bandwidth bits that are applicable for the current
operating band. This avoids use of reserved bits when determining the
length of the Support EHT-MCS And NSS Set field length.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This test case could fail if there was an old BSS entry from a previous
test case in the scan results.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
This test case was checking the exact key info bits in EAPOL-Key frames
during PTK rekeying as such, needs to be updated to match the
implementation change on the Secure bit setting.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
IEEE Std 802.11-2020 is ambiguous on how the Secure bit is set in
EAPOL-Key msg 1/4 and 2/4 in the case where 4-way handshake is use to
rekey the PTK. 12.7.2 describes this with "set to 1 once the initial key
exchange is complete" while 12.7.6 shows EAPOL-Key msg 1/4 and 2/4 using
Secure=0 without any consideration on whether the handshake is for
rekeying.
TGme seems to be moving towards clarifying this to use Secure=1 based on
there being a shared PTKSA between the Authenticator and the Supplicant.
In other words, this would use Secure=1 in EAPOL-Key msg 1/4 and 2/4 in
the case of rekeying. Change implementation to match that. This bit was
already practically ignored on the reception side, so this should not
have impact on actual functionality beyond this one bit changing its
value in the frame.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
There was a possible race condition here between the hostapd request
transmission and wpa_supplicant response command. Wait for the
wpa_supplicant event that indicates reception of the request before
issuing the DSCP_RESP command to avoid failures.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Check the full MBIE length against the buffer length explicitly before
the debug print. This is for locally generated data, so the bounds
checking is not critical here, but it is better to use proper checking
anyway to avoid static analyzer complaints.
Signed-off-by: Jouni Malinen <j@w1.fi>
Avoid the somewhat confusing mechanism of determining the bitfield index
from the assigned IP address to make this easier for static analyzers.
Signed-off-by: Jouni Malinen <j@w1.fi>
Limit the GAS comeback delay to 60000 TUs, i.e., about 60 seconds. This
is mostly to silence static analyzers that complain about unbounded
value from external sources even though this is clearly bounded by being
a 16-bit value.
Signed-off-by: Jouni Malinen <j@w1.fi>
The length of the URL, i.e., pos[0], is verified here to be within the
bounds of the recieved message, but that seemed to be done in a manner
that might bee too complex for static analyzers to understand.
Signed-off-by: Jouni Malinen <j@w1.fi>
Use local variables and common checking of the selector (or vendor
specific IE header) to make the bounds checking easier to understand.
Signed-off-by: Jouni Malinen <j@w1.fi>
BIT(r) is not sufficient here since it does not cover 64 bit values.
Write this out with 1ULL to be large enough for the shift operation.
Signed-off-by: Jouni Malinen <j@w1.fi>
The EHT changes made this checking inconsistent. If he_cap can be NULL
in case of EHT being enabled, better make sure it does not get
dereferenced without an explicit check.
Signed-off-by: Jouni Malinen <j@w1.fi>
