LibreSSL does not seem have SSL_CTX_set_num_tickets(), so comment out
these not really critical calls when building with that library.
Fixes: 81e2498889 ("OpenSSL: Limit the number of TLS 1.3 session tickets to one")
Fixes: decac7cd1e ("OpenSSL: Do not send out a TLS 1.3 session ticket if caching disabled")
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
EVP_PKEY_CTX_set_rsa_oaep_md() does not seem to be available in
LibreSSL, so for now, comment out this functionality whenever building
with that library.
Fixes: 36b11bbcff ("OpenSSL: RSA-OAEP-SHA-256 encryption/decryption")
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
SSL_CTX_set_num_tickets() is not available in boringSSL.
So protect the call to SSL_CTX_set_num_tickets() under
!defined(OPENSSL_IS_BORINGSSL) to fix the compilation error.
Fixes: decac7cd1e ("OpenSSL: Do not send out a TLS 1.3 session ticket if caching disabled")
Fixes: 81e2498889 ("OpenSSL: Limit the number of TLS 1.3 session tickets to one")
Signed-off-by: Sunil Ravi <sunilravi@google.com>
Add a notification message to indicate reason for TLS handshake failure
due to the server not supporting safe renegotiation (RFC 5746).
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
The TLS protocol design for renegotiation was identified to have a
significant security flaw in 2009 and an extension to secure this design
was published in 2010 (RFC 5746). However, some old RADIUS
authentication servers without support for this are still used commonly.
This is obviously not good from the security view point, but since there
are cases where the user of a network service has no realistic means for
getting the authentication server upgraded, TLS handshake may still need
to be allowed to be able to use the network.
OpenSSL 3.0 disabled the client side workaround by default and this
resulted in issues connection to some networks with insecure
authentication servers. With OpenSSL 3.0, the client is now enforcing
security by refusing to authenticate with such servers. The pre-3.0
behavior of ignoring this issue and leaving security to the server can
now be enabled with a new phase1 parameter allow_unsafe_renegotiation=1.
This should be used only when having to connect to a network that has an
insecure authentication server that cannot be upgraded.
The old (pre-2010) TLS renegotiation mechanism might open security
vulnerabilities if the authentication server were to allow TLS
renegotiation to be initiated. While this is unlikely to cause real
issues with EAP-TLS, there might be cases where use of PEAP or TTLS with
an authentication server that does not support RFC 5746 might result in
a security vulnerability.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
Some of the TLS library wrappers defined only an empty function for
tls_connection_set_success_data(). That could result in memory leaks in
TLS server cases, so update these to do the minimal thing and free the
provided buffer as unused.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
TLS 1.3 sends the OCSP response with the server Certificate message.
Since that Certificate message is not sent when resuming a session,
there can be no new OCSP response. Allow this since the OCSP response
was validated when checking the initial certificate exchange.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
One session ticket is sufficient for EAP-TLS, so do not bother
generating more than a single session ticket.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
A port of the trivial patch I wrote for FreeRADIUS to allow TLS decoding
in Wireshark for hostapd/wpa_supplicant:
df0eb0a884
Signed-off-by: Alexander Clouter <alex@coremem.com>
This may fail with FIPS builds because the FIPS requirement is that the
password must be at least 14 characters.
Signed-off-by: Juliusz Sosinowicz <juliusz@wolfssl.com>
Some API is not available when using FIPS. We need to allocate memory
and initialize the structs directly.
Signed-off-by: Juliusz Sosinowicz <juliusz@wolfssl.com>
Register a callback with wolfCrypt_SetCb_fips to inform the user of
errors in the wolfCrypt FIPS module.
Signed-off-by: Juliusz Sosinowicz <juliusz@wolfssl.com>
In some configurations the wc_Init*() functions may either allocate
memory or other system resources. These resources need to be freed.
Co-authored-by: JacobBarthelmeh <jacob@wolfssl.com>
Signed-off-by: Juliusz Sosinowicz <juliusz@wolfssl.com>
Add support for loading private keys and certificates in both PEM and
DER formats with wolfSSL.
Signed-off-by: Juliusz Sosinowicz <juliusz@wolfssl.com>
Add new crypto wrappers for performing RSA-OAEP-SHA-256 encryption and
decryption. These are needed for IMSI privacy.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
It looks like the OpenSSL callbacks for SSL_SESSION can end up calling
the remove callback for multiple SSL_SESSION entries that share the same
ex data. This could result in double freeing the session data on the
server side.
Track the SSL_SESSION ex data in a separate list and free the
allocations only if they are pointing to a valid allocated wpabuf
pointer.
Signed-off-by: Jouni Malinen <quic_jouni@quicinc.com>
The result (y2) was cleared and freed in error cases, but the pointer
itself was not cleared to NULL, so the error path would have returned a
pointer to freed memory. Fix this by properly clearing the return value
in error cases. In addition, simplify the function to avoid tracking
success case separately.
Signed-off-by: Jouni Malinen <j@w1.fi>
wc_HmacSetKey() seems to initialize everything that is needed for the
actual operation, but at least valgrind is reporting use of
uninitialized values when this was done on a data structure that was not
explicitly cleared.
Call wc_HmacInit() before wc_HmacSetKey() to avoid any unexpected
behavior from potentially uninitialized values.
Signed-off-by: Jouni Malinen <j@w1.fi>
SSL_CTX_set_dh_auto() is not available in BoringSSL even though it
claims to be based on OpenSSL 1.1.1 in OPENSSL_VERSION_NUMBER.
Fixes: ebb3055e13 ("OpenSSL: Generate DH parameters automatically if not set with dh_file")
Signed-off-by: Jouni Malinen <j@w1.fi>
Anything older than OpenSSL 1.0.2 should not be used anymore, so drop
the build compatibility options for those old versions.
Signed-off-by: Jouni Malinen <j@w1.fi>
Compilation was failing even with LibreSSL 2.7 versions, so there is no
point in maintaining these old compatibility options for older versions.
Signed-off-by: Jouni Malinen <j@w1.fi>
OpenSSL 3.0 has deprecated the low-level EC_KEY functionality, so use
the EVP API (EVP_EC_gen() and EV_PKEY_derive()) for the crypto_ecdh
wrappers.
Signed-off-by: Jouni Malinen <j@w1.fi>
OpenSSL 1.1.1 deprecated the function names with the _GFp postfix, so
move to the new function name. Add backwards compatibility wrappers for
older versions.
Signed-off-by: Jouni Malinen <j@w1.fi>
This is needed to work with TLS 1.3 key derivation. It looks the needed
functionality was added in wolfSSL 4.7.0.
Signed-off-by: Jouni Malinen <j@w1.fi>
priv_sz and pub_sz needs to be initialized to the buffer size before the
wc_DhGenerateKeyPair() call. The previous version happened to work in
some cases where a separate handled prime length was used, but not for
the generic case.
Signed-off-by: Jouni Malinen <j@w1.fi>
It looks like crypto_ecdh_set_peerkey() had started failing at some
point with a wolfSSL update due to ECC_TIMING_RESISTANT from
--enable-harden requiring RNG to be set.
Signed-off-by: Jouni Malinen <j@w1.fi>
Add support for EAP server to cache TLS sessions (see hostapd
configuration parameter tls_session_lifetime).
Signed-off-by: Juliusz Sosinowicz <juliusz@wolfssl.com>
The previous mechanism of unloaded the providers from tls_deinit() did
not work correctly for some cases. In particular, it was possible for
hostapd to end up unloading both providers and not being able to recover
from this if TLS server was not enabled.
Address this more cleanly by introducing a new crypto_unload() function
that will be called when the process is exiting.
Fixes: 097ca6bf0b ("OpenSSL: Unload providers on deinit")
Signed-off-by: Jouni Malinen <j@w1.fi>
Comment out the call to SSL_use_RSAPrivateKey_ASN1() function when using
OpenSSL 3.0 since that function was deprecated and there does not seem
to be any significant use case for supporting DER encoded RSAPrivateKey
structure in the private key blob.
Signed-off-by: Jouni Malinen <j@w1.fi>
OpenSSL 3.0 has deprecated low-level DH/DSA functions. Convert this
loading of the dh_file parameter to use newer API.
Signed-off-by: Jouni Malinen <j@w1.fi>
This was used earlier in the wpa_supplicant configuration, but there was
no real use for that, so remove the now completely unused parameter to
simplify implementation.
Signed-off-by: Jouni Malinen <j@w1.fi>
These operations do not really have any effect since
tls_connection_set_params() is used only in the TLS client case and the
client receives the DH parameters from the server instead of local
configuration.
Signed-off-by: Jouni Malinen <j@w1.fi>
BoringSSL commit dddb60eb9700110835ff6e2b429de40a17006429 ("Make most of
crypto/x509 opaque.") broke the direct access to these variables, so use
the accessor functions instead.
Signed-off-by: Jouni Malinen <j@w1.fi>
BoringSSL commit dddb60eb9700110835ff6e2b429de40a17006429 ("Make most of
crypto/x509 opaque.") broke the stack-allocated version, so move to
using X509_STORE_CTX_new() to get an opaque pointer to X509_STORE_CTX
instead.
Signed-off-by: Jouni Malinen <j@w1.fi>
Define the EC_GROUP_get_curve() wrapper for the older versions.
Fixes: 3c61f4db4c ("OpenSSL: Replace EC_GROUP_get_curve_GFp() calls with EC_GROUP_get_curve()")
Signed-off-by: Jouni Malinen <j@w1.fi>
