Both Android's libc and glibc support _FORTIFY_SOURCE, a compiler
and libc feature which inserts automatic bounds checking into
common C functions such as memcpy() and strcpy(). If a buffer
overflow occurs when calling a hardened libc function, the
automatic bounds checking will safely shutdown the program and
prevent memory corruption.
Android is experimenting with _FORTIFY_SOURCE=3, a new fortify
level which enhances memcpy() to prevent overflowing an element
of a struct. Under the enhancements, code such as
struct foo {
char empty[0];
char one[1];
char a[10];
char b[10];
};
int main() {
foo myfoo;
int n = atoi("11");
memcpy(myfoo.a, "01234567890123456789", n);
return 0;
}
will cleanly crash when the memcpy() call is made.
Fixup hostap code to support the new level. Specifically:
* Fixup sha1_transform so it works with the enhanced bounds checking.
The old memcpy() code was attempting to write to context.h0, but that
structure element is too small and the write was extending (by design)
into h1, h2, h3, and h4. Use explicit assignments instead of
overflowing the struct element.
* Modify most of the structures in ieee802_11_defs.h to use ISO C99
flexible array members (https://gcc.gnu.org/onlinedocs/gcc/Zero-Length.html)
instead of a zero length array. Zero length arrays have zero length,
and any attempt to call memcpy() on such elements will always overflow.
Flexible array members have no such limitation. The only element not
adjusted is probe_req, since doing so will generate a compile time error,
and it's not obvious to me how to fix it.
Signed-off-by: Nick Kralevich <nnk@google.com>
The internal TLS implementation started rejecting number of unsupported
configuration parameters recently, but those new error paths did not
free the allocated tlsv1_credentials buffer.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This function exposes internal state of the TLS negotiated parameters
for the sole purpose of being able to implement PRF for EAP-FAST. Since
tls_connection_prf() is now taking care of all TLS-based key derivation
cases, it is cleaner to keep this detail internal to each tls_*.c
wrapper implementation.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This is not needed anymore with the tls_connection_prf() being used to
handle all key derivation needs. tls_connection_get_keys() is a bit
misnamed for now, but it is only used to fetch the client and server
random for Session-Id derivation.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
tls_openssl.c is the only remaining TLS/crypto wrapper that needs the
internal PRF implementation for EAP-FAST (since
SSL_export_keying_material() is not available in older versions and does
not support server-random-before-client case). As such, it is cleaner to
assume that TLS libraries support tls_connection_prf() and move the
additional support code for the otherwise unsupported cases into
tls_openssl.c.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This replaces the internal CBC mode implementation in
aes_128_cbc_encrypt() and aes_128_cbc_decrypt() with the OpenSSL
implementation for CONFIG_TLS=openssl builds.
Signed-off-by: Jouni Malinen <j@w1.fi>
The local buffers may contain information used to generate parts of the
derived key, so clear these explicitly to minimize amount of unnecessary
private key-related material in memory.
Signed-off-by: Jouni Malinen <j@w1.fi>
The local hash[] buffer may contain parts of the derived key, so clear
it explicitly to minimize number of unnecessary copies of key material
in memory.
Signed-off-by: Jouni Malinen <j@w1.fi>
The local T[] buffer may contain parts of the derived key, so clear it
explicitly to minimize number of unnecessary copies of key material in
memory.
Signed-off-by: Jouni Malinen <j@w1.fi>
SChannel/CryptoAPI as a TLS/crypto library alternative was never
completed. Critical functionality is missing and there are bugs in this
implementation. Since there are no known plans of completing this
support, it is better to remove this code.
Signed-off-by: Jouni Malinen <j@w1.fi>
If OpenSSL reports that a presented leaf certificate is invalid,
but it has been explicitly pinned, accept it anyway.
Signed-off-by: Rohit Agrawal <rohit.agrawal.mn@gmail.com>
These were not supposed to include a newline at the end of the message
text since such formatting gets handled by tls_show_errors(). In
addition, change the message about the issuer's issuer to be more
accurate.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
If addition of a peer issuer certificate fails, the certs pointer would
be NULL when being passed to sk_X509_push() for peer issuer's issuer.
Fix this by skipping addition of issuer's issue if issuer addition
fails.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This is going to be required for OpenSSL 1.1.0 which makes the SSL
structure opaque. Older versions starting from OpenSSL 1.0.1 include
this function, so start using it now based on OPENSSL_VERSION_NUMBER.
Signed-off-by: Jouni Malinen <j@w1.fi>
OpenSSL 0.9.8 and newer includes SSL_CTX_get_cert_store() and
SSL_CTX_set_cert_store() helper functions, so there is no need to
dereference the SSL_CTX pointer to cert ssl_ctx->cert_store. This helps
in working with the future OpenSSL 1.1.0 release that makes the SSL_CTX
structure opaque.
Signed-off-by: Jouni Malinen <j@w1.fi>
This replaces the implementation in aes-wrap.c and aes-unwrap.c with
OpenSSL AES_wrap_key() and AES_unwrap_key() functions when building
hostapd or wpa_supplicant with OpenSSL.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
These have reached out-of-life status in the OpenSSL project and there
is no need to maintain support for them in hostapd/wpa_supplicant
either.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
Use OpenSSL HMAC_* functions to implement HMAC-MD5 instead of depending
on the src/crypto/md5.c implementation.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
omac1_aes_256() and omac1_aes_vector() can now be used to perform
256-bit CMAC operations similarly to the previously supported 128-bit
cases.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This is similar with domain_suffix_match, but required a full match of
the domain name rather than allowing suffix match (subdomains) or
wildcard certificates.
Signed-off-by: Jouni Malinen <j@w1.fi>
A new "CTRL-EVENT-EAP-PEER-ALT depth=<i> <alt name>" event is now used
to provide information about server certificate chain alternative
subject names for upper layers, e.g., to make it easier to configure
constraints on the server certificate. For example:
CTRL-EVENT-EAP-PEER-ALT depth=0 DNS:server.example.com
Currently, this includes DNS, EMAIL, and URI components from the
certificates. Similar information is priovided to D-Bus Certification
signal in the new altsubject argument which is a string array of these
items.
Signed-off-by: Jouni Malinen <j@w1.fi>
This allows ocsp=2 to be used with wpa_supplicant when built with GnuTLS
to request TLS status extension (OCSP stapling) to be used to validate
server certificate validity.
Signed-off-by: Jouni Malinen <j@w1.fi>
The server certificate will be rejected if it includes any EKU and none
of the listed EKUs is either TLS Web Server Authentication or ANY.
Signed-off-by: Jouni Malinen <j@w1.fi>
Certificate expiration is checked both within GnuTLS and in the
tls_gnutls.c implementation. The former was configured to use the
request to ignore time checks while the latter was not. Complete support
for this parameter by ignoring the internal expiration checks if
requested.
Signed-off-by: Jouni Malinen <j@w1.fi>
This allows private key and client certificate to be configured using
wpa_supplicant blobs instead of external files.
Signed-off-by: Jouni Malinen <j@w1.fi>
It looks like GnuTLS may return success on
gnutls_certificate_set_x509_*() functions with GNUTLS_X509_FMT_PEM even
when trying to read DER encoded information. Reverse the order of
parsing attempts so that we start with DER and then move to PEM if
GnuTLS reports failure on DER parsing. This seems to be more reliable
way of getting errors reported and both cases can now be handled.
Signed-off-by: Jouni Malinen <j@w1.fi>
This new wpa_supplicant and hostapd control interface command can be
used to determine which TLS library is used in the build and what is the
version of that library.
Signed-off-by: Jouni Malinen <j@w1.fi>
This allows wpa_supplicant to provide more information about peer
certificate validation results to upper layers similarly to the
mechanism used with OpenSSL.
Signed-off-by: Jouni Malinen <j@w1.fi>
This implementation uses GnuTLS function
gnutls_x509_crt_check_hostname(). It has a bit different rules regarding
matching (allows wildcards in some cases, but does not use suffix
matching) compared to the internal implementation used with OpenSSL.
However, these rules are sufficiently close to each other to be of
reasonable use for most cases.
Signed-off-by: Jouni Malinen <j@w1.fi>
After having checked all known GNUTLS_CERT_* error cases that we care
about, check that no other errors have been indicated by
gnutls_certificate_verify_peers2() as a reason to reject negotiation.
Signed-off-by: Jouni Malinen <j@w1.fi>
Make the debug output more useful for determining whuch version of
GnuTLS was used and what was negotiated for the session.
Signed-off-by: Jouni Malinen <j@w1.fi>
GnuTLS 2.10.0 added gnutls_certificate_set_verify_function() that can be
used to move peer certificate validation to an earlier point in the
handshake. Use that to get similar validation behavior to what was done
with OpenSSL, i.e., reject the handshake immediately after receiving the
peer certificate rather than at the completion of handshake.
Signed-off-by: Jouni Malinen <j@w1.fi>
GnuTLS project has marked 2.12.x obsolete since January 2014. There is
not much need for maintaining support for obsolete versions of the
library, so drop all #if/#endif blocks targeting 2.x.y versions. In
practice, none of these were requiring 2.12.x version with x greater
than 0, so 2.12.x remains supported for now.
In addition, add newer version (GnuTLS 3.0.18 and newer) to fetch client
and server random from the session since the old method is not supported
by new GnuTLS versions and as such, gets removed with rest of the old
ifdef blocks.
Signed-off-by: Jouni Malinen <j@w1.fi>
No one should be using GnuTLS versions older than 1.3.2 from 2006
anymore, so remove these unnecessary #if/#endif checks.
Signed-off-by: Jouni Malinen <j@w1.fi>
This was needed with very old GnuTLS versions, but has not been needed,
or used, since GnuTLS 1.3.2 which was released in 2006. As such, there
is no need to maintain this code anymore and it is better to just clean
the source code by removing all the related code.
Signed-off-by: Jouni Malinen <j@w1.fi>
This allows GnuTLS to be used with trusted CA certificate from
wpa_supplicant blob rather than an external certificate file.
Signed-off-by: Jouni Malinen <j@w1.fi>
This TLS configuration parameter is explicitly for OpenSSL. Instead of
ignoring it silently, reject any configuration trying to use it in
builds that use other options for TLS implementation.
Signed-off-by: Jouni Malinen <j@w1.fi>
NSS as a TLS/crypto library alternative was never completed and this
barely functional code does not even build with the current NSS version.
Taken into account that there has not been much interest in working on
this crypto wrapper over the years, it is better to just remove this
code rather than try to get it into somewhat more functional state.
Signed-off-by: Jouni Malinen <j@w1.fi>
Validation of these parameters has not been implemented with schannel.
Instead of ignoring them silently, reject the configuration to avoid
giving incorrect impression of the parameters being used if
wpa_supplicant is built with schannel instead of the default OpenSSL.
Signed-off-by: Jouni Malinen <j@w1.fi>
Validation of these parameters has not been implemented in the internal
TLS implementation. Instead of ignoring them silently, reject the
configuration to avoid giving incorrect impression of the parameters
being used if wpa_supplicant is built with the internal TLS
implementation instead of the default OpenSSL.
Signed-off-by: Jouni Malinen <j@w1.fi>
Validation of these parameters has not been implemented with GnuTLS.
Instead of ignoring them silently, reject the configuration to avoid
giving incorrect impression of the parameters being used if
wpa_supplicant is built with GnuTLS instead of the default OpenSSL.
Signed-off-by: Jouni Malinen <j@w1.fi>
This was supposed to use the iterations parameter from the caller
instead of the hardcoded 4096. In practice, this did not have problems
for normal uses since that 4096 value was used in all cases.
Signed-off-by: Jouni Malinen <j@w1.fi>
This moves the AES-SIV test case from tests/test-aes.c to be part of
wpa_supplicant module testing framework with a new
src/crypto/crypto_module_tests.c component. In addition, the second test
vector from RFC 5297 is also included for additional coverage.
Signed-off-by: Jouni Malinen <j@w1.fi>
This function is not used outside aes-siv.c. In addition, include the
aes_siv.h header to make sure that functions get declared consistently.
Signed-off-by: Jouni Malinen <j@w1.fi>
It isn't mandatory. If we need one and it's not present, the ENGINE will
try asking for it. Make sure it doesn't actually let an OpenSSL UI loose,
since we don't currently capture those.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
It needs to be available to ENGINE_by_id(), which in my case means it
needs to be /usr/lib64/openssl/engines/libpkcs11.so. But that's a system
packaging issue. If it isn't there, it will fail gracefully enough with:
ENGINE: engine pkcs11 not available [error:25066067:DSO support routines:DLFCN_LOAD:could not load the shared library]
TLS: Failed to set TLS connection parameters
EAP-TLS: Failed to initialize SSL.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
This means that if the PKCS#11 engine is installed in the right place
in the system, it'll automatically be invoked by ENGINE_by_id("pkcs11")
later, and things work without explictly configuring pkcs11_engine_path.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
If these start with "pkcs11:" then they are PKCS#11 URIs. These Just Work
in the normal private_key/ca_cert/client_cert configuration fields when
built with GnuTLS; make it work that way with OpenSSL too.
(Yes, you still need to explicitly set engine=1 and point to the engine,
but I'll work on that next...)
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
There's no reason I shouldn't be able to use PKCS#11 for just the CA cert,
or even the client cert, while the private key is still from a file.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
New versions of engine_pkcs11 will automatically use the system's
p11-kit-proxy.so to make the globally-configured PKCS#11 tokens available
by default. So invoking the engine without an explicit module path is
not an error.
Older engines will fail but gracefully enough, so although it's still an
error in that case there's no need for us to catch it for ourselves.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
This converts most of the remaining perror() and printf() calls from
hostapd and wpa_supplicant to use wpa_printf().
Signed-off-by: Jouni Malinen <j@w1.fi>
It's possible to jump through hoops to support it in older versions too,
but that seems a little unnecessary at this point.
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
'_t' suffix for gnutls_session and gnutls_transport_ptr was added in
GnuTLS 1.1.11 over ten years ago and the more recent versions of GnuTLS
have started forcing compiler warnings from the old names. Move to the
new names and don't bother about backwards compatibility with older
versions taken into account how long ago this change happened in GnuTLS.
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit d4913c585e ('OpenSSL: Fix EAP-FAST
peer regression') introduced a workaround to use a new SSL_CTX instance
set for TLSv1_method() when using EAP-FAST. While that works, it is
unnecessarily complex since there is not really a need to use a separate
SSL_CTX to be able to do that. Instead, simply use SSL_set_ssl_method()
to update the ssl_method for the SSL instance. In practice, this commit
reverts most of the tls_openssl.c changes from that earlier commit and
adds that single call into tls_connection_set_params() based on EAP-FAST
flag.
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit 35efa2479f ('OpenSSL: Allow TLS
v1.1 and v1.2 to be negotiated by default') changed from using
TLSv1_method() to SSLv23_method() to allow negotiation of TLS v1.0,
v1.1, and v1.2.
Unfortunately, it looks like EAP-FAST does not work with this due to
OpenSSL not allowing ClientHello extensions to be configured with
SSL_set_session_ticket_ext() when SSLv23_method() is used. Work around
this regression by initiating a separate SSL_CTX instance for EAP-FAST
phase 1 needs with TLSv1_method() while leaving all other EAP cases
using TLS to work with the new default that allows v1.1 and v1.2 to be
negotiated. This is not ideal and will hopefully get fixed in the future
with a new OpenSSL method, but until that time, this can be used allow
other methods use newer TLS versions while still allowing EAP-FAST to be
used even if it remains to be constraint to TLS v1.0 only.
Signed-off-by: Jouni Malinen <j@w1.fi>
Commit f5fa824e9a ('Update OpenSSL 0.9.8
patch for EAP-FAST support') changed the OpenSSL 0.9.8 patch to support
the new API that was introduced in OpenSSL 1.0.0 for EAP-FAST. As such,
there should be no valid users of the old API anymore and tls_openssl.c
can be cleaned up to use only the new API.
Signed-off-by: Jouni Malinen <j@w1.fi>
Previously, it was possible for the loop through the data components to
increment addr/len index at the last position beyond the declared size.
This resulted in reading beyond those arrays. The read values were not
used and as such, this was unlikely to cause noticeable issues, but
anyway, memory checkers can detect this and the correct behavior is to
stop increments before going beyond the arrays since no more bytes will
be processed after this anyway.
Signed-off-by: Jouni Malinen <j@w1.fi>
Use SSLv23_method() to enable TLS version negotiation for any version
equal to or newer than 1.0. If the old behavior is needed as a
workaround for some broken authentication servers, it can be configured
with phase1="tls_disable_tlsv1_1=1 tls_disable_tlsv1_2=1".
Signed-off-by: Jouni Malinen <j@w1.fi>
Add an implementation of Synthetic Initialization Vector (SIV)
Authenticated Encryption Using the Advanced Encryption Standard (AES).
This mode of AES is used to protect peering frames when using
the authenticated mesh peering exchange.
Signed-off-by: Javier Lopez <jlopex@gmail.com>
Signed-off-by: Jason Mobarak <x@jason.mobarak.name>
Signed-off-by: Bob Copeland <me@bobcopeland.com>
This extends the TLS wrapper code to allow OpenSSL cipherlist string to
be configured. In addition, the default value is now set to
DEFAULT:!EXP:!LOW to ensure cipher suites with low and export encryption
algoriths (40-64 bit keys) do not get enabled in default configuration
regardless of how OpenSSL build was configured.
Signed-off-by: Jouni Malinen <j@w1.fi>
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>
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>
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: Jouni Malinen <j@w1.fi>
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: Jouni Malinen <j@w1.fi>
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: Jouni Malinen <j@w1.fi>
Instead of using X509_print_fp() to print directly to stdout, print the
certificate dump to a memory BIO and use wpa_printf() to get this into
the debug log. This allows redirection of debug log to work better and
avoids undesired stdout prints when debugging is not enabled.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
This is similar to the existing functionality that parsed ASN.1-encoded
RSA public key by generating a similar public key instance from already
parsed n and e parameters.
Signed-off-by: Jouni Malinen <j@w1.fi>
Some OpenSSL versions have vulnerability in TLS heartbeat request
processing. Check the processed message to determine if the attack has
been used and if so, do not send the response to the peer. This does not
prevent the buffer read overflow within OpenSSL, but this prevents the
attacker from receiving the information.
This change is an additional layer of protection if some yet to be
identified paths were to expose this OpenSSL vulnerability. However, the
way OpenSSL is used for EAP-TLS/TTLS/PEAP/FAST in hostapd/wpa_supplicant
was already rejecting the messages before the response goes out and as
such, this additional change is unlikely to be needed to avoid the
issue.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
test-tls-4: Short 511-bit RSA-DHE prime
test-tls-5: Short 767-bit RSA-DHE prime
test-tls-6: Bogus RSA-DHE "prime" 15
test-tls-7: Very short 58-bit RSA-DHE prime in a long container
test-tls-8: Non-prime as RSA-DHE prime
Signed-off-by: Jouni Malinen <j@w1.fi>
Previously, EAP-SIM/AKA/AKA' did not work with number of crypto
libraries (GnuTLS, CryptoAPI, NSS) since the required FIPS 186-2 PRF
function was not implemented. This resulted in somewhat confusing error
messages since the placeholder functions were silently returning an
error. Fix this by using the internal implementation of FIP 186-2 PRF
(including internal SHA-1 implementation) with crypto libraries that do
not implement this in case EAP-SIM/AKA/AKA' is included in the build.
Signed-off-by: Jouni Malinen <jouni@qca.qualcomm.com>
The internal TLS server implementation and RADIUS server implementation
in hostapd can be configured to allow EAP clients to be tested to
perform TLS validation steps correctly. This functionality is not
included in the default build; CONFIG_TESTING_OPTIONS=y in
hostapd/.config can be used to enable this.
When enabled, the RADIUS server will configure special TLS test modes
based on the received User-Name attribute value in this format:
<user>@test-tls-<id>.<rest-of-realm>. For example,
anonymous@test-tls-1.example.com. When this special format is used, TLS
test modes are enabled. For other cases, the RADIUS server works
normally.
The following TLS test cases are enabled in this commit:
1 - break verify_data in the server Finished message
2 - break signed_params hash in ServerKeyExchange
3 - break Signature in ServerKeyExchange
Correctly behaving TLS client must abort connection if any of these
failures is detected and as such, shall not transmit continue the
session.
Signed-off-by: Jouni Malinen <j@w1.fi>
This allows the internal TLS implementation to write log entries to the
same authlog with rest of the RADIUS server and EAP server
functionality.
Signed-off-by: Jouni Malinen <j@w1.fi>
These can be used to disable TLSv1.1 and TLSv1.2 as a workaround for AAA
servers that have issues interoperating with newer TLS versions.
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
This reverts commit 51e3eafb68. There are
too many deployed AAA servers that include both id-kp-clientAuth and
id-kp-serverAuth EKUs for this change to be acceptable as a generic rule
for AAA authentication server validation. OpenSSL enforces the policy of
not connecting if only id-kp-clientAuth is included. If a valid EKU is
listed with it, the connection needs to be accepted.
Signed-off-by: Jouni Malinen <j@w1.fi>
This fixes issues in using a password that includes a UTF-8 character
with three-byte encoding with EAP methods that use NtPasswordHash
(anything using MSCHAPv2 or LEAP).
Signed-off-by: Jouni Malinen <j@w1.fi>
In function tls_verify_cb(), X509_STORE_CTX_get_current_cert() may
return NULL, and it will be dereferenced by X509_get_subject_name().
Signed-hostap: Eytan Lifshitz <eytan.lifshitz@intel.com>
These were somewhat more hidden to avoid direct use, but there are now
numerous places where these are needed and more justification to make
the extern int declarations available from wpa_debug.h. In addition,
this avoids some warnings from sparse.
Signed-hostap: Jouni Malinen <j@w1.fi>
If the extended key usage of the AAA server certificate indicates
that the certificate is for client use, reject the TLS handshake.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
OCSP response may not include all the needed CA certificates, so use the
ones received during TLS handshake.
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
It's not possible to get a raw private key from keystore anymore, so
this would fail every time anyway. Remove it so it doesn't confuse
anyone that looks at this code.
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
The new keystore ENGINE is usable to perform private key operations when
we can't get the actual private key data. This is the case when hardware
crypto is enabled: the private key never leaves the hardware.
Subsequently, we need to be able to talk to OpenSSL ENGINEs that aren't
PKCS#11 or OpenSC. This just changes a few #define variables to allow us
to talk to our keystore engine without having one of those enabled and
without using a PIN.
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
If SSL_CTX_new() fails in tls_init(), the per-SSL app-data allocation
could have been leaked when multiple TLS instances are allocated.
Signed-hostap: Jouni Malinen <j@w1.fi>
The new domain_suffix_match (and domain_suffix_match2 for Phase 2
EAP-TLS) can now be used to specify an additional constraint for the
server certificate domain name. If set, one of the dNSName values (or if
no dNSName is present, one of the commonName values) in the certificate
must have a suffix match with the specified value. Suffix match is done
based on full domain name labels, i.e., "example.com" matches
"test.example.com" but not "test-example.com".
Signed-hostap: Jouni Malinen <jouni@qca.qualcomm.com>
Check that SSL_clear_options and SSL_CTX_clear_options are defined
before using them to avoid compilation failures with older OpenSSL
versions that did not include these macros.
Signed-hostap: Jouni Malinen <j@w1.fi>
When using OpenSSL with TLS-based EAP methods, wpa_supplicant can now be
configured to use OCSP stapling (TLS certificate status request) with
ocsp=1 network block parameter. ocsp=2 can be used to require valid OCSP
response before connection is allowed to continue.
hostapd as EAP server can be configured to return cached OCSP response
using the new ocsp_stapling_response parameter and an external mechanism
for updating the response data (e.g., "openssl ocsp ..." command).
This allows wpa_supplicant to verify that the server certificate has not
been revoked as part of the EAP-TLS/PEAP/TTLS/FAST handshake before
actual data connection has been established (i.e., when a CRL could not
be fetched even if a distribution point were specified).
Signed-hostap: Jouni Malinen <j@w1.fi>
Store context for each tls_init() caller, so events are generated for
the correct wpa_s instance. The tls_global variable is retained for
older OpenSSL implementations that may not have app-data for SSL_CTX.
Signed-hostap: Paul Stewart <pstew@chromium.org>
Move the bignum comparison part into the bignum library to allow a
single implementation of rand generation for both ECC and FCC based
groups.
Signed-hostap: Jouni Malinen <j@w1.fi>
The groups 22, 23, and 24 are not based on a safe prime and generate a
prime order subgroup. As such, struct dh_group is also extended to
include the order for previously defined groups (q=(p-1)/2 since these
were based on a safe prime).
Signed-hostap: Jouni Malinen <j@w1.fi>
In addition to the trivial change in adding the new group ientifier,
this required changes to KDF and random number generation to support
cases where the length of the prime in bits is not a multiple of eight.
The binary presentation of the value needs to be shifted so that the
unused most significant bits are the zero padding rather than the extra
bits in the end of the array.
Signed-hostap: Jouni Malinen <j@w1.fi>
In addition to the mandatory group 19 (256-bit random ECP group) add
support for groups 20 (384-bit), 25 (192-bit), and 26 (224-bit).
Signed-hostap: Jouni Malinen <j@w1.fi>
This makes the SAE implementation a bit simpler by not having to build
the bignum for group order during execution.
Signed-hostap: Jouni Malinen <j@w1.fi>
This is a generic AES CCM implementation that can be used for other
purposes than just implementing CCMP, so it fits better in a separate
file in src/crypto.
Signed-hostap: Jouni Malinen <j@w1.fi>
AES uses the same 128-bit block size with 128, 192, 256 bit keys, so use
the fixed block size definition instead of trying to dynamically set the
block size based on key length. This fixes use of 192-bit and 256-bit
AES keys with crypto_cipher_*() API when using the internal AES
implementation.
Signed-hostap: Jouni Malinen <j@w1.fi>
This is otherwise identical to aes_gcm_ae() but does not use the
plain/crypt pointers since no data is encrypted.
Signed-hostap: Jouni Malinen <j@w1.fi>
This adds 192-bit and 256-bit key support to the internal AES
implementation and extends the AES-GCM functions to accept key length to
enable longer AES key use.
Signed-hostap: Jouni Malinen <j@w1.fi>
This is a generic AES GCM and GMAC implementation that can be used for
other purposes than just implementing GCMP, so it fits better in a
separate file in src/crypto.
Signed-hostap: Jouni Malinen <j@w1.fi>
Explicitly validate seed_len to skip memset call with zero length of
copied data at the end of the buffer. This is not really needed, but it
makes the code a bit easier for static analyzers. This is identical to
the commit a9ea17491a but for the OpenSSL
version of the function.
Signed-hostap: Jouni Malinen <j@w1.fi>
Commit c9e08af24f removed the only user of
the special case MD5 use that would be allowed in FIPS mode in
tls_prf_sha1_md5(). Commit 271dbf1594
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.
Signed-hostap: Jouni Malinen <j@w1.fi>
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.
Signed-hostap: Jouni Malinen <j@w1.fi>
intended-for: hostap-1
This can be used to implement workaround for authentication servers that
do not handle TLS extensions in ClientHello properly.
Signed-hostap: Jouni Malinen <j@w1.fi>
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.
Signed-hostap: Jouni Malinen <j@w1.fi>
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.
Signed-hostap: Jouni Malinen <j@w1.fi>
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.
Signed-hostap: Jouni Malinen <j@w1.fi>
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().
Signed-hostap: Jouni Malinen <j@w1.fi>
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.
Signed-hostap: Jouni Malinen <j@w1.fi>
This bit is set in the code path that handles keys and certs from places
other than OpenSSL authentication engines. Setting this bit causes
authentication to fail when the server provides certificates that don't
match the client certificate authority.
Commit d9cc4646eb added
crypto_hash_{init,update,finish}() wrappers for OpenSSL, but it
assumed the current HMAC API in OpenSSL. This was changed in 0.9.9
to return error codes from the functions while older versions used
void functions. Add support for the older versions, too.
Signed-hostap: Jouni Malinen <j@w1.fi>
Send an "EAP" signal via the new DBus interface under various
conditions during EAP authentication:
- During method selection (ACK and NAK)
- During certificate verification
- While sending and receiving TLS alert messages
- EAP success and failure messages
This provides DBus callers a number of new tools:
- The ability to probe an AP for available EAP methods
(given an identity).
- The ability to identify why the remote certificate was
not verified.
- The ability to identify why the remote peer refused
a TLS connection.
Signed-hostap: Paul Stewart <pstew@chromium.org>
Mark the debug print excessive and print it only in case the entropy
collection is used since this function can get called pretty frequently.
Signed-hostap: Jouni Malinen <j@w1.fi>
Prepare for multiple TLS PRF functions by renaming the SHA1+MD5 based
TLS PRF function to more specific name and add tls_prf() within the
internal TLS implementation as a wrapper for this for now.
Signed-hostap: Jouni Malinen <j@w1.fi>
Currently OpenSSL implementation of TLS in hostapd loads only top
certificate in server certificate file. Change this to try to the
whole chain first and only if that fails, revert to old behavior.
Signed-off-by: Maciej Szmigiero <mhej@o2.pl>
Avoid zero-length memset at the end of the buffer. This is not really
needed, but it makes the code a bit easier for static analyzers.
Signed-hostap: Jouni Malinen <j@w1.fi>
Avoid zero-length memset at the end of the buffer. This is not really
needed, but it makes the code a bit easier for static analyzers.
Signed-hostap: Jouni Malinen <j@w1.fi>
Explicitly validate seed_len to skip memset call with zero length
of copied data at the end of the buffer. This is not really needed,
but it makes the code a bit easier for static analyzers.
Signed-hostap: Jouni Malinen <j@w1.fi>
md->curlen cannot indicate full buffer size here since the buffered
data is processed whenever the full block size of data is available.
Avoid invalid warnings from static analyzers on memcpy() outside the
buffer length by verifying that curlen is smaller than block size.
Signed-hostap: Jouni Malinen <j@w1.fi>
Reassemble partial TLS records to make the internal TLS client
implementation more convenient for stream sockets.
Signed-hostap: Jouni Malinen <j@w1.fi>
The MS-CHAPv1 and MS-CHAPv2 RFCs specify that the password is a string
of "Unicode characters", which for Windows means UCS-2; thus the
password could be any even-length string of up to 512 bytes.
Instead of making the incompatible change of requiring the incoming
password to be UCS-2 encoded, assume the password is UTF-8 encoded and
convert it before using it in NtPasswordHash and
EncryptPwBlockWithPasswordHash.
Signed-off-by: Evan Broder <ebroder@mokafive.com>
These protocols seem to be abandoned: latest IETF drafts have expired
years ago and it does not seem likely that EAP-TTLSv1 would be
deployed. The implementation in hostapd/wpa_supplicant was not complete
and not fully tested. In addition, the TLS/IA functionality was only
available when GnuTLS was used. Since GnuTLS removed this functionality
in 3.0.0, there is no available TLS/IA implementation in the latest
version of any supported TLS library.
Remove the EAP-TTLSv1 and TLS/IA implementation to clean up unwanted
complexity from hostapd and wpa_supplicant. In addition, this removes
any potential use of the GnuTLS extra library.
This fixes some build issues in GnuTLS wrapper to be compatible with
at least following GnuTLS versions: 2.2.5, 2.4.3, 2.6.6, 2.8.6,
2.10.5, 2.12.11, 3.0.3.
eapol_test command line argument -o<file> can now be used to request
the received server certificate chain to be written to the specified
file. The certificates will be written in PEM format. [Bug 391]
Some compilers complain about fwrite calls if the return value is
not checked, so check the value even if it does not really make
much of a difference in this particular case.
This phase1 parameter for TLS-based EAP methods was already supported
with GnuTLS and this commit extends that support for OpenSSL and the
internal TLS implementation.
This can be used to avoid rejection of first two 4-way handshakes every
time hostapd (or wpa_supplicant in AP/IBSS mode) is restarted. A new
command line parameter, -e, can now be used to specify an entropy file
that will be used to maintain the needed state.
This allows keystore:// prefix to be used with client_cert and
private_key configuration parameters.
Signed-off-by: Dmitry Shmidt <dimitrysh@google.com>
If parsing of the certificate or private key succeeds using any of
the tried encoding types, clear the OpenSSL error queue without
showing the pending errors in debug log since they do not really
provide any useful output and can be confusing.
On Linux, verify that the kernel entropy pool is capable of providing
strong random data before allowing WPA/WPA2 connection to be
established. If 20 bytes of data cannot be read from /dev/random,
force first two 4-way handshakes to fail while collecting entropy
into the internal pool in hostapd. After that, give up on /dev/random
and allow the AP to function based on the combination of /dev/urandom
and whatever data has been collected into the internal entropy pool.
By default, make hostapd and wpa_supplicant maintain an internal
entropy pool that is fed with following information:
hostapd:
- Probe Request frames (timing, RSSI)
- Association events (timing)
- SNonce from Supplicants
wpa_supplicant:
- Scan results (timing, signal/noise)
- Association events (timing)
The internal pool is used to augment the random numbers generated
with the OS mechanism (os_get_random()). While the internal
implementation is not expected to be very strong due to limited
amount of generic (non-platform specific) information to feed the
pool, this may strengthen key derivation on some devices that are
not configured to provide strong random numbers through
os_get_random() (e.g., /dev/urandom on Linux/BSD).
This new mechanism is not supposed to replace proper OS provided
random number generation mechanism. The OS mechanism needs to be
initialized properly (e.g., hw random number generator,
maintaining entropy pool over reboots, etc.) for any of the
security assumptions to hold.
If the os_get_random() is known to provide strong ramdom data (e.g., on
Linux/BSD, the board in question is known to have reliable source of
random data from /dev/urandom), the internal hostapd random pool can be
disabled. This will save some in binary size and CPU use. However, this
should only be considered for builds that are known to be used on
devices that meet the requirements described above. The internal pool
is disabled by adding CONFIG_NO_RANDOM_POOL=y to the .config file.
This commit adds a new wrapper, random_get_bytes(), that is currently
defined to use os_get_random() as is. The places using
random_get_bytes() depend on the returned value being strong random
number, i.e., something that is infeasible for external device to
figure out. These values are used either directly as a key or as
nonces/challenges that are used as input for key derivation or
authentication.
The remaining direct uses of os_get_random() do not need as strong
random numbers to function correctly.
The length of the prime was used incorrectly and this resulted
in WPS DH operation failing whenever the public key ended up having
leading zeros (i.e., about every 1/256th time).
There are no subdirectories in any of these directories or plans
for adding ones. As such, there is no point in running the loop
that does not do anything and can cause problems with some shells.
The returned buffer length was hardcoded to be the prime length
which resulted in shorter results being padded in the end. However,
the results from DH code are supposed to be unpadded (and when used
with WPS, the padding is done in WPS code and it is added to the
beginning of the buffer). This fixes WPS key derivation errors
in about 1/256 of runs ("WPS: Incorrect Authenticator") when using
the internal crypto code.
This allows external programs (e.g., UI) to get more information
about server certificate chain used during TLS handshake. This can
be used both to automatically probe the authentication server to
figure out most likely network configuration and to get information
about reasons for failed authentications.
The follow new control interface events are used for this:
CTRL-EVENT-EAP-PEER-CERT
CTRL-EVENT-EAP-TLS-CERT-ERROR
In addition, there is now an option for matching the server certificate
instead of the full certificate chain for cases where a trusted CA is
not configured or even known. This can be used, e.g., by first probing
the network and learning the server certificate hash based on the new
events and then adding a network configuration with the server
certificate hash after user have accepted it. Future connections will
then be allowed as long as the same server certificate is used.
Authentication server probing can be done, e.g., with following
configuration options:
eap=TTLS PEAP TLS
identity=""
ca_cert="probe://"
Example set of control events for this:
CTRL-EVENT-EAP-STARTED EAP authentication started
CTRL-EVENT-EAP-PROPOSED-METHOD vendor=0 method=21
CTRL-EVENT-EAP-METHOD EAP vendor 0 method 21 (TTLS) selected
CTRL-EVENT-EAP-PEER-CERT depth=0 subject='/C=US/ST=California/L=San Francisco/CN=Server/emailAddress=server@kir.nu' hash=5a1bc1296205e6fdbe3979728efe3920798885c1c4590b5f90f43222d239ca6a
CTRL-EVENT-EAP-TLS-CERT-ERROR reason=8 depth=0 subject='/C=US/ST=California/L=San Francisco/CN=Server/emailAddress=server@kir.nu' err='Server certificate chain probe'
CTRL-EVENT-EAP-FAILURE EAP authentication failed
Server certificate matching is configured with ca_cert, e.g.:
ca_cert="hash://server/sha256/5a1bc1296205e6fdbe3979728efe3920798885c1c4590b5f90f43222d239ca6a"
This functionality is currently available only with OpenSSL. Other
TLS libraries (including internal implementation) may be added in
the future.
Undocumented (at least for the time being) TLS parameters can now
be provided in wpa_supplicant configuration to enable some workarounds
for being able to connect insecurely to some networks. phase1 and
phase2 network parameters can use following options:
tls_allow_md5=1
- allow MD5 signature to be used (disabled by default with GnuTLS)
tls_disable_time_checks=1
- ignore certificate expiration time
For now, only the GnuTLS TLS wrapper implements support for these.
This converts tls_connection_handshake(),
tls_connection_server_handshake(), tls_connection_encrypt(), and
tls_connection_decrypt() to use struct wpa_buf to allow higher layer
code to be cleaned up with consistent struct wpabuf use.
This message from tls_connection_handshake() is not really an error in
most cases, so do not show it if there was indeed no Application Data
available (which is a normal scenario and not an indication of any
error).
This allows libeap.a and libeap.so to be built by merging in multiple
libraries from src subdirectories. In addition, this avoids wasting
extra space and time for local builds.
The following defines are not really needed in most places, so
remove them to clean up source code and build scripts:
EAP_TLS_FUNCS
EAP_TLS_OPENSSL
EAP_TLS_GNUTLS
CONFIG_TLS_INTERNAL
Mainly, this is including header files to get definitions for functions
which is good to verify that the parameters match. None of these are
issues that would have shown as incorrect behavior of the program.
The current MinGW/w32api versions seem to provide all the needed CryptoAPI
functions, so the code for loading these dynamically from the DLL can be
removed.
Need to define the workspace buffer properly to allow compiler to handle
strict aliasing between the incoming unsigned char[64] buffer as an u32
array. The previous version built with strict aliasing enabled can
result in SHA-1 producing incorrect results and consequently, with
4-way handshake failing.
This is based on a report and patch from Dan Williams <dcbw@redhat.com>
but with a different type (the union) used as a fix to avoid needing
extra type casting.
Discovered as part of the investigation of:
https://bugzilla.redhat.com/show_bug.cgi?id=494262#c32
if sha1 is built with gcc without turning off strict aliasing, it will
fail to correctly generate the hashes and will fail its own testcases as
well.
Signed-off-by: Dan Williams <dcbw@redhat.com>
This functionality fits better with src/tls (i.e., internal TLS
implementation), so move it there to make crypto_internal.c more
of a wrapper like other crypto_*.c files.
Private keys can now be used in either unencrypted or encrypted
PKCS #8 encoding. Only the pbeWithMD5AndDES-CBC algorithm (PKCS #5)
is currently supported.
OpenSSL 0.9.7 does not include get_rfc3526_prime_1536() function, so
provide that functionality internally if needed. In addition, make
sha256_vector() building depend on whether SHA256 support is included
in the OpenSSL library. This with CONFIG_INTERNAL_SHA256=y in .config
allows OpenSSL without SHA256 support to be used.
Reorganize the TLS/crypto library segments into a single set of blocks
for each library instead of multiple locations handling library-specific
operations. Group crypto functionality together and get wpa_supplicant
and hostapd Makefile closer to eachother in order to make it easier to
eventually move this into a shared makefile.
Crypto library wrappers can now override the internal DH (group 5)
implementation. As a starting point, this is done with OpenSSL. The
new mechanism is currently available only for WPS (i.e., IKEv2 still
depends on the internal DH implementation).
This allows NSS to be used to derive EAP-TLS/PEAP/TTLS keying material.
NSS requires a patch from
https://bugzilla.mozilla.org/show_bug.cgi?id=507359
to provide the new API. In addition, that patch needs to be modified to
add the 16-bit context length value in SSL_ExportKeyingMaterial() only if
contextlen != 0 in order to match with the EAP-TLS/PEAP/TTLS use cases.
This issue seems to be coming from the unfortunate incompatibility in
draft-ietf-tls-extractor-07.txt (draft-ietf-tls-extractor-00.txt would
have used compatible PRF construction).
At this point, it is unclear how this will be resolved eventually, but
anyway, this shows a mechanism that can be used to implement EAP key
derivation with NSS with a small patch to NSS.
This brings in the first step in adding support for using NSS
(Mozilla Network Security Services) as the crypto and TLS library
with wpa_supplicant. This version is able to run through EAP-PEAP
and EAP-TTLS authentication, but does not yet implement any
certificate/private key configuration. In addition, this does not
implement proper key fetching functions either, so the end result
is not really of much use in real world yet.
The BLOCK_SIZE define can be made more specific by using AES_ prefix and
by moving it to aes.h. After this, most aes-*.c do not really need to
include anything from the internal aes_i.h header file. In other words,
aes_i.h can now be used only for the code that uses the internal AES
block operation implementation and none of the code that can use AES
implementation from an external library do not need to include this
header file.
Better not specify EVP_CIPHER again for the second init call since that
will override key length with the default value. The previous version
was likely to work since most use cases would be likely to use the
default key length. Anyway, better make this handle variable length
ciphers (mainly, RC4), too, just in case it is needed in the future.
This is not really of that much use since rc4_skip() can be used as
easily. In addition, rc4 has caused some symbol conflicts in the past,
so it is easier to live without that as an exported symbol.
wpa_supplicant can now be built with FIPS capable OpenSSL for FIPS mode
operation. Currently, this is only enabling the FIPS mode in OpenSSL
without providing any higher level enforcement in wpa_supplicant.
Consequently, invalid configuration will fail during the authentication
run. Proper configuration (e.g., WPA2-Enterprise with EAP-TLS) allows
the connection to be completed.
Instead of using low level, digest-specific functions, use the generic
EVP interface for digest functions. In addition, report OpenSSL errors
in more detail.
Some crypto libraries can return in these functions (e.g., if a specific
hash function is disabled), so we better provide the caller a chance to
check whether the call failed. The return values are not yet used
anywhere, but they will be needed for future changes.
This removes need for local configuration to ignore *.o and *~
and allows the src/*/.gitignore files to be removed (subdirectories
will inherit the rules from the root .gitignore).
It looks like GnuTLS (at least newer versions) is using random padding
on the application data and the previously used 100 byte extra buffer
for tls_connection_encrypt() calls was not enough to handle all cases.
This resulted in semi-random authentication failures with EAP-PEAP and
EAP-TTLS during Phase 2.
Increase the extra space for encryption from 100 to 300 bytes and add an
error message into tls_gnutls.c to make it easier to notice this issue
should it ever show up again even with the larger buffer.
Updated OpenSSL code for EAP-FAST to use an updated version of the
session ticket overriding API that was included into the upstream
OpenSSL 0.9.9 tree on 2008-11-15 (no additional OpenSSL patch is
needed with that version anymore).
When the TLS handshake had been completed earlier by the server in case of
abbreviated handshake, the output buffer length was left uninitialized. It
must be initialized to zero in this case. This code is used by EAP-FAST
server and the uninitialized length could have caused it to try to send a
very large frame (though, this would be terminated by the 50 roundtrip EAP
limit). This broke EAP-FAST server code in some cases when PAC was used to
establish the tunnel.
The internal TLS implementation can now use both PKCS #1 RSA private key
and PKCS #8 encapsulated RSA private key. PKCS #8 encrypted private key is
not yet supported.
The server handshake processing was still using SSL_read() to get OpenSSL
to perform the handshake. While this works for most cases, it caused some
issues for re-authentication. This is now changed to use SSL_accept() which
is more approriate here since we know that the handshake is still going on
and there will not be any tunneled data available. This resolves some of
the re-authentication issues and makes it possible for the server to notice
if TLS processing fails (SSL_read() did not return an error in many of
these cases while SSL_accept() does).
Set session id context to a unique value in order to avoid fatal errors
when client tries session resumption (SSL_set_session_id_context() must be
called for that to work), but disable session resumption with the unique
value for the time being since not all server side code is ready for it yet
(e.g., EAP-TTLS needs special Phase 2 processing when using abbreviated
handshake).
Changed EAP-TLS server not to call TLS library when processing the final
ACK (empty data) from the client in order to avoid starting a new TLS
handshake with SSL_accept().
It looks like this SSL_set_options() value was added in 0.9.9 and it does
not exist in stable releases of OpenSSL. Fix build by using #ifdef on this
variable before use.
The middle byte of the secret (key for PRF) is shared with key halfs in
case the key length is odd. This does not happen in any of the current
tls_prf() uses, but it's better to fix this function to avoid future issues
should someone end up defining a use that uses an odd length for the key.
