In some cases, we were calling `mbedtls_test_ssl_endpoint_free()` on an
uninitialized `mbedtls_test_ssl_endpoint` object if the test case failed
early, e.g. due to `psa_crypto_init()` failing. This was largely harmless,
but could have caused weird test results in case of failure, and was flagged
by Coverity.
Use a more systematic style for initializing the stack object as soon as
it's declared.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
There was a discrepancy between how `mbedtls_test_ssl_endpoint_init()` and
`mbedtls_test_ssl_perform_handshake()` handled client authentication:
`mbedtls_test_ssl_endpoint_init()` defaulted to
`MBEDTLS_SSL_VERIFY_REQUIRED` on both sides, whereas
`mbedtls_test_ssl_perform_handshake()` obeyed `options->srv_auth_mode` which
defaulted to no verification of the client certificate.
Make this more uniform. Now `mbedtls_test_ssl_endpoint_init()` obeys
`options->srv_auth_mode` on servers (still forcing verification on clients,
which is the library default anyway). Also, `options->srv_auth_mode` is now
enabled by default. Thus:
* Tests that call `mbedtls_test_ssl_perform_handshake()` now perform client
certificate verification, unless they disable it explicitly.
* Tests that call `mbedtls_test_ssl_endpoint_init()` on a server are
unchanged. (They would change if they were setting
`options->srv_auth_mode` explicitly, which previously was ignored, but
no test function did this.)
This means that a few test functions now perform client certificate
verification whereas they previously don't. This is harmless except in
`handshake_ciphersuite_select`, where one test case
`Handshake, select ECDH-RSA-WITH-AES-256-CBC-SHA384, opaque` fails with
client authentication because the test code doesn't deal with the weirdness
of static ECDH correctly with respect to client authentication. So keep
the previous behavior in `handshake_ciphersuite_select`, by explicitly
turning off client authentication.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
The DTLS context and the queues now conveyed inside the endpoint object.
Remove the unused parameters.
No behavior change.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
We were not making enough room. We want to move everything from the
place where we are going to insert the new record.
This was not causing failures because the code does not look at the
content after the inserted record, because it correctly returns an error
when seeing the inserted record. But as a matter on principle, the test
code should be doing what it says: just insert a new record but leave
a valid fragment after it.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
These are not optimal. For example, the tests should in principle be
able to run in builds without ECDSA, by using RSA certs instead. Ideally
PSK should work too.
However, getting optimal dependencies would be a lot of work that's
largely orthogonal to the purpose of this PR, so we'll settle for good
enough.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
Simulate the server closing the connection after a partial handshake
message.
These test cases don't send a close_notify alert. The test cases
"insert alert record" exercise what happens if the server sends an alert.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Create tests that coalesce the handshake messages in the first flight from
the server. This lets us test the behavior of the library when a handshake
record contains multiple handshake messages.
Only non-protected (non-encrypted, non-authenticated) handshake messages are
supported.
The test code works for all protocol versions, but it is only effective in
TLS 1.2. In TLS 1.3, there is only a single non-encrypted handshake record,
so we can't test records containing more than one handshake message without
a lot more work.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This is more flexible: the test data gets to decide whether we want to
assert the presence of a pattern or not.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
This should avoid running into a bug with printf format specifiers one
windows.
It's also a logical move for actual tests: I used the highest debug
level for discovery, but we don't need that all the time.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
Declare the same dependencies as for the previous TLS 1.3 tests, except
for part that varies with the cipher suite (ie AES-GCM).
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
This guards the definition of mbedtls_test_ssl_endpoint which we rely
on, so the function won't compile without it.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
Pre-existing but not having TLS 1.3 in the build does not seem to be
necessary actually. These test functions set the dtls flag when
calling `test_resize_buffers` and then `test_resize_buffers` sets the
`options.dtls` flag which eventually forces the TLS 1.2 version of the
protocol (in `mbedtls_test_ssl_endpoint_init()` call of
`mbedtls_ssl_config_defaults()` with `MBEDTLS_SSL_TRANSPORT_DATAGRAM`
as the transport).
Signed-off-by: Gabor Mezei <gabor.mezei@arm.com>
