This uncovered a bug that led to a double-free (in practice, in general could
be free() on any invalid value): initially the session structure is loaded
with `memcpy()` which copies the previous values of pointers peer_cert and
ticket to heap-allocated buffers (or any other value if the input is
attacker-controlled). Now if we exit before we got a chance to replace those
invalid values with valid ones (for example because the input buffer is too
small, or because the second malloc() failed), then the next call to
session_free() is going to call free() on invalid pointers.
This bug is fixed in this commit by always setting the pointers to NULL right
after they've been read from the serialised state, so that the invalid values
can never be used.
(An alternative would be to NULL-ify them when writing, which was rejected
mostly because we need to do it when reading anyway (as the consequences of
free(invalid) are too severe to take any risk), so doing it when writing as
well is redundant and a waste of code size.)
Also, while thinking about what happens in case of errors, it became apparent
to me that it was bad practice to leave the session structure in an
half-initialised state and rely on the caller to call session_free(), so this
commit also ensures we always clear the structure when loading failed.
This test appeared to be passing for the wrong reason, it's actually not
appropriate for the current implementation. The serialised data contains
values of pointers to heap-allocated buffers. There is no reason these should
be identical after a load-save pair. They just happened to be identical when I
first ran the test due to the place of session_free() in the test code and the
fact that the libc's malloc() reused the same buffers. The test no longer
passes if other malloc() implementations are used (for example, when compiling
with asan which avoids re-using the buffer, probably for better error
detection).
So, disable this test for now (we can re-enable it when we changed how
sessions are serialised, which will be done in a future PR, hence the name of
the dummy macro in depends_on). In the next commit we're going to add a test
that save-load is the identity instead - which will be more work in testing as
it will require checking each field manually, but at least is reliable.
This initial test ensures that a load-save function is the identity. It is so
far incomplete in that it only tests sessions without tickets or certificate.
This will be improved in the next commits.
When MBEDTLS_SSL_ENCRYPT_THEN_MAC is enabled, but not
MBEDTLS_SSL_SOME_MODES_USE_MAC, mbedtls_ssl_derive_keys() and
build_transforms() will attempt to use a non-existent `encrypt_then_mac`
field in the ssl_transform.
Compile [ 93.7%]: ssl_tls.c
[Error] ssl_tls.c@865,14: 'mbedtls_ssl_transform {aka struct mbedtls_ssl_transform}' ha
s no member named 'encrypt_then_mac'
[ERROR] ./mbed-os/features/mbedtls/src/ssl_tls.c: In function 'mbedtls_ssl_derive_keys'
:
./mbed-os/features/mbedtls/src/ssl_tls.c:865:14: error: 'mbedtls_ssl_transform {aka str
uct mbedtls_ssl_transform}' has no member named 'encrypt_then_mac'
transform->encrypt_then_mac = session->encrypt_then_mac;
^~
Change mbedtls_ssl_derive_keys() and build_transforms() to only access
`encrypt_then_mac` if `encrypt_then_mac` is actually present.
Add a regression test to detect when we have regressions with
configurations that do not include any MAC ciphersuites.
Fixes d56ed2491b ("Reduce size of `ssl_transform` if no MAC ciphersuite is enabled")
Part of the record encryption/decryption tests is to gradually
increase the space available at the front and/or at the back of
a record and observe when encryption starts to succeed. If exactly
one of the two parameters is varied at a time, the expectation is
that encryption will continue to succeed once it has started
succeeding (that's not true if both pre- and post-space are varied
at the same time).
Moreover, previously the test would take turns when choosing which
transform should be used for encryption, and which for decryption.
With the introduction of the CID feaature, this switching of transforms
doesn't align with the expectation of eventual success of the encryption,
since the overhead of encryption might be different for the parties,
because both parties may use different CIDs for their outgoing records.
This commit modifies the tests to not take turns between transforms,
but to always use the same transforms for encryption and decryption
during a single round of the test.
This commit adds tests to check the behavior of the record encryption
routine `ssl_encrypt_buf` when the buffer surrounding the plaintext is
too small to hold the expansion in the beginning and end (due to IV's,
padding, and MAC).
Each test starts successively increases the space available at the
beginning, end, or both, of the record buffer, and checks that the
record encryption either fails with a BUFFER_TOO_SMALL error, or
that it succeeds. Moreover, if it succeeds, it is checked that
decryption succeeds, too, and results in the original record.
This commit adds tests exercising mutually inverse pairs of
record encryption and decryption transformations for the various
transformation types allowed in TLS: Stream, CBC, and AEAD.
Add a test case calling ssl_set_hostname twice to test_suite_ssl.
When run in CMake build mode ASan, this catches the current leak,
but will hopefully be fine with the new version.
