We previously introduced a safety check ensuring that if a datagram had
already been dropped twice, it would no longer be dropped or delayed
after that.
This missed an edge case: if a datagram is dropped once, it can be
delayed any number of times. Since "delay" is not defined in terms of
time (x seconds) but in terms of ordering with respect to other messages
(will be forwarded after the next message is forwarded), depending on
the RNG results this could result in an endless loop where all messages
are delayed until the next, which is itself delayed, etc. and no message
is ever forwarded.
The probability of this happening n times in a row is (1/d)^n, where d
is the value passed as delay=d, so for delay=5 and n=5 it's around 0.03%
which seems small but we still happened on such an occurrence in real
life:
tests/ssl-opt.sh --seed 1625061502 -f 'DTLS proxy: 3d, min handshake, resumption$'
results (according to debug statements added for the investigation) in
the ClientHello of the second handshake being dropped once then delayed
5 times, after which the client stops re-trying and the test fails for
no interesting reason.
Make sure this doesn't happen again by putting a cap on the number of
times we fail to forward a given datagram immediately.
Signed-off-by: Manuel Pégourié-Gonnard <manuel.pegourie-gonnard@arm.com>
The sample program aescrypt2 shows bad practice: hand-rolled CBC
implementation, CBC+HMAC for AEAD, hand-rolled iterated SHA-2 for key
stretching, no algorithm agility. The new sample program pbcrypt does
the same thing, but better. So remove aescrypt2.
Fix#1906
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
According to the design in psa-driver-interface.md. Compiles without
issue in test_psa_crypto_drivers.
Signed-off-by: Steven Cooreman <steven.cooreman@silabs.com>
Call it “SHAKE256-512”, just like SHA3-512 has 512 bits of output.
SHAKE256-64 looks like it's 64 bits of output, but this is 64 bytes.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Define algorithms for PureEdDSA and for HashEdDSA, the EdDSA variants
defined by RFC 8032.
The encoding for HashEdDSA needs to encode the hash algorithm so that
the hash can be calculated by passing PSA_ALG_SIGN_GET_HASH(sig_alg)
to psa_hash_compute() or psa_hash_setup(). As a consequence,
Ed25519ph (using SHA-512) and Ed448ph (using SHAKE256) need to have
different algorithm encodings (the key is enough to tell them apart,
but it is not known while hashing). Another consequence is that the
API needs to recognize the Ed448 prehash (64 bytes of SHAKE256 output)
as a hash algorithm.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Add an elliptic curve family for the twisted Edwards curves
Edwards25519 and Edwards448 ("Goldilocks"). As with Montgomery curves,
since these are the only two curves in common use, the family has a
generic name.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
The PSA subsystem may consume global resources. It currently doesn't
consume any heap when no keys are registered, but it may do so in the
future. It does consume mutexes, which are reported as leaks when
mutex usage checking is enabled.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Test hook failure checks may print information to stdout, which messes
up the usage of query_config mode. Nothing interesting happens in
query_config mode anyway, so that's no loss.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Create utility functions to set up test hooks and report errors that
the test hooks might detect. Call them in ssl_client2 and ssl_server2.
Test hooks are potentially enabled by compiling with
MBEDTLS_TEST_HOOKS.
This commit only sets up the functions. It doesn't make them do
anything yet.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Part of build_default_make_gcc_and_cxx compares the list of headers
included by `programs/test/cpp_dummy_build.cpp` and the actual headers
present. Add in the missing `mbedtls/psa_config.h` file to this list so
that this test passes.
Signed-off-by: David Brown <david.brown@linaro.org>
The SSL test programs can now use the PSA RNG, and the PSA RNG can use
an external RNG. The build conditions hadn't been updated and didn't
cover the case when MBEDTLS_TEST_USE_PSA_CRYPTO_RNG is enabled but
MBEDTLS_USE_PSA_CRYPTO is disabled. Fix this.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Move the call to destroy the PSK to before freeing the SSL session
data and calling rng_free(), which deinitializes the PSA subsystem.
This particular ordering was chosen to make the ssl_client2 more
similar to ssl_server2. This fixes the client failing on the
psa_destroy_key() call in `ssl-opt.sh -f 'opaque psk on client'`.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
The SSL test programs can now use mbedtls_psa_get_random() rather than
entropy+DRBG as a random generator. This happens if
the configuration option MBEDTLS_USE_PSA_CRYPTO is enabled, or if
MBEDTLS_TEST_USE_PSA_CRYPTO_RNG is set at build time.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This brings them in line with PSA Crypto API 1.0.0
PSA_ALG_AEAD_WITH_DEFAULT_TAG_LENGTH -> PSA_ALG_AEAD_WITH_DEFAULT_LENGTH_TAG
PSA_ALG_AEAD_WITH_TAG_LENGTH -> PSA_ALG_AEAD_WITH_SHORTENED_TAG
Signed-off-by: Bence Szépkúti <bence.szepkuti@arm.com>
With MBEDTLS_USE_PSA_CRYPTO, some of the randomness for the TLS
connection is generated inside the PSA crypto subsystem, which has no
reproducible mode. Whether there is a nonzero amount of randomness
coming from inside the PSA subsystem rather than from the random
generator set by mbedtls_ssl_conf_rng() depends on the choice of
cipher suite and other connection parameters as well as the level of
support for MBEDTLS_USE_PSA_CRYPTO. Rather than give unreliable
results, conservatively abort with a clear error message.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
