mbedtls_ctr_drbg_free() left a mutex in the initialized state. This
caused a resource leak on platforms where mbedtls_mutex_init()
allocates resources.
To fix this, mbedtls_ctr_drbg_free() no longer reinitializes the
mutex. To preserve the property that mbedtls_ctr_drbg_free() leaves
the object in an initialized state, which is generally true throughout
the library except regarding mutex objects on some platforms, no
longer initialize the mutex in mbedtls_ctr_drbg_init(). Since the
mutex is only used after seeding, and seeding is only permitted once,
call mbedtls_mutex_init() as part of the seeding process.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Document the usage inside the library, and relate it with how it's
additionally used in the test code.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Subtract the number of calls to mbedtls_mutex_free() from the number
of calls to mbedtls_mutex_init(). A mutex leak will manifest as a
positive result at the end of the test case.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
If the mutex usage verification framework is enabled and it detects a
mutex usage error, report this error and mark the test as failed.
This detects most usage errors, but not all cases of using
uninitialized memory (which is impossible in full generality) and not
leaks due to missing free (which will be handled in a subsequent commit).
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
When using pthread mutexes (MBEDTLS_THREADING_C and
MBEDTLS_THREADING_PTHREAD enabled), and when test hooks are
enabled (MBEDTLS_TEST_HOOKS), set up wrappers around the
mbedtls_mutex_xxx abstraction. In this commit, the wrapper functions
don't do anything yet.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Some functions were not deinitializing the PSA subsystem. This could
lead to resource leaks at the level of individual test cases, and
possibly at the level of the whole test suite depending on the order
and selection of test cases.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Make USE_PSA_INIT() and USE_PSA_DONE() available in all test suites in
all cases, doing nothing if MBEDTLS_USE_PSA_CRYPTO is disabled. Use
those in preference to having explicit
defined(MBEDTLS_USE_PSA_CRYPTO) checks (but there may still be places
left where using the new macros would be better).
Also provide PSA_INIT() by symmetry with PSA_DONE(), functional
whenver MBEDTLS_PSA_CRYPTO_C is enabled, but currently unused.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
There are type annotations that indirectly depend on the
typing_extensions module (on Python 3.5-3.7: Protocol was added to the
core typing module in 3.8). The typing_extensions module is not
installed by default, so the code didn't run on a pristine Python
installation.
To avoid depending on a non-default module, make the dependency on
typing_extensions optional. (It's still required to run mypy, but
installing mypy takes care of providing typing_extensions.) If it
isn't available, provide a substitute definition that's just good
enough to get the scripts to run.
Move this ugly code to its own module to avoid the temptation of
spreading such ugliness all over the place. It's likely to be used in
other modules anyway.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
A temporary hack: at the time of writing, not all dependency symbols
are implemented yet. Skip test cases for which the dependency symbols are
not available. Once all dependency symbols are available, this comit
should be reverted.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
PSA_KEY_TYPE_RAW_DATA and PSA_KEY_TYPE_DERIVE are always supported.
Make this explicit by declaring PSA_WANT_KEY_TYPE_RAW_DATA and
PSA_WANT_KEY_TYPE_DERIVE unconditionally. This makes it easier to
infer dependencies in a systematic way.
Don't generate not-supported test cases for those key types. They
would always be skipped, which is noise and would make it impossible
to eventually validate that all test cases pass in at least one
configuration over the whole CI.
Don't remove the exception in set_psa_test_dependencies.py for now, to
get less noise in dependencies. This may be revised later if it is
deemed more important to be systematic.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
ECC curve dependency symbols include the key size in addition to the
curve family. Tweak the dependencies once the key size is known.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This test data file is automatically generated. We could do that as
part of the build, since the only requirement is Python and we have a
requirement on Python to build tests anyway (to generate the .c file
from the .function file). However, committing the generating file into
the repository has less impact on build scripts, and will be necessary
for some of the files generated by generate_psa_tests.py (at least the
storage format stability tests, for which stability is guaranteed by
the fact that the generated file doesn't change). To keep things
simple, for now, let's commit all the files generated by
generate_psa_tests.py into the repository.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Otherwise the generation is sensitive to trivial differences such as
running `tests/scripts/generate_psa_tests.py` vs
`./tests/scripts/generate_psa_tests.py` vs an absolute path.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Add support for ECC key types to the generation of not-supported test
cases in generate_psa_tests.py. For each curve, generate test cases
both for when ECC isn't supported and for when the curve isn't
supported.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
SECP192R1 is declared in the PSA API specification, but it's an old
one that Mbed TLS doesn't support and even OpenSSL doesn't support.
We don't have test vectors for it. Just skip it.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Asymmetric keys can't just be arbitrary byte strings: the public key
has to match the private key and the private key usually has
nontrivial constraints.
In order to have deterministic test data and not to rely on
cryptographic dependencies in the Python script, hard-code some test
keys.
In this commit, copy some test keys from test_suite_psa_crypto.data.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
For each declared key type, generate test cases for psa_import_key and
psa_generate_key when the corresponding type is not supported.
Some special cases:
* Public keys can never be generated.
* Omit key types that Mbed TLS does not support at all.
* ECC and FFDH, which depend on a curve/group, are not covered yet.
The generated test cases are written to
tests/suites/test_suite_psa_crypto_not_supported.generated.data .
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
This commit creates a script to generate test cases automatically
based on enumerating PSA key types, algorithms and other
classifications of cryptographic mechanisms.
Subsequent commits will implement the generation of test cases.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
To start with, test that key creation fails as intended when the key
type is not supported. This commit only covers psa_import_key and
psa_generate_key. A follow-up will cover psa_key_derivation_output_key.
My primary intent in creating this new test suite is to automatically
generate test cases by enumerating the key types and algorithms that
the library supports. But this commit only adds a few manually written
test cases, to get the ball rolling.
Move the relevant test cases of test_suite_psa_crypto.data that only
depend on generic knowledge about the API. Keep test cases that depend
more closely on the implementation, such as tests of non-supported key
sizes, in test_suite_psa_crypto.data.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
New Python module intended to gather knowledge about key types and
cryptographic mechanisms, such as the ability to create test data for
a given key type and the determination of whether an algorithm is
compatible with a key type.
This commit just creates a class for knowledge about key types.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Older versions of Pylint (<2.1.0) limited the names of various things
to 30 characters. We allowed up to 35 characters for method name due
to pre-existing identifiers.
Modern versions of Pylint no longer have an upper limit. Let's not be
pickier than Pylint's default configuration.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
At the moment it makes no difference, but it could if e.g. a new
algorithm was called 'foomask'.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
By default, exclude macros whose numerical value is not a valid member
of the semantic type (e.g. PSA_ALG_xxx_BASE is not itself an
algorithm, only an intermediate value used to construct others). But
do include them with include_intermediate=True, which
generate_psa_constants.py does.
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
Update with CHACHA20_POLY1305, CHACHA20. Add in CTR, which was missing,
and move ALG_XTS to its proper location alphabetically.
Signed-off-by: David Brown <david.brown@linaro.org>
Change the psa_crypto use of the CHACHA20 cipher to also use the new
MBEDTLS_PSA_BUILTIN_KE_TYPE_CHACHA20.
Signed-off-by: David Brown <david.brown@linaro.org>
Change a few conditionals in the psa library to be based on the
MBEDTLS_PSA_BUILTIN_KEY_TYPE_DES instead of the WANT macros. Future
additions of HW acceleration will need to be mindful of these
definitions if any of this code is needed in those instances.
Signed-off-by: David Brown <david.brown@linaro.org>
With the else branch commented out, both lines are unnecessary. We
could check for the invalid configuration in the future, once tests were
made to exclude this combination.
Signed-off-by: David Brown <david.brown@linaro.org>
There is no PSA ALG_ECB, only ALG_ECB_NO_PADDING. Fix one incorrect
declaration, and remove another that is just redundant.
Signed-off-by: David Brown <david.brown@linaro.org>
Only define MBEDTLS_CIPHER_MODE_CBC if one of the CBC modes is requested
and everything isn't covered by an accelerated version. This keeps this
from being defined in cases where everything needed would be
accelerated.
Signed-off-by: David Brown <david.brown@linaro.org>
When converting definitions to use the new PSA defines, one erroneously
was conditionalized on the WANT macro instead of on the BUILTIN macro.
Signed-off-by: David Brown <david.brown@linaro.org>
