Add a test case for doing an ECDH calculation by calling
mbedtls_ecdh_get_params on both keys, with keys belonging to
different groups. This should fail, but currently passes.
Calling psa_*_setup() twice on a MAC, cipher, or hash context should
result in a PSA_ERROR_BAD_STATE error because the operation has already
been set up.
Fixes#10
Extend hash bad order test in line with the new bad order tests for MAC
and cipher, covering more cases and making comments and test layout
consistent.
Ensure that when doing hash operations out of order, PSA_ERROR_BAD_STATE
is returned as documented in crypto.h and the PSA Crypto specification.
In multipart cipher tests, test that each step of psa_cipher_update
produces output of the expected length. The length is hard-coded in
the test data since it depends on the mode.
The length of the output of psa_cipher_finish is effectively tested
because it's the total output length minus the length produced by the
update steps.
Test data obtained with Python+PyCrypto:
AES.new(key, mode=AES.MODE_CTR, counter=Crypto.Util.Counter.new(128, initial_value=0x2a2a2a2a2a2a2a2a2a2a2a2a2a2a2a2a)).encrypt(plaintext.decode('hex')).encode('hex')
The output length can be equal to the input length.
This wasn't noticed at runtime because we happened to only test with
CBC with the first chunk being a partial block.
Some calls to psa_cipher_finish or psa_cipher_update append to a
buffer. Several of these calls were not calculating the offset into
the buffer or the remaining buffer size correctly.
This did not lead to buffer overflows before because the buffer sizes
were sufficiently large for our test inputs. This did not lead to
incorrect output when the test was designed to append but actually
wrote too early because all the existing test cases either have no
output from finish (stream cipher) or have no output from update (CBC,
with less than one block of input).
The test function pkcs1_rsaes_v15_encrypt gets its fake-random input
for padding from a test parameter. In one test case, the parameter was
too short, causing a fallback to rand(). The reference output depends
on this random input, so the test data was correct only for a platform
with one particular rand() implementation. Supply sufficient
fake-random input so that rand() isn't called.
Check generator validity (i.e. that alg has been initialized) before
allowing reads from the generator or allowing reads of the generator's
capacity.
This aligns our implementation with the documented error code behavior
in our crypto.h and the PSA Crypto API.
PSA spec now defines more generic PSA storage types instead of the ITS
specific ones. This is necessary in order to integrate with
the newer implementation of PSA ITS landing in Mbed OS soon.
Changes include the following:
- psa_status_t replaces psa_its_status_t
- psa_storage_info_t replaces psa_its_info_t
- psa_storage_uid_t replaces psa_its_uid_t
Resolve conflicts by performing the following.
- Take the upstream Mbed TLS ChangeLog verbatim.
- Reject changes to Makefiles and CMake that are related to using Mbed
Crypto as a submodule. It doesn't make sense to use Mbed Crypto as a
submodule of itself.
- Reject README changes, as Mbed Crypto has its own, different README.
- Reject PSA-related changes to config.h. We don't want to disable the
availability of the PSA Crypto API by default in the Mbed Crypto
config.h.
- Don't inadvertently revert dead code removal in
mbedtls_cipher_write_tag() which was added in f2a7529403 ("Fix
double return statement in cipher.c")
- Where Mbed Crypto already had some MBEDTLS_USE_PSA_CRYPTO code (from
past companion PRs) take the latest version from Mbed TLS which
includes integration with MBEDTLS_CHECK_PARAMS.
- Update the version of the shared library files to match what's
currently present in Mbed TLS.
- Reject removal of testing with PSA from config full tests.
- Resolve conflicts in test tests/suites/helpers.function, where both
Mbed Crypto and Mbed TLS both added documentation for TEST_ASSERT.
Combine text from both documentation efforts.
- Reject adding a submodule of ourselves.
- Reject addition of submodule tests in all.sh.
- Reject addition of submodule to library path in
tests/scripts/run-test-suites.pl.
- Avoid using USE_CRYPTO_SUBMODULE=1 in
component_test_use_psa_crypto_full_cmake_asan() in all.sh.
Don't unconditionally enable PSA Crypto for all tests. Only enable it in
tests that require it. This allows crypto tests to check that
psa_crypto_init() fails when it is supposed to fail, since we want to
perform some action in a test, and then call psa_crypto_init() and check
the result without it having been called previously.
The existing test `x509parse_crt()` for X.509 CRT parsing
so far used the generic parsing API `mbedtls_x509_crt_parse()`
capable of parsing both PEM encoded and DER encoded certficates,
but was actually only used with DER encoded input data. Moreover,
as the purpose of the test is the testing of the core DER X.509 parsing
functionality, not the PEM vs. DER dispatch (which is now already tested
in the various `x509_crt_info()` tests), the call can be replaced with a
direct call to `mbedtls_x509_parse_crt_der()`.
This commit does that, and further adds to the test an analogous
call to the new API `mbedtls_x509_parse_crt_der_nocopy()` to test
copyless parsing of X.509 certificates.
Test that freshly-initialized contexts exhibit default behavior through
the API. Do this without depending on the internal representation of the
contexts. This provides better portability of our tests on compilers
like MSVC.
For must-fail asymmetric decryption tests, add an output size parameter
so that tests can directly control what output buffer size they allocate
and use independently from the key size used. This enables better
testing of behavior with various output buffer sizes.
When RSA decrypting, unlike with RSA encrypting, we sometimes expect the
output length will be less than the key size. For instance, in the case
where the plaintext is zero-length we expect the output length of the
decryption to be zero-length as well, not key size in length.
For must-fail tests, we don't expect output-buffer-sized RSA-decryption,
only that the output length is less than or equal to the output size, so
these tests remain unchanged.
Change the must-pass tests to expect that the actual output size is
equal to the expected length of the output buffer instead of always
being the key size.
Merge a development version of Mbed TLS 2.16.0 that doesn't have
parameter validation into development.
The following conflicts were resolved:
- Update ChangeLog to include release notes merged from development so
far, with a version of "2.14.0+01b34fb316a5" and release date of
"xxxx-xx-xx" to show this is not a released version, but instead a
snapshot of the development branch equivalent to version of the 2.14.0
with additional commits from the mbedtls/development branch up through
01b34fb316 included. Entries added for unreleased versions of Mbed
Crypto remain at the top of the file for Mbed TLS 2.xx.x.
- Replace the Mbed Crypto version of
mbedtls_rsa_rsaes_pkcs1_v15_decrypt() with the version from Mbed TLS
which fixes timing variations and memory access variations that could
lead to a Bleichenbacher-style padding oracle attack. This will
prevent using psa_asymmetric_decrypt() with zero-length output buffers
until a follow up commit is made to restore this capability.
- In ssl_srv.c, include changes for both the new ECDH interface and
opaque PSK as already added to development previously.
In one place, exercise_key was used in a such a way that if the test
failed inside exercise_key, the test suite would correctly report the
test as failed but would not report the exact location of the failure.
Fix this.
Add documentation for exercise_key that explains how to use it.
Additional changes to temporarily enable running tests:
ssl_srv.c and test_suite_ecdh use mbedtls_ecp_group_load instead of
mbedtls_ecdh_setup
test_suite_ctr_drbg uses mbedtls_ctr_drbg_update instead of
mbedtls_ctr_drbg_update_ret
Test a few cases. The logic to combine the constraint is similar to
the logic to combine the source and target, so it's ok to have less
parameter domain coverage for constraints.
Split the testing into tests that exercise policies in
test_suite_psa_crypto and tests that exercise slot content (slot
states, key material) in test_suite_psa_crypto_slot_management.
Test various cases of source and target policies with and without
wildcards. Missing: testing of the policy constraint on psa_copy_key
itself.
Test several key types (raw data, AES, RSA). Test with the
source or target being persistent.
Add failure tests (incompatible policies, source slot empty, target
slot occupied).
Remove front matter from our EC key format, to make it just the contents
of an ECPoint as defined by SEC1 section 2.3.3.
As a consequence of the simplification, remove the restriction on not
being able to use an ECDH key with ECDSA. There is no longer any OID
specified when importing a key, so we can't reject importing of an ECDH
key for the purpose of ECDSA based on the OID.
Remove pkcs-1 and rsaEncryption front matter from RSA public keys. Move
code that was shared between RSA and other key types (like EC keys) to
be used only with non-RSA keys.
New function psa_copy_key().
Conflicts:
* library/psa_crypto.c: trivial conflicts due to consecutive changes.
* tests/suites/test_suite_psa_crypto.data: the same code
was added on both sides, but with a conflict resolution on one side.
* tests/suites/test_suite_psa_crypto_metadata.function: the same code
was added on both sides, but with a conflict resolution on one side.
You can use PSA_ALG_ANY_HASH to build the algorithm value for a
hash-and-sign algorithm in a policy. Then the policy allows usage with
this hash-and-sign family with any hash.
Test that PSA_ALG_ANY_HASH-based policies allow a specific hash, but
not a different hash-and-sign family. Test that PSA_ALG_ANY_HASH is
not valid for operations, only in policies.
Test for a subclass of public-key algorithm: those that perform
full-domain hashing, i.e. algorithms that can be broken down as
sign(key, hash(message)).
Remove the type and bits arguments to psa_allocate_key() and
psa_create_key(). They can be useful if the implementation wants to
know exactly how much space to allocate for the slot, but many
implementations (including ours) don't care, and it's possible to work
around their lack by deferring size-dependent actions to the time when
the key material is created. They are a burden to applications and
make the API more complex, and the benefits aren't worth it.
Change the API and adapt the implementation, the units test and the
sample code accordingly.
Change the key derivation API to take inputs in multiple steps,
instead of a single one-site-fits-poorly function.
Conflicts:
* include/psa/crypto.h: merge independent changes in the documentation
of psa_key_agreement (public_key from the work on public key formats
vs general description and other parameters in the work on key derivation).
* tests/suites/test_suite_psa_crypto.data: update the key agreement
tests from the work on key derivation to the format from the work on
public key formats.
* tests/suites/test_suite_psa_crypto_metadata.function: reconcile the
addition of unrelated ALG_IS_xxx macros
Get rid of "key selection" algorithms (of which there was only one:
raw key selection).
Encode key agreement by combining a raw key agreement with a KDF,
rather than passing the KDF as an argument of a key agreement macro.
Remove front matter from our EC key format, to make it just the contents
of an ECPoint as defined by SEC1 section 2.3.3.
As a consequence of the simplification, remove the restriction on not
being able to use an ECDH key with ECDSA. There is no longer any OID
specified when importing a key, so we can't reject importing of an ECDH
key for the purpose of ECDSA based on the OID.
You can use PSA_ALG_ANY_HASH to build the algorithm value for a
hash-and-sign algorithm in a policy. Then the policy allows usage with
this hash-and-sign family with any hash.
Test that PSA_ALG_ANY_HASH-based policies allow a specific hash, but
not a different hash-and-sign family. Test that PSA_ALG_ANY_HASH is
not valid for operations, only in policies.
Test for a subclass of public-key algorithm: those that perform
full-domain hashing, i.e. algorithms that can be broken down as
sign(key, hash(message)).
Remove pkcs-1 and rsaEncryption front matter from RSA public keys. Move
code that was shared between RSA and other key types (like EC keys) to
be used only with non-RSA keys.
Add new initializers for cipher operation objects and use them in our
tests and library code. Prefer using the macro initializers due to their
straightforwardness.
Add new initializers for MAC operation objects and use them in our tests
and library code. Prefer using the macro initializers due to their
straightforwardness.
Add new initializers for hash operation objects and use them in our
tests and library code. Prefer using the macro initializers due to their
straightforwardness.
Add new initializers for key policies and use them in our docs, example
programs, tests, and library code. Prefer using the macro initializers
due to their straightforwardness.
Change the way some lines are wrapped to cut at a more logical place.
This commit mainly rewrites multi-line calls to TEST_EQUAL, and also a
few calls to PSA_ASSERT.
This commit is the result of the following command, followed by
reindenting (but not wrapping lines):
perl -00 -i -pe 's/^( *)TEST_ASSERT\(([^;=]*)(?: |\n *)==([^;=]*)\);$/${1}TEST_EQUAL($2,$3);/gm' tests/suites/test_suite_psa_*.function
This commit is the result of the following command, followed by
reindenting (but not wrapping lines):
perl -00 -i -pe 's/^( *)TEST_ASSERT\(([^;=]*)(?: |\n *)==\s*PSA_SUCCESS\s*\);$/${1}PSA_ASSERT($2 );/gm' tests/suites/test_suite_psa_*.function
Cause a compilation error on ARRAY_LENGTH(p) where p is a pointer as
opposed to an array. This only works under GCC and compatible
compilers such as Clang. On other compilers, ARRAY_LENGTH works but
doesn't check the type of its argument.
Document when a context must be initialized or not, when it must be
set up or not, and whether it needs a private key or a public key will
do.
The implementation is sometimes more liberal than the documentation,
accepting a non-set-up context as a context that can't perform the
requested information. This preserves backward compatibility.
The MPI_VALIDATE_RET() macro cannot be used for parameter
validation of mbedtls_mpi_lsb() because this function returns
a size_t.
Use the underlying MBEDTLS_INTERNAL_VALIDATE_RET() insteaed,
returning 0 on failure.
Also, add a test for this behaviour.
For mbedtls_pk_parse_key and mbedtls_pk_parse_keyfile, the password is
optional. Clarify what this means: NULL is ok and means no password.
Validate parameters and test accordingly.
The test that mbedtls_aria_free() accepts NULL parameters
can be performed even if MBEDTLS_CHECK_PARAMS is unset, but
was previously included in the test case aria_invalid_params()
which is only executed if MBEDTLS_CHECK_PARAMS is set.
Parameter validation was previously performed and tested unconditionally
for the ChaCha/Poly modules. This commit therefore only needs go guard the
existing tests accordingly and use the appropriate test macros for parameter
validation.
This commit finishes the removal of support for direct access to key
slots in psa_crypto.c.
This marks the end of the necessary phase of the transition to key
handles. The code should subsequently be refactored to move key slot
management from psa_crypto.c to psa_crypto_slot_management.c.
Switch from the direct use of slot numbers to handles allocated by
psa_allocate_key.
The general principle for each function is:
* Change `psa_key_slot_t slot` to `psa_key_handle_t handle` or
`psa_key_id_t key_id` depending on whether it's used as a handle to
an open slot or as a persistent name for a key.
* Call psa_create_key() before using a slot, instead of calling
psa_set_key_lifetime to make a slot persistent.
Remove the unit test persistent_key_is_configurable which is no longer
relevant.
Switch from the direct use of slot numbers to handles allocated by
psa_allocate_key.
This commit does not affect persistent key tests except for the one
test function in test_suite_psa_crypto that uses persistent keys
(persistent_key_load_key_from_storage).
The general principle for each function is:
* Change `psa_key_slot_t slot` to `psa_key_handle_t handle`.
* Call psa_allocate_key() before setting the policy of the slot,
or before creating key material in functions that don't set a policy.
* Some PSA_ERROR_EMPTY_SLOT errors become PSA_ERROR_INVALID_HANDLE
because there is now a distinction between not having a valid
handle, and having a valid handle to a slot that doesn't contain key
material.
* In tests that use symmetric keys, calculate the max_bits parameters
of psa_allocate_key() from the key data size. In tests where the key
may be asymmetric, call an auxiliary macro KEY_BITS_FROM_DATA which
returns an overapproximation. There's no good way to find a good
value for max_bits with the API, I think the API should be tweaked.
Implement psa_allocate_key, psa_open_key, psa_create_key,
psa_close_key.
Add support for keys designated to handles to psa_get_key_slot, and
thereby to the whole API.
Allocated and non-allocated keys can coexist. This is a temporary
stage in order to transition from the use of direct slot numbers to
allocated handles only. Once all the tests and sample programs have
been migrated to use handles, the implementation will be simplified
and made more robust with support for handles only.
Previously, one could change the definition of AES_VALIDATE_RET() to return
some other code than MBEDTLS_ERR_AES_BAD_INPUT_DATA, and the test suite
wouldn't notice. Now this modification would make the suite fail as expected.
The test framework for validation of parameters depends on the macro
MBEDTLS_PARAM_FAILED() being set to its default value when building the
library. So far the test framework attempted to define this macro but this was
the wrong place - this definition wouldn't be picked by the library.
Instead, a different approach is taken: skip those tests when the macro is
defined in config.h, as in that case we have no way to know if it will indeed
end up calling mbedtls_param_failed() as we need it to.
This commit was tested by manually ensuring that aes_invalid_params:
- passes (and is not skipped) in the default configuration
- is skipped when MBEDTLS_PARAM_FAILED() is defined in config.h
The previous prototype gave warnings are the strings produced by #cond and
__FILE__ are const, so we shouldn't implicitly cast them to non-const.
While at it modifying most example programs:
- include the header that has the function declaration, so that the definition
can be checked to match by the compiler
- fix whitespace
- make it work even if PLATFORM_C is not defined:
- CHECK_PARAMS is not documented as depending on PLATFORM_C and there is
no reason why it should
- so, remove the corresponding #if defined in each program...
- and add missing #defines for mbedtls_exit when needed
The result has been tested (make all test with -Werror) with the following
configurations:
- full with CHECK_PARAMS with PLATFORM_C
- full with CHECK_PARAMS without PLATFORM_C
- full without CHECK_PARAMS without PLATFORM_C
- full without CHECK_PARAMS with PLATFORM_C
Additionally, it has been manually tested that adding
mbedtls_aes_init( NULL );
near the normal call to mbedtls_aes_init() in programs/aes/aescrypt2.c has the
expected effect when running the program.
It was inconsistent between files: sometimes 3 arguments, sometimes one.
Align to 1 argument for the macro and 3 for the function, because:
- we don't need 3 arguments for the macro, it can add __FILE__ and __LINE__
in its expansion, while the function needs them as parameters to be correct;
- people who re-defined the macro should have flexibility, and 3 arguments
can give the impression they they don't have as much as they actually do;
- the design document has the macro with 1 argument, so let's stick to that.
Change the use of setjmp and longjmp in signalling parameter validation failures
when using the MBEDTLS_CHECK_PARAMS config.h option. This change allows
all calls which might result in a call to the parameter validation failure
handler to always be caught, even without use of the new macros, by placing a
setjmp() in the outer function which calls the test function, which the handler
can jump to.
This has several benefits:
* it allows us to remove the clang compiler warning (-Wclobbered) caused
by local auto variables being in the same function as the call to setjmp.
* removes the need to wrap all function calls in the test functions with the
TEST_ASSERT() macro. Now all parameter validation function calls should be
caught.
The tests for the ECDH key exchange that use the context accessed it
directly. This can't work with the new context, where we can't make any
assumptions about the implementation of the context. This commit works
around this problem and comes with the cost of allocating an extra
structures on the stack when executing the test.
One of the tests is testing an older interface for the sake of backward
compatibility. The new ECDH context is not backward compatible and this
test doesn't make any sense for it, therefore we skip this test in
non-legacy mode.
