Commit graph

4665 commits

Author SHA1 Message Date
Ronald Cron 672279829a psa: Call export software implementation as a driver
Signed-off-by: Ronald Cron <ronald.cron@arm.com>
2021-02-02 11:29:08 +01:00
Ronald Cron 8328287956 psa: Move from validate_key to import_key entry point
In the course of the development of the PSA unified
driver interface, the validate_key entry point for
opaque drivers has been removed and replaced by an
import_key entry point. This commit takes into account
this change of specification.

Signed-off-by: Ronald Cron <ronald.cron@arm.com>
2021-02-02 11:29:07 +01:00
Gilles Peskine 612ffd2aa1 Fix copypasta in conditional directive in cipher_setup
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:22:38 +01:00
Gilles Peskine 80a970806f No dependencies for storage format tests
Storage format tests that only look at how the file is structured and
don't care about the format of the key material don't depend on any
cryptographic mechanisms.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:22:38 +01:00
Gilles Peskine b51d72f82e Run copy_fail even when the target policy is unsupported
The negative test cases for psa_copy_key() don't actually care whether
the target policy is supported. This is similar to _key_policy tests.
Add a similar rule.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:22:38 +01:00
Gilles Peskine fa37961413 Rename OMITTED_SYSTEMATIC_DEPENDENCIES
It isn't a set of dependencies, it's a set of symbols. So give it a
name that describes the symbol rather than a name that pretends it's a
collection of dependencies.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:22:38 +01:00
Gilles Peskine 0794572f96 More precise recognition of key policy negative tests
The test function asymmetric_signature_key_policy combines positive
and negative tests inside the code, so it doesn't take a status as its
last argument.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:22:38 +01:00
Gilles Peskine d6dc40cae1 Update hash dependencies in PSA test code
Switch dependencies on MBEDTLS_xxx to PSA_WANT_xxx for hash
algorithms.

Add a missing dependency in bad_order functions (it was previously
expressed in the .data file, but this is no longer the case when
dependencies in the .data file are determined automatically).

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:22:38 +01:00
Gilles Peskine 8405b5f60d Change PSA crypto test dependencies to PSA_WANT_xxx
Replace manually written dependencies on MBEDTLS_xxx with
PSA_WANT_xxx dependencies that are determined automatically from the
test data.

Run tests/scripts/set_psa_test_dependencies.py on
tests/suites/test_suite_psa_crypto*.data,
except for the dynamic secure element tests in
tests/suites/test_suite_psa_crypto_se_driver_hal*.data.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:21:40 +01:00
Gilles Peskine a87e178553 If there are no arguments, don't update the file content
This avoids having to bother with edge cases of the .data syntax.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine 72d8e0ad04 Key policy negative tests don't actually require the algorithm
It doesn't make much difference in practice, but to keep closer to
what the current code does, run negative key policy tests even if the
algorithm for the operation attempt is not supported.

In particular, this allows the following test cases to run:
* "PSA key policy: agreement + KDF, wrong agreement algorithm"
* "PSA key policy: raw agreement, wrong algorithm"
Without this exception, those two test cases would never run, because
they would depend on PSA_ALG_WANT_FFDH. Since FFDH is not implemented
yet, it isn't enabled in any configuration. There's no alternative to
FFDH for these particular test cases because ECDH is the only key
agreement that is implemented in Mbed TLS so far.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine 20987b9a46 Detect test cases that require something not to be supported
PSA_xxx_CATEGORY_yyy is used in metadata tests where it doesn't
involve any particular support, and elsewhere it's used as a value
that is definitely not supported but is in a plausible range. Such
symbols do not require any dependency.

If a test case is expects PSA_ERROR_NOT_SUPPORTED, its
dependencies (often including one negative dependency) cannot be
determined automatically, so leave that test case alone.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine f032fa9d94 Special handling for some always-on or alias symbols
Some symbols don't require a dependency symbol:

* Modifiers such as truncated MAC
* Always-on features such as the raw data key type
* Aliases or special values such as RSA PKCS#1v1.5 raw

I'm not convinced that all of these warrant special handling in the
script, rather than having the expected symbol defined somewhere. But
for now I prefer to minimize changes to the header files.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine 81dec0047f Remove classic crypto dependencies
Don't remove cipher-related dependencies because the corresponding
PSA_WANT_xxx dependencies are not implemented yet.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine e4f539caf9 Don't add cipher-related PSA_WANT_xxx because they aren't implemented yet
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine 9bbba5ea69 Add a function to remove classic crypto dependencies
A list of symbols to remove will be added in a subsequent commit.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine 2d2e924401 Implement basic update of PSA_WANT dependencies
Remove any existing PSA_WANT_xxx dependency. Add PSA_WANT_xxx
dependencies based on the PSA_KEY_TYPE_xxx and PSA_ALG_xxx symbols
used in the test case arguments.

PSA_ECC_FAMILY_xxx and PSA_DH_GROUP_xxx are not implemented yet in the
PSA conditional inclusion mechanism in Mbed TLS, so this script
doesn't handle them yet.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine 45e9e73e66 Add a possibility to keep manual dependencies for certain functions
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine 82ebaa4d36 Implement stanza parsing in .data files
Parse the existing dependencies. For now, just write them back.
Subsequent commits will implement the dependency processing that is
the goal of this program.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine bdffaeaf3a New script to determine PSA crypto test dependencies automatically
This commit only contains a framework to rewrite .data files. No
actual modification of the content is implemented yet.

For now, command line parsing is trivial: just a list of file names,
with no options.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Gilles Peskine 6fea21d515 Declare a code dependency in the code instead of in the data
When using the test function persistent_key_load_key_from_storage with
DERIVE_KEY, there's a dependency on HKDF-SHA-256. Since this
dependency is in the code, declare it there rather than with the data.
If the depenency is not met, mark the test as skipped since it can't
create the key to be tested.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-02-01 13:19:26 +01:00
Janos Follath a209f34faf
Merge pull request #3996 from stevew817/feature/allow_reading_external_keys
Allow loading external wrapped keys
2021-01-29 13:34:11 +00:00
paul-elliott-arm a14d1642f4
Merge pull request #4008 from stevew817/bugfix/fix_dependencies
Fix warnings and dependencies when using _ALT in test
2021-01-29 12:24:54 +00:00
Ronald Cron 88a8035860
Merge pull request #3953 from gilles-peskine-arm/python-mypy-mkdir
Python upscale: pass mypy and create library directory
2021-01-29 12:07:53 +01:00
Ronald Cron 318515b384
Merge pull request #3984 from gabor-mezei-arm/3268_update_macros_for_ouput_buffer_size_renames
Rename existing support macros for output buffer sizes for PSA Crypto API 1.0.0
2021-01-29 09:31:59 +01:00
Ronald Cron 0426e2545d
Merge pull request #4020 from gilles-peskine-arm/ssl_test_lib-hmac_drg
Support HMAC_DRBG in SSL test programs
2021-01-29 08:58:23 +01:00
Steven Cooreman 29eecbf6e1 move _bad_order() test case dependencies to the function itself
Signed-off-by: Steven Cooreman <steven.cooreman@silabs.com>
2021-01-28 19:41:25 +01:00
Steven Cooreman 107409f470 Apply review feedback
Signed-off-by: Steven Cooreman <steven.cooreman@silabs.com>
2021-01-26 12:01:22 +01:00
Gilles Peskine 75829a4296 Explain the explicit dependency on instrument_record_status.h
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-25 13:46:14 +01:00
gabor-mezei-arm cbcec21684
Rename output buffer size macros
Rename existing support macros for output buffer sizes for PSA Crypto API 1.0.0

Signed-off-by: gabor-mezei-arm <gabor.mezei@arm.com>
2021-01-21 13:17:25 +01:00
Gilles Peskine 1af872d23b Move the fake PSA external RNG to its own header and source files
Move the declaration of the functions needed to use the test
implementation of mbedtls_psa_external_get_random() to a new header
file. Before, they were declared in
tests/include/test/psa_crypto_helpers.h, but this header file can't be
included in sample programs because it also includes headers from the
library directory which is not on the include path for sample
programs.

This fixes the build of the sample programs when
MBEDTLS_PSA_CRYPTO_EXTERNAL_RNG and MBEDTLS_USE_PSA_CRYPTO are
enabled.

Move the implementation of the functions to a separate .c file as
well. This isn't strictly necessary, but makes the structure of the
source code easier to understand.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-20 20:02:01 +01:00
Gilles Peskine e1d51bd99d Depend on all test headers when building tests
There were explicit dependencies on header files for some test suites,
dating back from when only a few test suites depended on anything in
tests/include. The noted dependencies were still correct, but now that
tests/include is more populated, they were only the tip of the
iceberg. Just keep it simple and depend on all the headers.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-20 19:49:23 +01:00
Gilles Peskine 0370c17105 mypy: require at least version 0.780
0.780 works. The previous release, 0.770, does not.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-20 12:03:24 +01:00
Gilles Peskine 4738b96d75 Use $PYTHON when running mypy
Make sure to run mypy with the desired Python version.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-19 21:45:32 +01:00
Gilles Peskine c3b178768f Use can_mypy rather than just checking for mypy
As indicated in the comments in the can_mypy function, we don't just
need a mypy executable to be present, we need it to work.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-19 21:43:24 +01:00
Gilles Peskine bdde5d002c Use Python to check the version of pylint
This reduces dependencies, doesn't require maintainers to know awk,
and makes the version parsing more robust.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-19 21:42:05 +01:00
Gilles Peskine 6d82a7ef9f Say we're running pylint
Now that the script might additionally run mypy, it's more
user-friendly to indicate what's going on at the beginning as well.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-19 21:23:11 +01:00
Gilles Peskine 2991b5f6c0 Minor documentation fixes
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-19 21:23:11 +01:00
Dave Rodgman 6fbff5b557
Merge pull request #3698 from darrenkrahn/development
Mark basic constraints critical as appropriate.
2021-01-17 18:06:18 +00:00
Darren Krahn 9c134cef35 Add build instructions for new test data.
Signed-off-by: Darren Krahn <dkrahn@google.com>
2021-01-13 22:04:45 -08:00
Gilles Peskine a222434952 Test SSL with non-deterministic ECDSA
In component_test_no_hmac_drbg, the fact that HMAC_DRBG is disabled
doesn't affect the SSL code, but the fact that deterministic ECDSA is
disabled does. So run some ECDSA-related SSL tests.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-13 20:31:24 +01:00
Gilles Peskine ba74904c48 SSL test programs: support HMAC_DRBG
Support HMAC_DRBG in ssl_client2 and ssl_server2, in addition to
CTR_DRBG. CTR_DRBG is still used if present, but it's now possible to
run the SSL test programs with CTR_DRBG disabled.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-13 20:30:03 +01:00
Steven Cooreman 9b31218109 Testing a maximum size key is RAM intensive, and should be dependent
Signed-off-by: Steven Cooreman <steven.cooreman@silabs.com>
2021-01-11 17:59:36 +01:00
Steven Cooreman affc38c90d Make test cases that use PKCS7 padding dependent on it
Signed-off-by: Steven Cooreman <steven.cooreman@silabs.com>
2021-01-11 17:56:19 +01:00
Steven Cooreman 98435ddf84 Allow loading wrapped keys even when SE support is compiled in
Signed-off-by: Steven Cooreman <steven.cooreman@silabs.com>
2021-01-11 11:33:10 +01:00
Gilles Peskine c85c20147b Don't call TEST_ASSERT in PSA_DONE
TEST_ASSERT jumps to the exit label, so it must not be called from
cleanup code executed after the exit label. It's legitimate (and
indeed very common) to call PSA_DONE in cleanup code, so PSA_DONE must
not jump to exit.

Define an auxiliary function test_fail_if_psa_leaking() that calls
test_fail() with the error message provided by
mbedtls_test_helper_is_psa_leaking(). This function currently needs to
be in helpers.function rather than in a PSA-specific helper file
because it calls test_fail which is defined in helpers.function.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-06 20:49:42 +01:00
Gilles Peskine 6beb327a5e external_rng_failure_sign: more robust buffer management
Don't microoptimize memory usage in tests: use separate buffers for
the input and the output. Allocate the input buffer dynamically
because the size is a parameter of the test case. Allocate the output
buffer dynamically because it's generally good practice in tests so
that a memory sanitizer can detect a buffer overflow.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-06 20:16:26 +01:00
Gilles Peskine 3aa5a6414e Fix a test dependency
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-06 20:06:36 +01:00
Gilles Peskine 1631514b8e Add missing dependencies on key types
Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-06 18:22:40 +01:00
Gilles Peskine ba0c1ffb72 Depend on the built-in implementation when injecting RNG failure
When verifying the impact of a forced RNG failure, depend on the built-in
implementation of the algorithm that uses randomization, whether it's
because the algorithm is randomized or because our implementation uses
randomization for (e.g.) blinding. An external implementation could use
its own randomness source which is not affected by the forced failure of
the RNG driver.

Signed-off-by: Gilles Peskine <Gilles.Peskine@arm.com>
2021-01-06 18:22:40 +01:00