mbedtls/tests/scripts/analyze_outcomes.py

#!/usr/bin/env python3

"""Analyze the test outcomes from a full CI run.

This script can also run on outcomes from a partial run, but the results are
less likely to be useful.
"""

import argparse
import sys
import traceback
import re
import subprocess
import os

import check_test_cases

class Results:
    """Process analysis results."""

    def __init__(self):
        self.error_count = 0
        self.warning_count = 0

    @staticmethod
    def log(fmt, *args, **kwargs):
        sys.stderr.write((fmt + '\n').format(*args, **kwargs))

    def error(self, fmt, *args, **kwargs):
        self.log('Error: ' + fmt, *args, **kwargs)
        self.error_count += 1

    def warning(self, fmt, *args, **kwargs):
        self.log('Warning: ' + fmt, *args, **kwargs)
        self.warning_count += 1

class TestCaseOutcomes:
    """The outcomes of one test case across many configurations."""
    # pylint: disable=too-few-public-methods

    def __init__(self):
        # Collect a list of witnesses of the test case succeeding or failing.
        # Currently we don't do anything with witnesses except count them.
        # The format of a witness is determined by the read_outcome_file
        # function; it's the platform and configuration joined by ';'.
        self.successes = []
        self.failures = []

    def hits(self):
        """Return the number of times a test case has been run.

        This includes passes and failures, but not skips.
        """
        return len(self.successes) + len(self.failures)

def execute_reference_driver_tests(ref_component, driver_component, outcome_file):
    """Run the tests specified in ref_component and driver_component. Results
    are stored in the output_file and they will be used for the following
    coverage analysis"""
    # If the outcome file already exists, we assume that the user wants to
    # perform the comparison analysis again without repeating the tests.
    if os.path.exists(outcome_file):
        Results.log("Outcome file (" + outcome_file + ") already exists. " + \
                    "Tests will be skipped.")
        return

    shell_command = "tests/scripts/all.sh --outcome-file " + outcome_file + \
                    " " + ref_component + " " + driver_component
    Results.log("Running: " + shell_command)
    ret_val = subprocess.run(shell_command.split(), check=False).returncode

    if ret_val != 0:
        Results.log("Error: failed to run reference/driver components")
        sys.exit(ret_val)

def analyze_coverage(results, outcomes):
    """Check that all available test cases are executed at least once."""
    available = check_test_cases.collect_available_test_cases()
    for key in available:
        hits = outcomes[key].hits() if key in outcomes else 0
        if hits == 0:
            # Make this a warning, not an error, as long as we haven't
            # fixed this branch to have full coverage of test cases.
            results.warning('Test case not executed: {}', key)

def analyze_driver_vs_reference(outcomes, component_ref, component_driver,
                                ignored_suites, ignored_test=None):
    """Check that all tests executed in the reference component are also
    executed in the corresponding driver component.
    Skip:
    - full test suites provided in ignored_suites list
    - only some specific test inside a test suite, for which the corresponding
      output string is provided
    """
    available = check_test_cases.collect_available_test_cases()
    result = True

    for key in available:
        # Continue if test was not executed by any component
        hits = outcomes[key].hits() if key in outcomes else 0
        if hits == 0:
            continue
        # Skip ignored test suites
        full_test_suite = key.split(';')[0] # retrieve full test suite name
        test_string = key.split(';')[1] # retrieve the text string of this test
        test_suite = full_test_suite.split('.')[0] # retrieve main part of test suite name
        if test_suite in ignored_suites or full_test_suite in ignored_suites:
            continue
        if ((full_test_suite in ignored_test) and
                (test_string in ignored_test[full_test_suite])):
            continue
        # Search for tests that run in reference component and not in driver component
        driver_test_passed = False
        reference_test_passed = False
        for entry in outcomes[key].successes:
            if component_driver in entry:
                driver_test_passed = True
            if component_ref in entry:
                reference_test_passed = True
        if(reference_test_passed and not driver_test_passed):
            Results.log(key)
            result = False
    return result

def analyze_outcomes(outcomes):
    """Run all analyses on the given outcome collection."""
    results = Results()
    analyze_coverage(results, outcomes)
    return results

def read_outcome_file(outcome_file):
    """Parse an outcome file and return an outcome collection.

An outcome collection is a dictionary mapping keys to TestCaseOutcomes objects.
The keys are the test suite name and the test case description, separated
by a semicolon.
"""
    outcomes = {}
    with open(outcome_file, 'r', encoding='utf-8') as input_file:
        for line in input_file:
            (platform, config, suite, case, result, _cause) = line.split(';')
            key = ';'.join([suite, case])
            setup = ';'.join([platform, config])
            if key not in outcomes:
                outcomes[key] = TestCaseOutcomes()
            if result == 'PASS':
                outcomes[key].successes.append(setup)
            elif result == 'FAIL':
                outcomes[key].failures.append(setup)
    return outcomes

def do_analyze_coverage(outcome_file, args):
    """Perform coverage analysis."""
    del args # unused
    outcomes = read_outcome_file(outcome_file)
    Results.log("\n*** Analyze coverage ***\n")
    results = analyze_outcomes(outcomes)
    return results.error_count == 0

def do_analyze_driver_vs_reference(outcome_file, args):
    """Perform driver vs reference analyze."""
    execute_reference_driver_tests(args['component_ref'], \
                                    args['component_driver'], outcome_file)

    ignored_suites = ['test_suite_' + x for x in args['ignored_suites']]

    outcomes = read_outcome_file(outcome_file)
    Results.log("\n*** Analyze driver {} vs reference {} ***\n".format(
        args['component_driver'], args['component_ref']))
    return analyze_driver_vs_reference(outcomes, args['component_ref'],
                                       args['component_driver'], ignored_suites,
                                       args['ignored_tests'])

# List of tasks with a function that can handle this task and additional arguments if required
TASKS = {
    'analyze_coverage':                 {
        'test_function': do_analyze_coverage,
        'args': {}
        },
    # There are 2 options to use analyze_driver_vs_reference_xxx locally:
    # 1. Run tests and then analysis:
    #   - tests/scripts/all.sh --outcome-file "$PWD/out.csv" <component_ref> <component_driver>
    #   - tests/scripts/analyze_outcomes.py out.csv analyze_driver_vs_reference_xxx
    # 2. Let this script run both automatically:
    #   - tests/scripts/analyze_outcomes.py out.csv analyze_driver_vs_reference_xxx
    'analyze_driver_vs_reference_hash': {
        'test_function': do_analyze_driver_vs_reference,
        'args': {
            'component_ref': 'test_psa_crypto_config_reference_hash_use_psa',
            'component_driver': 'test_psa_crypto_config_accel_hash_use_psa',
            'ignored_suites': [
                'shax', 'mdx', # the software implementations that are being excluded
                'md.psa',  # purposefully depends on whether drivers are present
            ],
            'ignored_tests': {
            }
        }
    },
    'analyze_driver_vs_reference_ecp_light_only': {
        'test_function': do_analyze_driver_vs_reference,
        'args': {
            'component_ref': 'test_psa_crypto_config_reference_ecc_ecp_light_only',
            'component_driver': 'test_psa_crypto_config_accel_ecc_ecp_light_only',
            'ignored_suites': [
                'ecdsa',
                'ecdh',
                'ecjpake',
            ],
            'ignored_tests': {
                'test_suite_random': [
                    'PSA classic wrapper: ECDSA signature (SECP256R1)',
                ],
                # In the accelerated test ECP_C is not set (only ECP_LIGHT is)
                # so we must ignore disparities in the tests for which ECP_C
                # is required.
                'test_suite_ecp': [
                    'ECP check public-private #1 (OK)',
                    'ECP check public-private #2 (group none)',
                    'ECP check public-private #3 (group mismatch)',
                    'ECP check public-private #4 (Qx mismatch)',
                    'ECP check public-private #5 (Qy mismatch)',
                    'ECP check public-private #6 (wrong Qx)',
                    'ECP check public-private #7 (wrong Qy)',
                    'ECP gen keypair [#1]',
                    'ECP gen keypair [#2]',
                    'ECP gen keypair [#3]',
                    'ECP gen keypair wrapper',
                    'ECP point muladd secp256r1 #1',
                    'ECP point muladd secp256r1 #2',
                    'ECP point multiplication Curve25519 (element of order 2: origin) #3',
                    'ECP point multiplication Curve25519 (element of order 4: 1) #4',
                    'ECP point multiplication Curve25519 (element of order 8) #5',
                    'ECP point multiplication Curve25519 (normalized) #1',
                    'ECP point multiplication Curve25519 (not normalized) #2',
                    'ECP point multiplication rng fail Curve25519',
                    'ECP point multiplication rng fail secp256r1',
                    'ECP test vectors Curve25519',
                    'ECP test vectors Curve448 (RFC 7748 6.2, after decodeUCoordinate)',
                    'ECP test vectors brainpoolP256r1 rfc 7027',
                    'ECP test vectors brainpoolP384r1 rfc 7027',
                    'ECP test vectors brainpoolP512r1 rfc 7027',
                    'ECP test vectors secp192k1',
                    'ECP test vectors secp192r1 rfc 5114',
                    'ECP test vectors secp224k1',
                    'ECP test vectors secp224r1 rfc 5114',
                    'ECP test vectors secp256k1',
                    'ECP test vectors secp256r1 rfc 5114',
                    'ECP test vectors secp384r1 rfc 5114',
                    'ECP test vectors secp521r1 rfc 5114',
                ],
                'test_suite_pkparse': [
                    # This is a known difference for Montgomery curves: in
                    # reference component private keys are parsed using
                    # mbedtls_mpi_read_binary_le(), while in driver version they
                    # they are imported in PSA and there the parsing is done
                    # through mbedtls_ecp_read_key(). Unfortunately the latter
                    # fixes the errors which are intentionally set on the parsed
                    # key and therefore the following test case is not failing
                    # as expected.
                    # This cause the following test to be guarded by ECP_C and
                    # not being executed on the driver version.
                    ('Key ASN1 (OneAsymmetricKey X25519, doesn\'t match masking '
                     'requirements, from RFC8410 Appendix A but made into version 0)'),
                ],
            },
        }
    },
    'analyze_driver_vs_reference_no_ecp_at_all': {
        'test_function': do_analyze_driver_vs_reference,
        'args': {
            'component_ref': 'test_psa_crypto_config_reference_ecc_no_ecp_at_all',
            'component_driver': 'test_psa_crypto_config_accel_ecc_no_ecp_at_all',
            'ignored_suites': [
                # Ignore test suites for the modules that are disabled in the
                # accelerated test case.
                'ecp',
                'ecdsa',
                'ecdh',
                'ecjpake',
            ],
            'ignored_tests': {
                'test_suite_random': [
                    'PSA classic wrapper: ECDSA signature (SECP256R1)',
                ],
                'test_suite_psa_crypto': [
                    'PSA key derivation: HKDF-SHA-256 -> ECC secp256r1',
                    'PSA key derivation: HKDF-SHA-256 -> ECC secp256r1 (1 redraw)',
                    'PSA key derivation: HKDF-SHA-256 -> ECC secp256r1, exercise ECDSA',
                    'PSA key derivation: HKDF-SHA-256 -> ECC secp384r1',
                    'PSA key derivation: HKDF-SHA-256 -> ECC secp521r1 #0',
                    'PSA key derivation: HKDF-SHA-256 -> ECC secp521r1 #1',
                    'PSA key derivation: bits=7 invalid for ECC BRAINPOOL_P_R1 (ECC enabled)',
                    'PSA key derivation: bits=7 invalid for ECC SECP_K1 (ECC enabled)',
                    'PSA key derivation: bits=7 invalid for ECC SECP_R1 (ECC enabled)',
                    'PSA key derivation: bits=7 invalid for ECC SECP_R2 (ECC enabled)',
                    'PSA key derivation: bits=7 invalid for ECC SECT_K1 (ECC enabled)',
                    'PSA key derivation: bits=7 invalid for ECC SECT_R1 (ECC enabled)',
                    'PSA key derivation: bits=7 invalid for ECC SECT_R2 (ECC enabled)',
                ],
                'test_suite_pkparse': [
                    # See description provided for the analyze_driver_vs_reference_all_ec_algs
                    # case above.
                    ('Key ASN1 (OneAsymmetricKey X25519, doesn\'t match masking '
                     'requirements, from RFC8410 Appendix A but made into version 0)'),
                    # When PK_PARSE_C and ECP_C are defined then PK_PARSE_EC_COMPRESSED
                    # is automatically enabled in build_info.h (backward compatibility)
                    # even if it is disabled in config_psa_crypto_no_ecp_at_all(). As a
                    # consequence compressed points are supported in the reference
                    # component but not in the accelerated one, so they should be skipped
                    # while checking driver's coverage.
                    'Parse EC Key #10a (SEC1 PEM, secp384r1, compressed)',
                    'Parse EC Key #11a (SEC1 PEM, secp521r1, compressed)',
                    'Parse EC Key #12a (SEC1 PEM, bp256r1, compressed)',
                    'Parse EC Key #13a (SEC1 PEM, bp384r1, compressed)',
                    'Parse EC Key #14a (SEC1 PEM, bp512r1, compressed)',
                    'Parse EC Key #2a (SEC1 PEM, secp192r1, compressed)',
                    'Parse EC Key #8a (SEC1 PEM, secp224r1, compressed)',
                    'Parse EC Key #9a (SEC1 PEM, secp256r1, compressed)',
                    'Parse Public EC Key #2a (RFC 5480, PEM, secp192r1, compressed)',
                    'Parse Public EC Key #3a (RFC 5480, secp224r1, compressed)',
                    'Parse Public EC Key #4a (RFC 5480, secp256r1, compressed)',
                    'Parse Public EC Key #5a (RFC 5480, secp384r1, compressed)',
                    'Parse Public EC Key #6a (RFC 5480, secp521r1, compressed)',
                    'Parse Public EC Key #7a (RFC 5480, brainpoolP256r1, compressed)',
                    'Parse Public EC Key #8a (RFC 5480, brainpoolP384r1, compressed)',
                    'Parse Public EC Key #9a (RFC 5480, brainpoolP512r1, compressed)',
                ],
            }
        }
    },
    'analyze_driver_vs_reference_ffdh_alg': {
        'test_function': do_analyze_driver_vs_reference,
        'args': {
            'component_ref': 'test_psa_crypto_config_reference_ffdh',
            'component_driver': 'test_psa_crypto_config_accel_ffdh',
            'ignored_suites': ['dhm'],
            'ignored_tests': {'test_suite_pkparse': ['DH group family: RFC 7919']}
        }
    },
}

def main():
    try:
        parser = argparse.ArgumentParser(description=__doc__)
        parser.add_argument('outcomes', metavar='OUTCOMES.CSV',
                            help='Outcome file to analyze')
        parser.add_argument('task', default='all', nargs='?',
                            help='Analysis to be done. By default, run all tasks. '
                                 'With one or more TASK, run only those. '
                                 'TASK can be the name of a single task or '
                                 'comma/space-separated list of tasks. ')
        parser.add_argument('--list', action='store_true',
                            help='List all available tasks and exit.')
        options = parser.parse_args()

        if options.list:
            for task in TASKS:
                Results.log(task)
            sys.exit(0)

        result = True

        if options.task == 'all':
            tasks = TASKS.keys()
        else:
            tasks = re.split(r'[, ]+', options.task)

            for task in tasks:
                if task not in TASKS:
                    Results.log('Error: invalid task: {}'.format(task))
                    sys.exit(1)

        for task in TASKS:
            if task in tasks:
                if not TASKS[task]['test_function'](options.outcomes, TASKS[task]['args']):
                    result = False

        if result is False:
            sys.exit(1)
        Results.log("SUCCESS :-)")
    except Exception: # pylint: disable=broad-except
        # Print the backtrace and exit explicitly with our chosen status.
        traceback.print_exc()
        sys.exit(120)

if __name__ == '__main__':
    main()