diff options
Diffstat (limited to 'Lib/test/test_hashlib.py')
| -rw-r--r-- | Lib/test/test_hashlib.py | 139 |
1 files changed, 97 insertions, 42 deletions
diff --git a/Lib/test/test_hashlib.py b/Lib/test/test_hashlib.py index b83ae181718..6e4e41e4fbb 100644 --- a/Lib/test/test_hashlib.py +++ b/Lib/test/test_hashlib.py @@ -208,6 +208,11 @@ class HashLibTestCase(unittest.TestCase): return itertools.chain.from_iterable(constructors) @property + def shake_constructors(self): + for shake_name in self.shakes: + yield from self.constructors_to_test.get(shake_name, ()) + + @property def is_fips_mode(self): return get_fips_mode() @@ -376,21 +381,50 @@ class HashLibTestCase(unittest.TestCase): self.assertIsInstance(h.digest(), bytes) self.assertEqual(hexstr(h.digest()), h.hexdigest()) - def test_digest_length_overflow(self): - # See issue #34922 - large_sizes = (2**29, 2**32-10, 2**32+10, 2**61, 2**64-10, 2**64+10) - for cons in self.hash_constructors: - h = cons(usedforsecurity=False) - if h.name not in self.shakes: - continue - if HASH is not None and isinstance(h, HASH): - # _hashopenssl's take a size_t - continue - for digest in h.digest, h.hexdigest: - self.assertRaises(ValueError, digest, -10) - for length in large_sizes: - with self.assertRaises((ValueError, OverflowError)): - digest(length) + def test_shakes_zero_digest_length(self): + for constructor in self.shake_constructors: + with self.subTest(constructor=constructor): + h = constructor(b'abcdef', usedforsecurity=False) + self.assertEqual(h.digest(0), b'') + self.assertEqual(h.hexdigest(0), '') + + def test_shakes_invalid_digest_length(self): + # See https://github.com/python/cpython/issues/79103. + for constructor in self.shake_constructors: + with self.subTest(constructor=constructor): + h = constructor(usedforsecurity=False) + # Note: digest() and hexdigest() take a signed input and + # raise if it is negative; the rationale is that we use + # internally PyBytes_FromStringAndSize() and _Py_strhex() + # which both take a Py_ssize_t. + for negative_size in (-1, -10, -(1 << 31), -sys.maxsize): + self.assertRaises(ValueError, h.digest, negative_size) + self.assertRaises(ValueError, h.hexdigest, negative_size) + + def test_shakes_overflow_digest_length(self): + # See https://github.com/python/cpython/issues/135759. + + exc_types = (OverflowError, ValueError) + # HACL* accepts an 'uint32_t' while OpenSSL accepts a 'size_t'. + openssl_overflown_sizes = (sys.maxsize + 1, 2 * sys.maxsize) + # https://github.com/python/cpython/issues/79103 restricts + # the accepted built-in lengths to 2 ** 29, even if OpenSSL + # accepts such lengths. + builtin_overflown_sizes = openssl_overflown_sizes + ( + 2 ** 29, 2 ** 32 - 10, 2 ** 32, 2 ** 32 + 10, + 2 ** 61, 2 ** 64 - 10, 2 ** 64, 2 ** 64 + 10, + ) + + for constructor in self.shake_constructors: + with self.subTest(constructor=constructor): + h = constructor(usedforsecurity=False) + if HASH is not None and isinstance(h, HASH): + overflown_sizes = openssl_overflown_sizes + else: + overflown_sizes = builtin_overflown_sizes + for invalid_size in overflown_sizes: + self.assertRaises(exc_types, h.digest, invalid_size) + self.assertRaises(exc_types, h.hexdigest, invalid_size) def test_name_attribute(self): for cons in self.hash_constructors: @@ -1009,49 +1043,67 @@ class HashLibTestCase(unittest.TestCase): def test_sha256_gil(self): gil_minsize = hashlib_helper.find_gil_minsize(['_sha2', '_hashlib']) + data = b'1' + b'#' * gil_minsize + b'1' + expected = hashlib.sha256(data).hexdigest() + m = hashlib.sha256() m.update(b'1') m.update(b'#' * gil_minsize) m.update(b'1') - self.assertEqual( - m.hexdigest(), - '1cfceca95989f51f658e3f3ffe7f1cd43726c9e088c13ee10b46f57cef135b94' - ) + self.assertEqual(m.hexdigest(), expected) - m = hashlib.sha256(b'1' + b'#' * gil_minsize + b'1') - self.assertEqual( - m.hexdigest(), - '1cfceca95989f51f658e3f3ffe7f1cd43726c9e088c13ee10b46f57cef135b94' - ) + @threading_helper.reap_threads + @threading_helper.requires_working_threading() + def test_threaded_hashing_fast(self): + # Same as test_threaded_hashing_slow() but only tests some functions + # since otherwise test_hashlib.py becomes too slow during development. + for name in ['md5', 'sha1', 'sha256', 'sha3_256', 'blake2s']: + if constructor := getattr(hashlib, name, None): + with self.subTest(name): + self.do_test_threaded_hashing(constructor, is_shake=False) + if shake_128 := getattr(hashlib, 'shake_128', None): + self.do_test_threaded_hashing(shake_128, is_shake=True) + @requires_resource('cpu') @threading_helper.reap_threads @threading_helper.requires_working_threading() - def test_threaded_hashing(self): + def test_threaded_hashing_slow(self): + for algorithm, constructors in self.constructors_to_test.items(): + is_shake = algorithm in self.shakes + for constructor in constructors: + with self.subTest(constructor.__name__, is_shake=is_shake): + self.do_test_threaded_hashing(constructor, is_shake) + + def do_test_threaded_hashing(self, constructor, is_shake): # Updating the same hash object from several threads at once # using data chunk sizes containing the same byte sequences. # # If the internal locks are working to prevent multiple # updates on the same object from running at once, the resulting # hash will be the same as doing it single threaded upfront. - hasher = hashlib.sha1() - num_threads = 5 - smallest_data = b'swineflu' - data = smallest_data * 200000 - expected_hash = hashlib.sha1(data*num_threads).hexdigest() - - def hash_in_chunks(chunk_size): - index = 0 - while index < len(data): - hasher.update(data[index:index + chunk_size]) - index += chunk_size + + # The data to hash has length s|M|q^N and the chunk size for the i-th + # thread is s|M|q^(N-i), where N is the number of threads, M is a fixed + # message of small length, and s >= 1 and q >= 2 are small integers. + smallest_size, num_threads, s, q = 8, 5, 2, 10 + + smallest_data = os.urandom(smallest_size) + data = s * smallest_data * (q ** num_threads) + + h1 = constructor(usedforsecurity=False) + h2 = constructor(data * num_threads, usedforsecurity=False) + + def update(chunk_size): + for index in range(0, len(data), chunk_size): + h1.update(data[index:index + chunk_size]) threads = [] - for threadnum in range(num_threads): - chunk_size = len(data) // (10 ** threadnum) + for thread_num in range(num_threads): + # chunk_size = len(data) // (q ** thread_num) + chunk_size = s * smallest_size * q ** (num_threads - thread_num) self.assertGreater(chunk_size, 0) - self.assertEqual(chunk_size % len(smallest_data), 0) - thread = threading.Thread(target=hash_in_chunks, - args=(chunk_size,)) + self.assertEqual(chunk_size % smallest_size, 0) + thread = threading.Thread(target=update, args=(chunk_size,)) threads.append(thread) for thread in threads: @@ -1059,7 +1111,10 @@ class HashLibTestCase(unittest.TestCase): for thread in threads: thread.join() - self.assertEqual(expected_hash, hasher.hexdigest()) + if is_shake: + self.assertEqual(h1.hexdigest(16), h2.hexdigest(16)) + else: + self.assertEqual(h1.hexdigest(), h2.hexdigest()) def test_get_fips_mode(self): fips_mode = self.is_fips_mode |