def move_mergeable_units(merge_directory, corpus_directory):
"""Move new units in |merge_directory| into |corpus_directory|."""
initial_units = set(
os.path.basename(filename)
for filename in shell.get_files_list(corpus_directory))
for unit_path in shell.get_files_list(merge_directory):
unit_name = os.path.basename(unit_path)
if unit_name in initial_units and is_sha1_hash(unit_name):
continue
dest_path = os.path.join(corpus_directory, unit_name)
shell.move(unit_path, dest_path)
def legalize_corpus_files(directory):
"""Convert the name of every corpus file in |directory| to a name that is
allowed on Windows."""
# Iterate through return value of legalize_filenames to convert every
# filename.
files_list = shell.get_files_list(directory)
legalize_filenames(files_list)
def find_mutator_plugin():
"""Sets LD_PRELOAD to the path of a usable mutator plugin shared object.
This should only be called after a call to get_mutator_plugin."""
paths = shell.get_files_list(_get_mutator_plugins_unpacked_dir())
# This function should not be called unless there is an unpacked plugin.
for path in paths:
if os.path.basename(path) == MUTATOR_SHARED_OBJECT_FILENAME:
return path
return None
def generate_new_testcase_mutations_using_radamsa(
corpus_directory, new_testcase_mutations_directory,
generation_timeout):
"""Generate new testcase mutations based on Radamsa."""
radamsa_path = get_radamsa_path()
if not radamsa_path:
# Mutations using radamsa are not supported on current platform, bail out.
return
radamsa_runner = new_process.ProcessRunner(radamsa_path)
files_list = shell.get_files_list(corpus_directory)
filtered_files_list = [
f for f in files_list if os.path.getsize(f) <= CORPUS_INPUT_SIZE_LIMIT
]
if not filtered_files_list:
# No mutations to do on an empty corpus or one with very large files.
return
old_corpus_size = shell.get_directory_file_count(
new_testcase_mutations_directory)
expected_completion_time = time.time() + generation_timeout
for i in range(RADAMSA_MUTATIONS):
original_file_path = random_choice(filtered_files_list)
original_filename = os.path.basename(original_file_path)
output_path = os.path.join(
new_testcase_mutations_directory,
get_radamsa_output_filename(original_filename, i))
result = radamsa_runner.run_and_wait(
['-o', output_path, original_file_path], timeout=RADAMSA_TIMEOUT)
if (os.path.exists(output_path)
and os.path.getsize(output_path) > CORPUS_INPUT_SIZE_LIMIT):
# Skip large files to avoid furthur mutations and impact fuzzing
# efficiency.
shell.remove_file(output_path)
elif result.return_code or result.timed_out:
logs.log_warn('Radamsa failed to mutate or timed out.',
output=result.output)
# Check if we exceeded our timeout. If yes, do no more mutations and break.
if time.time() > expected_completion_time:
break
new_corpus_size = shell.get_directory_file_count(
new_testcase_mutations_directory)
logs.log('Added %d tests using Radamsa mutations.' %
(new_corpus_size - old_corpus_size))
def unpack_seed_corpus_if_needed(
fuzz_target_path,
corpus_directory,
max_bytes=float("inf"),
force_unpack=False,
max_files_for_unpack=MAX_FILES_FOR_UNPACK,
):
"""If seed corpus available, unpack it into the corpus directory if needed,
ie: if corpus exists and either |force_unpack| is True, or the number of files
in corpus_directory is less than |max_files_for_unpack|. Uses
|fuzz_target_path| to find the seed corpus. If max_bytes is specified, then
seed corpus files larger than |max_bytes| will not be unpacked.
"""
seed_corpus_archive_path = get_seed_corpus_path(fuzz_target_path)
if not seed_corpus_archive_path:
return
num_corpus_files = len(shell.get_files_list(corpus_directory))
if not force_unpack and num_corpus_files > max_files_for_unpack:
return
if force_unpack:
logs.log("Forced unpack: %s." % seed_corpus_archive_path)
archive_iterator = archive.iterator(seed_corpus_archive_path)
# Unpack seed corpus recursively into the root of the main corpus directory.
idx = 0
for seed_corpus_file in archive_iterator:
# Ignore directories.
if seed_corpus_file.name.endswith("/"):
continue
# Allow callers to opt-out of unpacking large files.
if seed_corpus_file.size > max_bytes:
continue
output_filename = "%016d" % idx
output_file_path = os.path.join(corpus_directory, output_filename)
with open(output_file_path, "wb") as file_handle:
shutil.copyfileobj(seed_corpus_file.handle, file_handle)
idx += 1
logs.log("Unarchiving %d files from seed corpus %s." %
(idx, seed_corpus_archive_path))
def generate_new_testcase_mutations_using_radamsa(
corpus_directory, new_testcase_mutations_directory,
expected_completion_time):
"""Generate new testcase mutations based on Radamsa."""
radamsa_path = get_radamsa_path()
if not radamsa_path:
# Mutations using radamsa are not supported on current platform, bail out.
return
radamsa_runner = new_process.ProcessRunner(radamsa_path)
files_list = shell.get_files_list(corpus_directory)
if not files_list:
# No mutations to do on an empty corpus, bail out.
return
old_corpus_size = shell.get_directory_file_count(
new_testcase_mutations_directory)
for i in range(RADAMSA_MUTATIONS):
original_file_path = engine_common.random_choice(files_list)
original_filename = os.path.basename(original_file_path)
output_path = os.path.join(
new_testcase_mutations_directory,
'radamsa-%08d-%s' % (i + 1, original_filename))
result = radamsa_runner.run_and_wait(
['-o', output_path, original_file_path], timeout=RADAMSA_TIMEOUT)
if result.return_code or result.timed_out:
logs.log_error('Radamsa failed to mutate or timed out.',
output=result.output)
# Check if we exceeded our timeout. If yes, do no more mutations and break.
if time.time() > expected_completion_time:
break
new_corpus_size = shell.get_directory_file_count(
new_testcase_mutations_directory)
logs.log('Added %d tests using Radamsa mutations.' %
(new_corpus_size - old_corpus_size))
