def _wait_for_operation(response, project, zone):
"""Wait for the given operation to complete."""
if 'status' in response and response['status'] == 'DONE':
return
if 'kind' not in response or response['kind'] != 'compute#operation':
logs.log_error('Compute api response not an operation.')
return
api = _get_api()
operation = response['name']
start_time = datetime.datetime.utcnow()
while not dates.time_has_expired(start_time, seconds=OPERATION_TIMEOUT):
operation_func = api.zoneOperations().get(
operation=operation, project=project, zone=zone)
response = _execute_api_call_with_retries(operation_func)
if 'status' not in response:
logs.log_error('Invalid compute engine operation %s.' % str(operation))
return
if response['status'] == 'DONE':
return
time.sleep(POLL_INTERVAL)
logs.log_error('Compute engine operation %s timed out.' % str(operation))
def check_public_testcase(self, blob_info, testcase):
"""Check public testcase."""
if blob_info.key() != testcase.minimized_keys:
return False
if not testcase.bug_information:
return False
issue_tracker = issue_tracker_utils.get_issue_tracker_for_testcase(
testcase)
issue = issue_tracker.get_issue(testcase.bug_information)
if not issue:
return False
# If the issue is explicitly marked as view restricted to committers only
# (OSS-Fuzz only), then don't allow public download.
if 'restrict-view-commit' in issue.labels:
return False
# For OSS-Fuzz, delay the disclosure of the reproducer by 30 days.
# If the deadline had previously exceeded, the reproducer was made public
# already so exclude that case.
if (utils.is_oss_fuzz() and 'deadline-exceeded' not in issue.labels
and issue.closed_time and not dates.time_has_expired(
issue.closed_time, days=_OSS_FUZZ_REPRODUCER_DELAY)):
return False
return True
def wait_until_good_state():
"""Check battery and make sure it is charged beyond minimum level and
temperature thresholds."""
# Battery levels are not applicable on GCE.
if environment.is_android_cuttlefish() or settings.is_automotive():
return
# Make sure device is online.
adb.wait_for_device()
# Skip battery check if done recently.
last_battery_check_time = persistent_cache.get_value(
LAST_BATTERY_CHECK_TIME_KEY,
constructor=datetime.datetime.utcfromtimestamp)
if last_battery_check_time and not dates.time_has_expired(
last_battery_check_time, seconds=BATTERY_CHECK_INTERVAL):
return
# Initialize variables.
battery_level_threshold = environment.get_value('LOW_BATTERY_LEVEL_THRESHOLD',
LOW_BATTERY_LEVEL_THRESHOLD)
battery_temperature_threshold = environment.get_value(
'MAX_BATTERY_TEMPERATURE_THRESHOLD', MAX_BATTERY_TEMPERATURE_THRESHOLD)
device_restarted = False
while True:
battery_information = get_battery_level_and_temperature()
if battery_information is None:
logs.log_error('Failed to get battery information, skipping check.')
return
battery_level = battery_information['level']
battery_temperature = battery_information['temperature']
logs.log('Battery information: level (%d%%), temperature (%.1f celsius).' %
(battery_level, battery_temperature))
if (battery_level >= battery_level_threshold and
battery_temperature <= battery_temperature_threshold):
persistent_cache.set_value(LAST_BATTERY_CHECK_TIME_KEY, time.time())
return
logs.log('Battery in bad battery state, putting device in sleep mode.')
if not device_restarted:
adb.reboot()
device_restarted = True
# Change thresholds to expected levels (only if they were below minimum
# thresholds).
if battery_level < battery_level_threshold:
battery_level_threshold = EXPECTED_BATTERY_LEVEL
if battery_temperature > battery_temperature_threshold:
battery_temperature_threshold = EXPECTED_BATTERY_TEMPERATURE
# Stopping shell should help with shutting off a lot of services that would
# otherwise use up the battery. However, we need to turn it back on to get
# battery status information.
adb.stop_shell()
time.sleep(BATTERY_CHARGE_INTERVAL)
adb.start_shell()
def update_tests_if_needed():
"""Updates layout tests every day."""
data_directory = environment.get_value('FUZZ_DATA')
error_occured = False
expected_task_duration = 60 * 60 # 1 hour.
retry_limit = environment.get_value('FAIL_RETRIES')
temp_archive = os.path.join(data_directory, 'temp.zip')
tests_url = environment.get_value('WEB_TESTS_URL')
# Check if we have a valid tests url.
if not tests_url:
return
# Layout test updates are usually disabled to speedup local testing.
if environment.get_value('LOCAL_DEVELOPMENT'):
return
# |UPDATE_WEB_TESTS| env variable can be used to control our update behavior.
if not environment.get_value('UPDATE_WEB_TESTS'):
return
last_modified_time = persistent_cache.get_value(
TESTS_LAST_UPDATE_KEY, constructor=datetime.datetime.utcfromtimestamp)
if (last_modified_time is not None
and not dates.time_has_expired(last_modified_time,
days=TESTS_UPDATE_INTERVAL_DAYS)):
return
logs.log('Updating layout tests.')
tasks.track_task_start(tasks.Task('update_tests', '', ''),
expected_task_duration)
# Download and unpack the tests archive.
for _ in range(retry_limit):
try:
shell.remove_directory(data_directory, recreate=True)
storage.copy_file_from(tests_url, temp_archive)
archive.unpack(temp_archive, data_directory, trusted=True)
shell.remove_file(temp_archive)
error_occured = False
break
except:
logs.log_error(
'Could not retrieve and unpack layout tests archive. Retrying.'
)
error_occured = True
if not error_occured:
persistent_cache.set_value(TESTS_LAST_UPDATE_KEY,
time.time(),
persist_across_reboots=True)
tasks.track_task_end()
def beat(previous_state, log_filename):
"""Run a cycle of heartbeat checks to ensure bot is running."""
# Handle case when run_bot.py script is stuck. If yes, kill its process.
task_end_time = tasks.get_task_end_time()
if psutil and task_end_time and dates.time_has_expired(
task_end_time, seconds=tasks.TASK_COMPLETION_BUFFER):
# Get absolute path to |run_bot| script. We use this to identify unique
# instances of bot running on a particular host.
startup_scripts_directory = environment.get_startup_scripts_directory()
bot_file_path = os.path.join(startup_scripts_directory, 'run_bot')
for process in psutil.process_iter():
try:
command_line = ' '.join(process.cmdline())
except (psutil.AccessDenied, psutil.NoSuchProcess, OSError):
continue
# Find the process running the main bot script.
if bot_file_path not in command_line:
continue
process_id = process.pid
logs.log('Killing stale bot (pid %d) which seems to have stuck.' %
process_id)
try:
process_handler.terminate_root_and_child_processes(process_id)
except Exception:
logs.log_error('Failed to terminate stale bot processes.')
# Minor cleanup to avoid disk space issues on bot restart.
process_handler.terminate_stale_application_instances()
shell.clear_temp_directory()
shell.clear_testcase_directories()
# Concerned stale processes should be killed. Now, delete the stale task.
tasks.track_task_end()
# Figure out when the log file was last modified.
try:
current_state = str(os.path.getmtime(log_filename))
except Exception:
current_state = None
# Only update the heartbeat if the log file was modified.
if current_state and current_state != previous_state:
# Try updating the heartbeat. If an error occurs, just
# wait and return None.
if not data_handler.update_heartbeat():
return None
# Heartbeat is successfully updated.
return current_state
def add_test_accounts_if_needed():
"""Add test account to work with GmsCore, etc."""
last_test_account_check_time = persistent_cache.get_value(
constants.LAST_TEST_ACCOUNT_CHECK_KEY,
constructor=datetime.datetime.utcfromtimestamp)
needs_test_account_update = (last_test_account_check_time is None
or dates.time_has_expired(
last_test_account_check_time,
seconds=ADD_TEST_ACCOUNT_CHECK_INTERVAL))
if not needs_test_account_update:
return
config = db_config.get()
if not config:
return
test_account_email = config.test_account_email
test_account_password = config.test_account_password
if not test_account_email or not test_account_password:
return
adb.run_as_root()
wifi.configure(force_enable=True)
if not app.is_installed(ADD_TEST_ACCOUNT_PKG_NAME):
logs.log('Installing helper apk for adding test account.')
android_directory = environment.get_platform_resources_directory()
add_test_account_apk_path = os.path.join(android_directory,
ADD_TEST_ACCOUNT_APK_NAME)
app.install(add_test_account_apk_path)
logs.log('Trying to add test account.')
output = adb.run_shell_command(
'am instrument -e account %s -e password %s -w %s' %
(test_account_email, test_account_password,
ADD_TEST_ACCOUNT_CALL_PATH),
timeout=ADD_TEST_ACCOUNT_TIMEOUT)
if not output or test_account_email not in output:
logs.log('Failed to add test account, probably due to wifi issues.')
return
logs.log('Test account added successfully.')
persistent_cache.set_value(constants.LAST_TEST_ACCOUNT_CHECK_KEY,
time.time())
def _is_data_bundle_up_to_date(data_bundle, data_bundle_directory):
"""Return true if the data bundle is up to date, false otherwise."""
sync_file_path = _get_data_bundle_sync_file_path(data_bundle_directory)
if environment.is_trusted_host() and data_bundle.sync_to_worker:
from clusterfuzz._internal.bot.untrusted_runner import file_host
worker_sync_file_path = file_host.rebase_to_worker_root(sync_file_path)
shell.remove_file(sync_file_path)
file_host.copy_file_from_worker(worker_sync_file_path, sync_file_path)
if not os.path.exists(sync_file_path):
return False
last_sync_time = datetime.datetime.utcfromtimestamp(
utils.read_data_from_file(sync_file_path))
# Check if we recently synced.
if not dates.time_has_expired(
last_sync_time, seconds=_DATA_BUNDLE_SYNC_INTERVAL_IN_SECONDS):
return True
# For search index data bundle, we don't sync them from bucket. Instead, we
# rely on the fuzzer to generate testcases periodically.
if _is_search_index_data_bundle(data_bundle.name):
return False
# Check when the bucket url had last updates. If no new updates, no need to
# update directory.
bucket_url = data_handler.get_data_bundle_bucket_url(data_bundle.name)
last_updated_time = storage.last_updated(bucket_url)
if last_updated_time and last_sync_time > last_updated_time:
logs.log('Data bundle %s has no new content from last sync.' %
data_bundle.name)
return True
return False
def get(self):
"""Handle a get request."""
try:
grouper.group_testcases()
except:
logs.log_error('Error occurred while grouping test cases.')
return
# Free up memory after group task run.
utils.python_gc()
# Get a list of jobs excluded from bug filing.
excluded_jobs = _get_excluded_jobs()
# Get a list of all jobs. This is used to filter testcases whose jobs have
# been removed.
all_jobs = data_handler.get_all_job_type_names()
for testcase_id in data_handler.get_open_testcase_id_iterator():
try:
testcase = data_handler.get_testcase_by_id(testcase_id)
except errors.InvalidTestcaseError:
# Already deleted.
continue
# Skip if testcase's job is removed.
if testcase.job_type not in all_jobs:
continue
# Skip if testcase's job is in exclusions list.
if testcase.job_type in excluded_jobs:
continue
# Skip if we are running progression task at this time.
if testcase.get_metadata('progression_pending'):
continue
# If the testcase has a bug filed already, no triage is needed.
if _is_bug_filed(testcase):
continue
# Check if the crash is important, i.e. it is either a reproducible crash
# or an unreproducible crash happening frequently.
if not _is_crash_important(testcase):
continue
# Require that all tasks like minimizaton, regression testing, etc have
# finished.
if not data_handler.critical_tasks_completed(testcase):
continue
# For testcases that are not part of a group, wait an additional time to
# make sure it is grouped.
# The grouper runs prior to this step in the same cron, but there is a
# window of time where new testcases can come in after the grouper starts.
# This delay needs to be longer than the maximum time the grouper can take
# to account for that.
# FIXME: In future, grouping might be dependent on regression range, so we
# would have to add an additional wait time.
if not testcase.group_id and not dates.time_has_expired(
testcase.timestamp, hours=data_types.MIN_ELAPSED_TIME_SINCE_REPORT):
continue
# If this project does not have an associated issue tracker, we cannot
# file this crash anywhere.
issue_tracker = issue_tracker_utils.get_issue_tracker_for_testcase(
testcase)
if not issue_tracker:
issue_filer.notify_issue_update(testcase, 'new')
continue
# If there are similar issues to this test case already filed or recently
# closed, skip filing a duplicate bug.
if _check_and_update_similar_bug(testcase, issue_tracker):
continue
# Clean up old triage messages that would be not applicable now.
testcase.delete_metadata(TRIAGE_MESSAGE_KEY, update_testcase=False)
# File the bug first and then create filed bug metadata.
try:
issue_filer.file_issue(testcase, issue_tracker)
except Exception as e:
logs.log_error('Failed to file issue for testcase %d.' % testcase_id)
_add_triage_message(testcase,
f'Failed to file issue due to exception: {str(e)}')
continue
_create_filed_bug_metadata(testcase)
issue_filer.notify_issue_update(testcase, 'new')
logs.log('Filed new issue %s for testcase %d.' %
(testcase.bug_information, testcase_id))
def _check_and_update_similar_bug(testcase, issue_tracker):
"""Get list of similar open issues and ones that were recently closed."""
# Get similar testcases from the same group.
similar_testcases_from_group = []
if testcase.group_id:
group_query = data_types.Testcase.query(
data_types.Testcase.group_id == testcase.group_id)
similar_testcases_from_group = ndb_utils.get_all_from_query(
group_query, batch_size=data_types.TESTCASE_ENTITY_QUERY_LIMIT // 2)
# Get testcases with the same crash params. These might not be in the a group
# if they were just fixed.
same_crash_params_query = data_types.Testcase.query(
data_types.Testcase.crash_type == testcase.crash_type,
data_types.Testcase.crash_state == testcase.crash_state,
data_types.Testcase.security_flag == testcase.security_flag,
data_types.Testcase.project_name == testcase.project_name,
data_types.Testcase.status == 'Processed')
similar_testcases_from_query = ndb_utils.get_all_from_query(
same_crash_params_query,
batch_size=data_types.TESTCASE_ENTITY_QUERY_LIMIT // 2)
for similar_testcase in itertools.chain(similar_testcases_from_group,
similar_testcases_from_query):
# Exclude ourself from comparison.
if similar_testcase.key.id() == testcase.key.id():
continue
# Exclude similar testcases without bug information.
if not similar_testcase.bug_information:
continue
# Get the issue object given its ID.
issue = issue_tracker.get_issue(similar_testcase.bug_information)
if not issue:
continue
# If the reproducible issue is not verified yet, bug is still valid and
# might be caused by non-availability of latest builds. In that case,
# don't file a new bug yet.
if similar_testcase.open and not similar_testcase.one_time_crasher_flag:
return True
# If the issue is still open, no need to file a duplicate bug.
if issue.is_open:
return True
# If the issue indicates that this crash needs to be ignored, no need to
# file another one.
policy = issue_tracker_policy.get(issue_tracker.project)
ignore_label = policy.label('ignore')
if ignore_label in issue.labels:
_add_triage_message(
testcase,
('Skipping filing a bug since similar testcase ({testcase_id}) in '
'issue ({issue_id}) is blacklisted with {ignore_label} label.'
).format(
testcase_id=similar_testcase.key.id(),
issue_id=issue.id,
ignore_label=ignore_label))
return True
# If the issue is recently closed, wait certain time period to make sure
# our fixed verification has completed.
if (issue.closed_time and not dates.time_has_expired(
issue.closed_time, hours=data_types.MIN_ELAPSED_TIME_SINCE_FIXED)):
_add_triage_message(
testcase,
('Delaying filing a bug since similar testcase '
'({testcase_id}) in issue ({issue_id}) was just fixed.').format(
testcase_id=similar_testcase.key.id(), issue_id=issue.id))
return True
return False
def flash_to_latest_build_if_needed():
"""Wipes user data, resetting the device to original factory state."""
if environment.get_value('LOCAL_DEVELOPMENT'):
# Don't reimage local development devices.
return
run_timeout = environment.get_value('RUN_TIMEOUT')
if run_timeout:
# If we have a run timeout, then we are already scheduled to bail out and
# will be probably get re-imaged. E.g. using frameworks like Tradefed.
return
# Check if a flash is needed based on last recorded flash time.
last_flash_time = persistent_cache.get_value(
constants.LAST_FLASH_TIME_KEY,
constructor=datetime.datetime.utcfromtimestamp)
needs_flash = last_flash_time is None or dates.time_has_expired(
last_flash_time, seconds=FLASH_INTERVAL)
if not needs_flash:
return
is_google_device = settings.is_google_device()
if is_google_device is None:
logs.log_error('Unable to query device. Reimaging failed.')
adb.bad_state_reached()
elif not is_google_device:
# We can't reimage these, skip.
logs.log('Non-Google device found, skipping reimage.')
return
# Check if both |BUILD_BRANCH| and |BUILD_TARGET| environment variables
# are set. If not, we don't have enough data for reimaging and hence
# we bail out.
branch = environment.get_value('BUILD_BRANCH')
target = environment.get_value('BUILD_TARGET')
if not target:
# We default to userdebug configuration.
build_params = settings.get_build_parameters()
if build_params:
target = build_params.get('target') + '-userdebug'
# Cache target in environment. This is also useful for cases when
# device is bricked and we don't have this information available.
environment.set_value('BUILD_TARGET', target)
if not branch or not target:
logs.log_warn(
'BUILD_BRANCH and BUILD_TARGET are not set, skipping reimage.')
return
image_directory = environment.get_value('IMAGES_DIR')
build_info = fetch_artifact.get_latest_artifact_info(branch, target)
if not build_info:
logs.log_error('Unable to fetch information on latest build artifact for '
'branch %s and target %s.' % (branch, target))
return
if environment.is_android_cuttlefish():
download_latest_build(build_info, FLASH_CUTTLEFISH_REGEXES, image_directory)
adb.recreate_cuttlefish_device()
adb.connect_to_cuttlefish_device()
else:
download_latest_build(build_info, FLASH_IMAGE_REGEXES, image_directory)
# We do one device flash at a time on one host, otherwise we run into
# failures and device being stuck in a bad state.
flash_lock_key_name = 'flash:%s' % socket.gethostname()
if not locks.acquire_lock(flash_lock_key_name, by_zone=True):
logs.log_error('Failed to acquire lock for reimaging, exiting.')
return
logs.log('Reimaging started.')
logs.log('Rebooting into bootloader mode.')
for _ in range(FLASH_RETRIES):
adb.run_as_root()
adb.run_command(['reboot-bootloader'])
time.sleep(FLASH_REBOOT_BOOTLOADER_WAIT)
adb.run_fastboot_command(['oem', 'off-mode-charge', '0'])
adb.run_fastboot_command(['-w', 'reboot-bootloader'])
for partition, partition_image_filename in FLASH_IMAGE_FILES:
partition_image_file_path = os.path.join(image_directory,
partition_image_filename)
adb.run_fastboot_command(
['flash', partition, partition_image_file_path])
if partition in ['bootloader', 'radio']:
adb.run_fastboot_command(['reboot-bootloader'])
# Disable ramdump to avoid capturing ramdumps during kernel crashes.
# This causes device lockup of several minutes during boot and we intend
# to analyze them ourselves.
adb.run_fastboot_command(['oem', 'ramdump', 'disable'])
adb.run_fastboot_command('reboot')
time.sleep(FLASH_REBOOT_WAIT)
if adb.get_device_state() == 'device':
break
logs.log_error('Reimaging failed, retrying.')
locks.release_lock(flash_lock_key_name, by_zone=True)
if adb.get_device_state() != 'device':
logs.log_error('Unable to find device. Reimaging failed.')
adb.bad_state_reached()
logs.log('Reimaging finished.')
# Reset all of our persistent keys after wipe.
persistent_cache.delete_value(constants.BUILD_PROP_MD5_KEY)
persistent_cache.delete_value(constants.LAST_TEST_ACCOUNT_CHECK_KEY)
persistent_cache.set_value(constants.LAST_FLASH_BUILD_KEY, build_info)
persistent_cache.set_value(constants.LAST_FLASH_TIME_KEY, time.time())
