import logging import logging.config import requests import features import time import os import random from endpoints.notificationhelper import spawn_notification from collections import defaultdict from peewee import JOIN_LEFT_OUTER from app import app, config_provider, storage, OVERRIDE_CONFIG_DIRECTORY, secscan_api from workers.worker import Worker from data.database import (Image, ImageStorage, ImageStorageLocation, ImageStoragePlacement, db_random_func, UseThenDisconnect, RepositoryTag, Repository, ExternalNotificationEvent, RepositoryNotification) from util.secscan.api import SecurityConfigValidator logger = logging.getLogger(__name__) BATCH_SIZE = 20 INDEXING_INTERVAL = 10 API_METHOD_INSERT = '/v1/layers' API_METHOD_VERSION = '/v1/versions/engine' def _get_images_to_export_list(version): Parent = Image.alias() ParentImageStorage = ImageStorage.alias() rimages = [] # Collect the images without parents candidates = (Image .select(Image.id, Image.docker_image_id, ImageStorage.uuid) .join(ImageStorage) .where(Image.security_indexed_engine < version, Image.parent >> None, ImageStorage.uploading == False) .limit(BATCH_SIZE*10) .alias('candidates')) images = (Image .select(candidates.c.id, candidates.c.docker_image_id, candidates.c.uuid) .from_(candidates) .order_by(db_random_func()) .tuples() .limit(BATCH_SIZE)) for image in images: rimages.append({'image_id': image[0], 'docker_image_id': image[1], 'storage_uuid': image[2], 'parent_docker_image_id': None, 'parent_storage_uuid': None}) # Collect the images with analyzed parents. candidates = (Image .select(Image.id, Image.docker_image_id, ImageStorage.uuid, Parent.docker_image_id.alias('parent_docker_image_id'), ParentImageStorage.uuid.alias('parent_storage_uuid')) .join(Parent, on=(Image.parent == Parent.id)) .join(ParentImageStorage, on=(ParentImageStorage.id == Parent.storage)) .switch(Image) .join(ImageStorage) .where(Image.security_indexed_engine < version, Parent.security_indexed == True, Parent.security_indexed_engine >= version, ImageStorage.uploading == False) .limit(BATCH_SIZE*10) .alias('candidates')) images = (Image .select(candidates.c.id, candidates.c.docker_image_id, candidates.c.uuid, candidates.c.parent_docker_image_id, candidates.c.parent_storage_uuid) .from_(candidates) .order_by(db_random_func()) .tuples() .limit(BATCH_SIZE)) for image in images: rimages.append({'image_id': image[0], 'docker_image_id': image[1], 'storage_uuid': image[2], 'parent_docker_image_id': image[3], 'parent_storage_uuid': image[4]}) # Shuffle the images, otherwise the images without parents will always be on the top random.shuffle(rimages) return rimages def _get_storage_locations(uuid): query = (ImageStoragePlacement .select() .join(ImageStorageLocation) .switch(ImageStoragePlacement) .join(ImageStorage, JOIN_LEFT_OUTER) .where(ImageStorage.uuid == uuid)) return [location.location.name for location in query] def _update_image(image, indexed, version): query = (Image .select() .join(ImageStorage) .where(Image.docker_image_id == image['docker_image_id'], ImageStorage.uuid == image['storage_uuid'])) (Image .update(security_indexed=indexed, security_indexed_engine=version) .where(Image.id << [row.id for row in query]) .execute()) class SecurityWorker(Worker): def __init__(self): super(SecurityWorker, self).__init__() validator = SecurityConfigValidator(app, config_provider) if validator.valid(): secscan_config = app.config.get('SECURITY_SCANNER') self._api = secscan_config['ENDPOINT'] self._target_version = secscan_config['ENGINE_VERSION_TARGET'] self._default_storage_locations = app.config['DISTRIBUTED_STORAGE_PREFERENCE'] self._cert = validator.cert() self._keys = validator.keypair() self.add_operation(self._index_images, INDEXING_INTERVAL) logger.warning('Failed to validate security scan configuration') def _get_image_url(self, image): """ Gets the download URL for an image and if the storage doesn't exist, marks the image as unindexed. """ path = storage.image_layer_path(image['storage_uuid']) locations = self._default_storage_locations if not storage.exists(locations, path): locations = _get_storage_locations(image['storage_uuid']) if not storage.exists(locations, path): logger.warning('Could not find a valid location to download layer %s', image['docker_image_id']+'.'+image['storage_uuid']) _update_image(image, False, self._target_version) return None uri = storage.get_direct_download_url(locations, path) if uri is None: # Handle local storage local_storage_enabled = False for storage_type, _ in app.config.get('DISTRIBUTED_STORAGE_CONFIG', {}).values(): if storage_type == 'LocalStorage': local_storage_enabled = True if local_storage_enabled: uri = path else: logger.warning('Could not get image URL and local storage was not enabled') return None return uri def _new_request(self, image): url = self._get_image_url(image) if url is None: return None request = { 'ID': '%s.%s' % (image['docker_image_id'], image['storage_uuid']), 'Path': url, } if image['parent_docker_image_id'] is not None and image['parent_storage_uuid'] is not None: request['ParentID'] = '%s.%s' % (image['parent_docker_image_id'], image['parent_storage_uuid']) return request def _analyze_image(self, image): """ Analyzes an image by passing it to Clair. Returns the vulnerabilities detected (if any) or None on error. """ request = self._new_request(image) if request is None: return None # Analyze the image. try: logger.info('Analyzing %s', request['ID']) # Using invalid certificates doesn't return proper errors because of # https://github.com/shazow/urllib3/issues/556 httpResponse = requests.post(self._api + API_METHOD_INSERT, json=request, cert=self._keys, verify=self._cert) jsonResponse = httpResponse.json() except (requests.exceptions.RequestException, ValueError): logger.exception('An exception occurred when analyzing layer ID %s', request['ID']) return None # Handle any errors from the security scanner. if httpResponse.status_code != 201: if 'OS and/or package manager are not supported' in jsonResponse.get('Message', ''): # The current engine could not index this layer logger.warning('A warning event occurred when analyzing layer ID %s : %s', request['ID'], jsonResponse['Message']) # Hopefully, there is no version lower than the target one running _update_image(image, False, self._target_version) else: logger.warning('Got non-201 when analyzing layer ID %s: %s', request['ID'], jsonResponse) return None # Verify that the version matches. api_version = jsonResponse['Version'] if api_version < self._target_version: logger.warning('An engine runs on version %d but the target version is %d') # Mark the image as analyzed. logger.debug('Layer %s analyzed successfully; Loading vulnerabilities for layer', image['image_id']) _update_image(image, True, api_version) # Lookup the vulnerabilities for the image, now that it is analyzed. try: response = secscan_api.call('layers/%s/vulnerabilities', None, request['ID']) logger.debug('Got response %s for vulnerabilities for layer %s', response.status_code, image['image_id']) if response.status_code == 404: return None except (requests.exceptions.RequestException, ValueError): logger.exception('Failed to get vulnerability response for %s', image['image_id']) return None return response.json() def _index_images(self): logger.debug('Started indexing') with UseThenDisconnect(app.config): while True: # Lookup the images to index. images = [] try: logger.debug('Looking up images to index') images = _get_images_to_export_list(self._target_version) except Image.DoesNotExist: pass if not images: logger.debug('No more images left to analyze') return logger.debug('Found %d images to index', len(images)) for image in images: # Analyze the image, retrieving the vulnerabilities (if any). sec_data = self._analyze_image(image) if sec_data is None: continue if not sec_data['Vulnerabilities']: continue # Dispatch events for any detected vulnerabilities logger.debug('Got vulnerabilities for layer %s: %s', image['image_id'], sec_data) event = ExternalNotificationEvent.get(name='vulnerability_found') matching = (RepositoryTag .select(RepositoryTag, Repository) .distinct() .join(Repository) .join(RepositoryNotification) .where(RepositoryNotification.event == event, RepositoryTag.image == image['image_id'], RepositoryTag.hidden == False, RepositoryTag.lifetime_end_ts >> None)) repository_map = defaultdict(list) for tag in matching: repository_map[tag.repository_id].append(tag) for repository_id in repository_map: tags = repository_map[repository_id] for vuln in sec_data['Vulnerabilities']: event_data = { 'tags': [tag.name for tag in tags], 'vulnerability': { 'id': vuln['ID'], 'description': vuln['Description'], 'link': vuln['Link'], 'priority': vuln['Priority'], }, } spawn_notification(tags[0].repository, 'vulnerability_found', event_data) if __name__ == '__main__': if not features.SECURITY_SCANNER: logger.debug('Security scanner disabled; skipping SecurityWorker') while True: time.sleep(100000) logging.config.fileConfig('conf/logging_debug.conf', disable_existing_loggers=False) worker = SecurityWorker() worker.start()