def reload_config_info(self):
"""
Refresh the configuration values.
Reload the configuration, get configuration values from self.cfg
and set them as attributes of self.
To be called at least once.
"""
# TODO: merge this method into the constructor, there is no reason to
# keep it separated anymore.
self.cfg.load()
self.client_id = self.cfg.get('User', 'client_id')
self.username = self.cfg.get('User', 'username')
self.priv_key = self.cfg.get('Application Paths',
'client_priv_key_file')
self.host = self.cfg.get('System', 'server_hostname')
self.port = self.cfg.getint('System', 'server_port')
self.server_certificate = self.cfg.get('Application Paths',
'server_certificate')
self.storage_hostname = self.cfg.get('System', 'storage_endpoint')
self.refused_declare_max = self.cfg.getint('System',
'refused_declare_max')
self.refused_declare_waiting_time = self.cfg.getint(
'System', 'refused_declare_waiting_time')
self.commit_threshold_seconds = self.cfg.getint(
'Client', 'commit_threshold_seconds')
self.commit_threshold_operations = self.cfg.getint(
'Client', 'commit_threshold_operations')
self.commit_threshold_bytes = self.cfg.getint(
'Client', 'commit_threshold_bytes')
temp = self.cfg.get('Application Paths', 'transaction_cache_db')
self.transaction_cache = TransactionCache(temp)
self.integrity_manager = IntegrityManager(None)
is_firt_startup = self._internal_facade.is_first_startup()
self.cryptoAdapter = Adapter(self.cfg,
self.warebox,
self._input_queue,
self._lockfile_fd,
enc_dir='enc',
first_startup=is_firt_startup)
self.worker_pool = WorkerPool(self.warebox, self, self.cfg,
self.cryptoAdapter)
self.temp_dir = self.cryptoAdapter.get_enc_dir()
self._ui_controller.update_config_info(self.cfg)
def reload_config_info(self):
"""
Refresh the configuration values.
Reload the configuration, get configuration values from self.cfg
and set them as attributes of self.
To be called at least once.
"""
# TODO: merge this method into the constructor, there is no reason to
# keep it separated anymore.
self.cfg.load()
self.client_id = self.cfg.get('User', 'client_id')
self.username = self.cfg.get('User', 'username')
self.priv_key = self.cfg.get('Application Paths', 'client_priv_key_file')
self.host = self.cfg.get('System', 'server_hostname')
self.port = self.cfg.getint('System', 'server_port')
self.server_certificate = self.cfg.get('Application Paths', 'server_certificate')
self.storage_hostname = self.cfg.get('System', 'storage_endpoint')
self.refused_declare_max = self.cfg.getint(
'System', 'refused_declare_max')
self.refused_declare_waiting_time = self.cfg.getint(
'System', 'refused_declare_waiting_time')
self.commit_threshold_seconds = self.cfg.getint(
'Client', 'commit_threshold_seconds')
self.commit_threshold_operations = self.cfg.getint(
'Client', 'commit_threshold_operations')
self.commit_threshold_bytes = self.cfg.getint(
'Client', 'commit_threshold_bytes')
temp = self.cfg.get('Application Paths', 'transaction_cache_db')
self.transaction_cache = TransactionCache(temp)
self.integrity_manager = IntegrityManager(None)
is_firt_startup = self._internal_facade.is_first_startup()
self.cryptoAdapter = Adapter(self.cfg,
self.warebox,
self._input_queue,
self._lockfile_fd,
enc_dir='enc',
first_startup=is_firt_startup)
self.worker_pool = WorkerPool(self.warebox,
self,
self.cfg,
self.cryptoAdapter)
self.temp_dir = self.cryptoAdapter.get_enc_dir()
self._ui_controller.update_config_info(self.cfg)
def __init__(self,
warebox,
operation_queue,
server_session,
cfg,
cryptoAdapter,
worker_pool):
"""
@param warebox:
Instance of filerockclient.warebox.Warebox.
@param cfg:
Instance of filerockclient.config.ConfigManager.
@param operation_queue:
A threading queue, where worker receives the operations
@param worker_pool:
Instance of filerockclient.workers.worker.Worker
"""
Thread.__init__(self, name=self.__class__.__name__)
self.cfg = cfg
self.operation_queue = operation_queue
self._server_session = server_session
self.warebox = warebox
self.child = None
self.input_queue = None
self.communicationQueue = None
self.child_logger = None
self.child_logs_queue = None
self.cryptoAdapter = cryptoAdapter
self.integrity_manager = IntegrityManager(None)
self._worker_pool = worker_pool
self.must_die = threading.Event()
self.last_send = datetime.now()
self.communicationQueue = None
self.child_logs_queue = None
def __init__(self,
warebox,
operation_queue,
server_session,
cfg,
cryptoAdapter,
worker_pool):
"""
@param warebox:
Instance of filerockclient.warebox.Warebox.
@param cfg:
Instance of filerockclient.config.ConfigManager.
@param operation_queue:
A threading queue, where worker receives the operations
@param worker_pool:
Instance of filerockclient.workers.worker.Worker
"""
Thread.__init__(self, name=self.__class__.__name__)
self.cfg = cfg
self.operation_queue = operation_queue
self._server_session = server_session
self.warebox = warebox
self.child = None
self.input_queue = None
self.communicationQueue = None
self.child_logger = None
self.child_logs_queue = None
self.cryptoAdapter = cryptoAdapter
self.integrity_manager = IntegrityManager(None)
self._worker_pool = worker_pool
self.must_die = threading.Event()
self.last_send = datetime.now()
self.communicationQueue = None
self.child_logs_queue = None
class Worker(Thread):
'''A worker takes something to do and does it, and then again takes
something to do etc.
'''
def __init__(self,
warebox,
operation_queue,
server_session,
cfg,
cryptoAdapter,
worker_pool):
"""
@param warebox:
Instance of filerockclient.warebox.Warebox.
@param cfg:
Instance of filerockclient.config.ConfigManager.
@param operation_queue:
A threading queue, where worker receives the operations
@param worker_pool:
Instance of filerockclient.workers.worker.Worker
"""
Thread.__init__(self, name=self.__class__.__name__)
self.cfg = cfg
self.operation_queue = operation_queue
self._server_session = server_session
self.warebox = warebox
self.child = None
self.input_queue = None
self.communicationQueue = None
self.child_logger = None
self.child_logs_queue = None
self.cryptoAdapter = cryptoAdapter
self.integrity_manager = IntegrityManager(None)
self._worker_pool = worker_pool
self.must_die = threading.Event()
self.last_send = datetime.now()
self.communicationQueue = None
self.child_logs_queue = None
def run(self):
"""
Serves file operations until termination request is received
"""
try:
self.name += "_%s" % self.ident
self.logger = logging.getLogger("FR.%s" % self.getName())
self.logger.debug(u'Started.')
while not self._termination_requested():
self._serve_file_operations()
self.logger.debug(u"I'm terminated.")
finally:
self._terminate_child()
def _serve_file_operations(self):
"""
Blocks on operation queue until a message is received
If the message is a POISON_PILL the worker terminate it self
If the message is a non aborted file operation,
an abort handler is registered to it and the operation is handled
"""
try:
file_operation = self.operation_queue.get()
self.warebox._check_blacklisted_dir()
if file_operation == 'POISON_PILL':
self._on_poison_pill()
elif file_operation.is_aborted():
self.logger.debug(u"Got an already aborted operation, "
"giving up: %s" % file_operation)
else:
try:
self.logger.debug(u"Got an operation to handle: %s"
% file_operation)
file_operation.register_abort_handler(self.on_operation_abort)
self._handle_file_operation(file_operation)
except Exception as e:
self.logger.error(
u"Some problem occurred with the operation : %r" % e)
raise e
finally:
self.logger.debug("Releasing a worker")
self._worker_pool.release_worker()
def _handle_file_operation(self, file_operation):
if file_operation.verb == 'UPLOAD':
self._handle_upload_file_operation(file_operation)
elif file_operation.verb == 'DOWNLOAD':
self._handle_download_file_operation(file_operation)
elif file_operation.verb == 'DELETE_LOCAL':
self._handle_delete_local_file_operation(file_operation)
elif file_operation.verb == 'CREATE_DIRECTORIES':
self._handle_operation_create_directories(file_operation)
elif file_operation.verb == 'RESOLVE_DELETION_CONFLICTS':
self._handle_operation_resolve_deletion_conflicts(file_operation)
else:
self.logger.warning(u"I should not handle a '%s' operation! "
"I'm rejecting it", file_operation.verb)
file_operation.reject()
def _send_percentage(self, file_operation, status, percentage):
now = datetime.now()
delta = now - self.last_send
if ((delta.seconds + delta.microseconds/1000000.) > 0.5) or (percentage == 100):
file_operation.notify_pathname_status_change(status, {'percentage': percentage})
self.last_send = now
def _handle_network_transfer_operation(self, file_operation):
'''By locking we mantain the following invariant: if the
EventQueue tries to abort this operation due to a conflicting
operation, then EventQueue waits until this operation either
aborts or completes.
This preserves the ordering of execution for the conflicting
operations - that is, the EventQueue doesn't emit
the conflicting operation while this one is still working.
'''
with file_operation.lock:
if not file_operation.is_aborted():
self.logger.debug(u"Starting child process to handle file"
" operation: %s" % file_operation)
try:
self._spawn_child(file_operation)
self.input_queue.put(('FileOperation', file_operation))
except Exception as e:
self.logger.error(
u"Could not spawn a child process: %r" % e)
raise OperationRejection(file_operation)
else:
self.logger.debug(u"Got an already aborted operation, "
"giving up: %s" % file_operation)
return False
if file_operation.verb == 'UPLOAD':
status = PStatuses.UPLOADING
else:
status = PStatuses.DOWNLOADING
self._send_percentage(file_operation, status, 0)
termination = False
max_retry = 3
while not termination:
message, content = self.communicationQueue.get()
self.logger.debug(u'Worker send back %s with content %s'
% (message, content))
if message == 'completed':
termination = True
if file_operation.verb == 'DOWNLOAD':
return {'actual_etag': content['actual_etag']}
else:
return True
elif message == 'interrupted':
self.logger.debug(u"Child has been terminated by "
"Software Operation: %s"
% file_operation)
file_operation.abort()
termination = True
return False
elif message == 'failed':
self.logger.error(u"Child has been terminated, "
"Assuming failure for operation: %s"
% file_operation)
max_retry -= 1
if max_retry == 0:
raise OperationRejection(file_operation)
self.input_queue.put(('FileOperation', file_operation))
elif message == 'percentage':
self._send_percentage(file_operation, status, content)
elif message == 'log':
level, msg = content
self.child_logger[level](msg)
elif message == 'ShuttingDown':
self.logger.debug(u"Get a shutting down message from process")
termination = True
return False
elif message == 'DIED':
self.child = None
termination = True
raise OperationRejection(file_operation)
def _handle_upload_file_operation(self, operation):
try:
success = self._handle_network_transfer_operation(operation)
if success:
CryptoUtils.clean_env(operation, self.logger)
self.logger.debug(u"Operation has been completed "
"successfully: %s" % operation)
self.logger.info(u'Synchronized pathname: %s "%s", which '
'will be persisted after a commit'
% (operation.verb, operation.pathname))
operation.notify_pathname_status_change(PStatuses.UPLOADED,
{'percentage': 100})
operation.complete()
except Exception as e:
self.logger.error(u"Error while uploading: %r."
" Rejecting the operation: %s" % (e, operation))
operation.reject()
def _handle_operation_create_directories(self, task):
operations = sorted(task.operations, key=lambda op: op.pathname)
actual_etag = 'd41d8cd98f00b204e9800998ecf8427e'
for operation in operations:
# Check integrity
operation.notify_pathname_status_change(PStatuses.DOWNLOADING)
res = self._check_download_integrity(operation, actual_etag)
if not res['valid']:
# Detected an integrity error. Badly bad.
self._server_session.signal_download_integrity_error(
operation, res['reason'],
res['expected_etag'], res['expected_basis'],
res['actual_etag'], res['computed_basis'])
return
# The directory is valid, create it
self.warebox.make_directory(operation.pathname)
lmtime = self.warebox.get_last_modification_time(operation.pathname)
operation.lmtime = lmtime
operation.notify_pathname_status_change(PStatuses.ALIGNED)
operation.complete()
self.logger.info(u'Synchronized pathname: %s "%s"'
% (operation.verb, operation.pathname))
self.logger.debug(u"Operation has been completed "
"successfully: %s" % operation)
task.complete()
def _handle_download_file_operation(self, operation):
try:
# Note: it is a normal file, not a directory
CryptoUtils.set_temp_file(operation, self.cfg)
success = self._handle_network_transfer_operation(operation)
if success:
actual_etag = success['actual_etag']
res = self._check_download_integrity(operation, actual_etag)
if not res['valid']:
# Detected an integrity error. Badly bad.
self._server_session.signal_download_integrity_error(
operation, res['reason'],
res['expected_etag'], res['expected_basis'],
res['actual_etag'], res['computed_basis'])
return
if operation.to_decrypt:
# We have not finished yet, leaving the rest to decrypter.
# Note: the decrypter duplicates the following ending logic
self.cryptoAdapter.put(operation)
return
# It is a valid cleartext file, move it to the warebox
self.warebox.move(operation.temp_pathname,
operation.pathname,
operation.conflicted)
self.logger.debug(u"Operation has been completed "
"successfully: %s" % operation)
self.logger.info(u'Synchronized pathname: %s "%s"'
% (operation.verb, operation.pathname))
lmtime = self.warebox.get_last_modification_time(operation.pathname)
operation.lmtime = lmtime
operation.notify_pathname_status_change(PStatuses.ALIGNED)
operation.complete()
except Exception as e:
self.logger.error(u"Error while downloading: %r."
" Rejecting the operation: %s" % (e, operation))
self.logger.error(u"Stacktrace: %r" % traceback.format_exc())
operation.reject()
finally:
# Just in case the move had failed for any reason
if not operation.to_decrypt:
if operation.temp_pathname is not None \
and os.path.exists(operation.temp_pathname):
_try_remove(operation.temp_pathname, self.logger)
def _check_download_integrity(self, operation, actual_etag):
pathname = operation.pathname
proof = operation.download_info['proof']
basis = operation.download_info['trusted_basis']
self.integrity_manager.trusted_basis = basis
result = {}
result['valid'] = None
result['reason'] = None
result['expected_etag'] = operation.storage_etag
result['expected_basis'] = basis
result['actual_etag'] = actual_etag
result['computed_basis'] = None
if operation.storage_etag != actual_etag:
# Note: the etag inside the proof will be
# checked by the integrity manager.
self.logger.debug(
u"Invalid etag of download operation. "
"Expected etag %s but found %s. %s"
% (operation.storage_etag, actual_etag, operation))
result['valid'] = False
result['reason'] = "Expected etag different from actual etag"
return result
try:
self.integrity_manager.addOperation('DOWNLOAD',
pathname,
proof,
actual_etag)
result['valid'] = True
result['computed_basis'] = basis
except MalformedProofException as e:
self.logger.debug(u"Invalid proof of download operation: "
"%s. %r, %r" % (e, operation, proof))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
except WrongBasisFromProofException as e:
self.logger.debug(u"Integrity check of download operation failed."
"Basis %s was expected but the proof computed "
"%s. Error details: %s. %r, %r" %
(basis, e.operation_basis, e, operation, proof))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
result['computed_basis'] = e.operation_basis
except Exception as e:
self.logger.debug(u"Integrity check of download operation failed"
" with unknown reason: %s. %r, %r"
% (e, operation, proof))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
finally:
self.integrity_manager.clear()
return result
def _handle_delete_local_file_operation(self, task):
pathnames = sorted(task.pathname2proof.keys())
#self.logger.debug("Going to delete pathnames: %s" % pathnames)
for pathname in pathnames:
basis = task.trusted_basis
proof = task.pathname2proof[pathname]
res = self._check_deletelocal_integrity(pathname, proof, basis)
if not res['valid']:
# Detected an integrity error. Badly bad.
self._server_session.signal_deletelocal_integrity_error(
pathname, proof, res['reason'],
res['expected_basis'], res['computed_basis'])
return
roots = {}
for pathname in pathnames:
found_ancestor = False
for root in roots:
if pathname.startswith(root):
found_ancestor = True
break
if not found_ancestor:
roots[pathname] = True
try:
for pathname in roots.iterkeys():
self.warebox.delete_tree(pathname)
except Exception as e:
self.logger.error(
u"Caught an operating system exception while "
u"modifying the filesystem. Are you locking the Warebox? % r"
% e)
raise
task.complete()
def _check_deletelocal_integrity(self, pathname, proof, trusted_basis):
self.integrity_manager.trusted_basis = trusted_basis
result = {}
result['valid'] = None
result['reason'] = None
result['expected_basis'] = trusted_basis
result['computed_basis'] = None
try:
self.integrity_manager.addOperation('DELETE_LOCAL',
pathname,
proof,
None)
result['valid'] = True
result['computed_basis'] = trusted_basis
except MalformedProofException as e:
self.logger.debug(u"Invalid proof of delete_local operation: "
"%s. %s, %s" % (e, pathname, proof.raw))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
except WrongBasisFromProofException as e:
self.logger.debug(u"Integrity check of delete_local operation failed."
"Basis %s was expected but the proof computed "
"%s. Error details: %s. %r, %r" %
(trusted_basis, e.operation_basis, e, pathname, proof.raw))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
result['computed_basis'] = e.operation_basis
except Exception as e:
self.logger.debug(u"Integrity check of delete_local operation failed"
" with unknown reason: %s. %r, %r"
% (e, pathname, proof.raw))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
finally:
self.integrity_manager.clear()
return result
def _find_new_name(self, pathname):
# TODO: try harder in finding a name that is available
curr_time = datetime.now().strftime('%Y-%m-%d %H_%M_%S')
suffix = ' (Conflicted on %s)' % curr_time
if pathname.endswith('/'):
new_pathname = pathname[:-1] + suffix + '/'
else:
basename, ext = os.path.splitext(pathname)
new_pathname = basename + suffix + ext
return new_pathname
def _rename_conflicting_pathname(self, pathname, prefix=None):
new_pathname = self.warebox.rename(pathname, pathname, prefix)
return new_pathname
def _handle_operation_resolve_deletion_conflicts(self, task):
"""Solve deletion conflicts by renaming the local file to a new
pathname. The old pathname will result implicitly deleted.
Side effect on Content_to_upload to add the renamed files.
Deletion conflicts are tough to resolve. A conflicting pathname:
a) has been deleted by the server
b) has an ancestor folder that has been deleted by the server
c) both
It must be checked if it's safe leaving the file in its original
folder (that is, if it still exists).
"""
conflicts = task.deletion_conflicts
content_to_delete_locally = task.content_to_delete_locally
for pathname in conflicts:
basis = task.trusted_basis
proof = task.pathname2proof[pathname]
res = self._check_deletelocal_integrity(pathname, proof, basis)
if not res['valid']:
# Detected an integrity error. Badly bad.
self._server_session.signal_deletelocal_integrity_error(
pathname, proof, res['reason'],
res['expected_basis'], res['computed_basis'])
return
try:
backupped_folders = {}
for pathname in conflicts:
missing_ancestor_folders = filter(lambda p: pathname.startswith(p), content_to_delete_locally)
# Is it safe leaving the file in its original folder?
if len(missing_ancestor_folders) > 0:
# No, it's been deleted. Backup the whole deleted subtree
missing_ancestor_folders = sorted(missing_ancestor_folders)
highest_missing_folder = missing_ancestor_folders[0]
if not highest_missing_folder in backupped_folders:
backup_folder = self._find_new_name(highest_missing_folder)
self.warebox.make_directory(backup_folder)
backupped_folders[highest_missing_folder] = backup_folder
backup_folder = backupped_folders[highest_missing_folder]
new_pathname = pathname.replace(highest_missing_folder, backup_folder, 1)
self.warebox.make_directories_to(new_pathname)
if not self.warebox.is_directory(new_pathname):
self.warebox.rename(pathname, new_pathname)
else:
# Yes, just rename the file
new_pathname = self._rename_conflicting_pathname(pathname, 'Deleted')
self.logger.warning(
u"Conflict detected for pathname %r, which has been "
u"remotely deleted. Moved the local copy to: %r"
% (pathname, new_pathname))
task.complete()
except Exception:
self.logger.error(
u"Caught an operating system exception while modifying the "
u"filesystem. Are you locking the Warebox?")
raise
def _start_child_logger(self):
self._stop_child_logger()
self.child_logs_queue = multiprocessing.Queue()
self.child_logger = LogsReceiver(self.getName(), self.child_logs_queue)
self.child_logger.start()
if self.child_logger is None:
logger = logging.getLogger(u'FR.WorkerChild of %s' % self.getName())
self.child_logger = {
'info': logger.info,
'debug': logger.debug,
'warning': logger.warning,
'error': logger.error,
'critical': logger.critical
}
def _stop_child_logger(self):
if self.child_logger is not None:
self.child_logger.stop()
self.child_logs_queue.put(('log', ('debug', 'Die please!')))
self.child_logger.join()
self.child_logger = None
if self.child_logs_queue is not None:
self.child_logs_queue.close()
self.child_logs_queue.join_thread()
self.child_logs_queue = None
def _create_multiprocessing_queues(self):
self._destroy_multiprocessing_queues()
self.input_queue = Queue.Queue()
self.communicationQueue = Queue.Queue()
def _destroy_multiprocessing_queue(self, queue):
if queue is not None:
while not queue.empty():
queue.get_nowait()
# queue.close()
# queue.join_thread()
queue = None
def _destroy_multiprocessing_queues(self):
self._destroy_multiprocessing_queue(self.input_queue)
self._destroy_multiprocessing_queue(self.communicationQueue)
def _spawn_child(self, file_operation):
if self.child is None or not self.child.is_alive():
# self.terminationEvent = multiprocessing.Event()
self.terminationEvent = threading.Event()
self._create_multiprocessing_queues()
self._start_child_logger()
try:
self.logger.debug(u"Allocating child process to handle file "
"operation: %s" % file_operation)
self.child = WorkerChild(self.warebox,
self.input_queue,
self.communicationQueue,
self.terminationEvent,
self._send_percentage,
#self.child_logs_queue,
self.cfg,
self._worker_pool)
self.child.start()
self.logger.debug(u"Child process Started to handle file "
"operation: %s" % file_operation)
except Exception:
self._stop_child_logger()
raise
def on_operation_abort(self, file_operation):
self.logger.debug(u'Abort detected for the operation I am handling: '
'%s. Terminating child process...' % file_operation)
self.abort_operation()
def abort_operation(self):
if self.child is not None and self.child.is_alive():
try:
if self.terminationEvent is not None:
self.terminationEvent.set()
# else:
# self.child.terminate()
except:
pass
def _terminate_child(self):
self.stop_network_transfer()
if self.child is not None:
self.input_queue.put(('PoisonPill', None))
self.child.join(5)
self.child = None
def _clean_env(self):
self._terminate_child()
self._stop_child_logger()
self._destroy_multiprocessing_queues()
def _on_poison_pill(self):
self.logger.debug(u"Got poison pill.")
self.must_die.set()
self._clean_env()
def terminate(self):
'''
Signal the worker that the time has come.
'''
self.abort_operation()
def stop_network_transfer(self):
self.abort_operation()
def _termination_requested(self):
return self.must_die.wait(0.01)
class Worker(Thread):
'''A worker takes something to do and does it, and then again takes
something to do etc.
'''
def __init__(self,
warebox,
operation_queue,
server_session,
cfg,
cryptoAdapter,
worker_pool):
"""
@param warebox:
Instance of filerockclient.warebox.Warebox.
@param cfg:
Instance of filerockclient.config.ConfigManager.
@param operation_queue:
A threading queue, where worker receives the operations
@param worker_pool:
Instance of filerockclient.workers.worker.Worker
"""
Thread.__init__(self, name=self.__class__.__name__)
self.cfg = cfg
self.operation_queue = operation_queue
self._server_session = server_session
self.warebox = warebox
self.child = None
self.input_queue = None
self.communicationQueue = None
self.child_logger = None
self.child_logs_queue = None
self.cryptoAdapter = cryptoAdapter
self.integrity_manager = IntegrityManager(None)
self._worker_pool = worker_pool
self.must_die = threading.Event()
self.last_send = datetime.now()
self.communicationQueue = None
self.child_logs_queue = None
def run(self):
"""
Serves file operations until termination request is received
"""
try:
self.name += "_%s" % self.ident
self.logger = logging.getLogger("FR.%s" % self.getName())
self.logger.debug(u'Started.')
while not self._termination_requested():
self._serve_file_operations()
self.logger.debug(u"I'm terminated.")
finally:
self._terminate_child()
def _serve_file_operations(self):
"""
Blocks on operation queue until a message is received
If the message is a POISON_PILL the worker terminate it self
If the message is a non aborted file operation,
an abort handler is registered to it and the operation is handled
"""
file_operation = self.operation_queue.get()
if file_operation == 'POISON_PILL':
self._on_poison_pill()
return
assert type(file_operation) in [PathnameOperation,
CreateDirectoriesTask,
DeleteLocalTask,
ResolveDeletionConflictsTask]
self.logger.debug(u"worker executing: %s", file_operation)
if __debug__:
self._worker_pool.track_assign_worker_to_pathname(
self.ident,
file_operation.pathname)
try:
self.warebox._check_blacklisted_dir()
if file_operation.is_aborted():
self.logger.debug(u"Got an already aborted operation, "
"giving up: %s" % file_operation)
else:
self.logger.debug(u"Got an operation to handle: %s",
file_operation)
file_operation.register_abort_handler(self.on_operation_abort)
self._handle_file_operation(file_operation)
except Exception:
self.logger.error(u"Some problem occurred in worker "
u"handling operation %r" % file_operation)
raise
finally:
self.logger.debug("Releasing a worker")
if __debug__:
self._worker_pool.track_assert_assigned(
self.ident, file_operation.pathname)
self._worker_pool.release_worker()
if __debug__:
self._worker_pool.track_release_worker(
self.ident,
file_operation.pathname)
def _handle_file_operation(self, file_operation):
if file_operation.verb == 'UPLOAD':
self._handle_upload_file_operation(file_operation)
elif file_operation.verb == 'DOWNLOAD':
self._handle_download_file_operation(file_operation)
elif file_operation.verb == 'DELETE_LOCAL':
self._handle_delete_local_file_operation(file_operation)
elif file_operation.verb == 'CREATE_DIRECTORIES':
self._handle_operation_create_directories(file_operation)
elif file_operation.verb == 'RESOLVE_DELETION_CONFLICTS':
self._handle_operation_resolve_deletion_conflicts(file_operation)
else:
self.logger.warning(u"I should not handle a '%s' operation! "
"I'm rejecting it", file_operation.verb)
file_operation.reject()
def _send_percentage(self, file_operation, status, percentage):
now = datetime.now()
delta = now - self.last_send
if ((delta.seconds + delta.microseconds/1000000.) > 0.5) \
or (percentage == 100):
file_operation.notify_pathname_status_change(
status, {'percentage': percentage})
self.last_send = now
def _handle_network_transfer_operation(self, file_operation):
'''By locking we mantain the following invariant: if the
EventQueue tries to abort this operation due to a conflicting
operation, then EventQueue waits until this operation either
aborts or completes.
This preserves the ordering of execution for the conflicting
operations - that is, the EventQueue doesn't emit
the conflicting operation while this one is still working.
'''
with file_operation.lock:
if not file_operation.is_aborted():
self.logger.debug(u"Starting child process to handle file"
" operation: %s" % file_operation)
try:
self._spawn_child(file_operation)
self.input_queue.put(('FileOperation', file_operation))
except Exception as e:
self.logger.error(
u"Could not spawn a child process: %r" % e)
raise OperationRejection(file_operation)
else:
self.logger.debug(u"Got an already aborted operation, "
"giving up: %s" % file_operation)
return False
if file_operation.verb == 'UPLOAD':
status = PStatuses.UPLOADING
else:
status = PStatuses.DOWNLOADING
self._send_percentage(file_operation, status, 0)
termination = False
max_retry = 3
while not termination:
message, content = self.communicationQueue.get()
self.logger.debug(u'Worker send back %s with content %s'
% (message, content))
if message == 'completed':
termination = True
if file_operation.verb == 'DOWNLOAD':
return {'actual_etag': content['actual_etag']}
else:
return True
elif message == 'interrupted':
self.logger.debug(u"Child has been terminated by "
"Software Operation: %s"
% file_operation)
file_operation.abort()
termination = True
return False
elif message == 'failed':
self.logger.error(u"Child has been terminated, "
"Assuming failure for operation: %s"
% file_operation)
max_retry -= 1
if max_retry == 0:
raise OperationRejection(file_operation)
self.input_queue.put(('FileOperation', file_operation))
elif message == 'percentage':
self._send_percentage(file_operation, status, content)
elif message == 'log':
level, msg = content
self.child_logger[level](msg)
elif message == 'ShuttingDown':
self.logger.debug(u"Get a shutting down message from process")
termination = True
return False
elif message == 'DIED':
self.child = None
termination = True
raise OperationRejection(file_operation)
def _handle_upload_file_operation(self, operation):
try:
success = self._handle_network_transfer_operation(operation)
if success:
CryptoUtils.clean_env(operation, self.logger)
self.logger.debug(u"Operation has been completed "
"successfully: %s" % operation)
self.logger.info(u'Synchronized pathname: %s "%s", which '
'will be persisted after a commit'
% (operation.verb, operation.pathname))
operation.notify_pathname_status_change(PStatuses.UPLOADED,
{'percentage': 100})
operation.complete()
except Exception as e:
self.logger.error(u"Error while uploading: %r."
" Rejecting the operation: %s" % (e, operation))
operation.reject()
def _handle_operation_create_directories(self, task):
operations = sorted(task.operations, key=lambda op: op.pathname)
actual_etag = 'd41d8cd98f00b204e9800998ecf8427e'
for operation in operations:
# Check integrity
operation.notify_pathname_status_change(PStatuses.DOWNLOADING)
res = self._check_download_integrity(operation, actual_etag)
if not res['valid']:
# Detected an integrity error. Badly bad.
self._server_session.signal_download_integrity_error(
operation, res['reason'],
res['expected_etag'], res['expected_basis'],
res['actual_etag'], res['computed_basis'])
return
# The directory is valid, create it
self._make_directories_to(operation.pathname)
lmtime = self.warebox.get_last_modification_time(operation.pathname)
operation.lmtime = lmtime
operation.notify_pathname_status_change(PStatuses.ALIGNED)
operation.complete()
self.logger.info(u'Synchronized pathname: %s "%s"'
% (operation.verb, operation.pathname))
self.logger.debug(u"Operation has been completed "
"successfully: %s" % operation)
task.complete()
def _handle_download_file_operation(self, operation):
try:
# Note: it is a normal file, not a directory
CryptoUtils.set_temp_file(operation, self.cfg)
success = self._handle_network_transfer_operation(operation)
if success:
actual_etag = success['actual_etag']
res = self._check_download_integrity(operation, actual_etag)
if not res['valid']:
# Detected an integrity error. Badly bad.
self._server_session.signal_download_integrity_error(
operation, res['reason'],
res['expected_etag'], res['expected_basis'],
res['actual_etag'], res['computed_basis'])
return
self._make_directories_to(operation.pathname)
if operation.to_decrypt:
# We have not finished yet, leaving the rest to decrypter.
# Note: the decrypter duplicates the following ending logic
self.cryptoAdapter.put(operation)
return
# It is a valid cleartext file, move it to the warebox
self.warebox.move(operation.temp_pathname,
operation.pathname,
operation.conflicted)
self.logger.debug(u"Operation has been completed "
"successfully: %s" % operation)
self.logger.info(u'Synchronized pathname: %s "%s"'
% (operation.verb, operation.pathname))
lmtime = self.warebox.get_last_modification_time(operation.pathname)
operation.lmtime = lmtime
operation.notify_pathname_status_change(PStatuses.ALIGNED)
operation.complete()
except Exception as e:
self.logger.error(u"Error while downloading: %r."
" Rejecting the operation: %s" % (e, operation))
self.logger.error(u"Stacktrace: %r" % traceback.format_exc())
operation.reject()
finally:
# Just in case the move had failed for any reason
if not operation.to_decrypt:
if operation.temp_pathname is not None \
and os.path.exists(operation.temp_pathname):
_try_remove(operation.temp_pathname, self.logger)
def _make_directories_to(self, pathname):
"""Make sure that the full path for this file exists. It may be
either a file or a directory.
There are a couple of reasons why the full path may not exist:
1) Data has been loaded on the storage with an external tool
that doesn't explicitly support directories, so they have
not been created. Treating these directories as something
to download is wrong: they wouldn't pass the integrity
check (they really doesn't exist in the trusted dataset),
so it is better to make them as new local modifications.
2) The user has deleted while offline a directory that is
needed by a download. We call it "hierarchy conflict" and
is uncovered by the diff algorithm, since case 1 would
still remain unhandled.
"""
self.warebox.make_directories_to(pathname)
def _check_download_integrity(self, operation, actual_etag):
pathname = operation.pathname
proof = operation.download_info['proof']
basis = operation.download_info['trusted_basis']
self.integrity_manager.trusted_basis = basis
result = {}
result['valid'] = None
result['reason'] = None
result['expected_etag'] = operation.storage_etag
result['expected_basis'] = basis
result['actual_etag'] = actual_etag
result['computed_basis'] = None
if operation.storage_etag != actual_etag:
# Note: the etag inside the proof will be
# checked by the integrity manager.
self.logger.debug(
u"Invalid etag of download operation. "
"Expected etag %s but found %s. %s"
% (operation.storage_etag, actual_etag, operation))
result['valid'] = False
result['reason'] = "Expected etag different from actual etag"
return result
try:
self.integrity_manager.addOperation('DOWNLOAD',
pathname,
proof,
actual_etag)
result['valid'] = True
result['computed_basis'] = basis
except MalformedProofException as e:
self.logger.debug(u"Invalid proof of download operation: "
"%s. %r, %r" % (e, operation, proof))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
except WrongBasisFromProofException as e:
self.logger.debug(u"Integrity check of download operation failed."
"Basis %s was expected but the proof computed "
"%s. Error details: %s. %r, %r" %
(basis, e.operation_basis, e, operation, proof))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
result['computed_basis'] = e.operation_basis
except Exception as e:
self.logger.debug(u"Integrity check of download operation failed"
" with unknown reason: %s. %r, %r"
% (e, operation, proof))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
finally:
self.integrity_manager.clear()
return result
def _handle_delete_local_file_operation(self, task):
pathnames = sorted(task.pathname2proof.keys())
#self.logger.debug("Going to delete pathnames: %s" % pathnames)
for pathname in pathnames:
basis = task.trusted_basis
proof = task.pathname2proof[pathname]
res = self._check_deletelocal_integrity(pathname, proof, basis)
if not res['valid']:
# Detected an integrity error. Badly bad.
self._server_session.signal_deletelocal_integrity_error(
pathname, proof, res['reason'],
res['expected_basis'], res['computed_basis'])
return
roots = {}
for pathname in pathnames:
found_ancestor = False
for root in roots:
if pathname.startswith(root):
found_ancestor = True
break
if not found_ancestor:
roots[pathname] = True
try:
for pathname in roots.iterkeys():
self.warebox.delete_tree(pathname)
except Exception as e:
self.logger.error(
u"Caught an operating system exception while "
u"modifying the filesystem. Are you locking the Warebox? % r"
% e)
raise
task.complete()
def _check_deletelocal_integrity(self, pathname, proof, trusted_basis):
self.integrity_manager.trusted_basis = trusted_basis
result = {}
result['valid'] = None
result['reason'] = None
result['expected_basis'] = trusted_basis
result['computed_basis'] = None
try:
self.integrity_manager.addOperation('DELETE_LOCAL',
pathname,
proof,
None)
result['valid'] = True
result['computed_basis'] = trusted_basis
except MalformedProofException as e:
self.logger.debug(u"Invalid proof of delete_local operation: "
"%s. %s, %s" % (e, pathname, proof.raw))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
except WrongBasisFromProofException as e:
self.logger.debug(u"Integrity check of delete_local operation failed."
"Basis %s was expected but the proof computed "
"%s. Error details: %s. %r, %r" %
(trusted_basis, e.operation_basis, e, pathname, proof.raw))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
result['computed_basis'] = e.operation_basis
except Exception as e:
self.logger.debug(u"Integrity check of delete_local operation"
" failed with unknown reason: %s. %r, %r"
% (e, pathname, proof.raw))
self.logger.debug(traceback.format_exc())
result['valid'] = False
result['reason'] = "%s" % e
finally:
self.integrity_manager.clear()
return result
def _find_new_name(self, pathname):
# TODO: try harder in finding a name that is available
curr_time = datetime.now().strftime('%Y-%m-%d %H_%M_%S')
suffix = ' (Conflicted on %s)' % curr_time
if pathname.endswith('/'):
new_pathname = pathname[:-1] + suffix + '/'
else:
basename, ext = os.path.splitext(pathname)
new_pathname = basename + suffix + ext
return new_pathname
def _rename_conflicting_pathname(self, pathname, prefix=None):
new_pathname = self.warebox.rename(pathname, pathname, prefix)
return new_pathname
def _handle_operation_resolve_deletion_conflicts(self, task):
"""Solve deletion conflicts by renaming the local file to a new
pathname. The old pathname will result implicitly deleted.
Deletion conflicts are tough to resolve. A conflicting pathname:
a) has been deleted by the server
b) has an ancestor folder that has been deleted by the server
c) both
It must be checked if it's safe leaving the file in its original
folder (that is, if it still exists).
"""
conflicts = task.deletion_conflicts
content_to_delete_locally = task.content_to_delete_locally
for pathname in conflicts:
basis = task.trusted_basis
proof = task.pathname2proof[pathname]
res = self._check_deletelocal_integrity(pathname, proof, basis)
if not res['valid']:
# Detected an integrity error. Badly bad.
self._server_session.signal_deletelocal_integrity_error(
pathname, proof, res['reason'],
res['expected_basis'], res['computed_basis'])
return
try:
backupped_folders = {}
for pathname in conflicts:
missing_ancestor_folders = filter(lambda p: pathname.startswith(p), content_to_delete_locally)
# Is it safe leaving the file in its original folder?
if len(missing_ancestor_folders) > 0:
# No, it's been deleted. Backup the whole deleted subtree
missing_ancestor_folders = sorted(missing_ancestor_folders)
highest_missing_folder = missing_ancestor_folders[0]
if not highest_missing_folder in backupped_folders:
backup_folder = self._find_new_name(highest_missing_folder)
self.warebox.make_directory(backup_folder)
backupped_folders[highest_missing_folder] = backup_folder
backup_folder = backupped_folders[highest_missing_folder]
new_pathname = pathname.replace(highest_missing_folder, backup_folder, 1)
self.warebox.make_directories_to(new_pathname)
if not self.warebox.is_directory(new_pathname):
self.warebox.rename(pathname, new_pathname)
else:
# Yes, just rename the file
new_pathname = self._rename_conflicting_pathname(pathname, 'Deleted')
self.logger.warning(
u"Conflict detected for pathname %r, which has been "
u"remotely deleted. Moved the local copy to: %r"
% (pathname, new_pathname))
task.complete()
except Exception:
self.logger.error(
u"Caught an operating system exception while modifying the "
u"filesystem. Are you locking the Warebox?")
raise
def _start_child_logger(self):
self._stop_child_logger()
self.child_logs_queue = multiprocessing.Queue()
self.child_logger = LogsReceiver(self.getName(), self.child_logs_queue)
self.child_logger.start()
if self.child_logger is None:
logger = logging.getLogger(u'FR.WorkerChild of %s' % self.getName())
self.child_logger = {
'info': logger.info,
'debug': logger.debug,
'warning': logger.warning,
'error': logger.error,
'critical': logger.critical
}
def _stop_child_logger(self):
if self.child_logger is not None:
self.child_logger.stop()
self.child_logs_queue.put(('log', ('debug', 'Die please!')))
self.child_logger.join()
self.child_logger = None
if self.child_logs_queue is not None:
self.child_logs_queue.close()
self.child_logs_queue.join_thread()
self.child_logs_queue = None
def _create_multiprocessing_queues(self):
self._destroy_multiprocessing_queues()
self.input_queue = Queue.Queue()
self.communicationQueue = Queue.Queue()
def _destroy_multiprocessing_queue(self, queue):
if queue is not None:
while not queue.empty():
queue.get_nowait()
# queue.close()
# queue.join_thread()
queue = None
def _destroy_multiprocessing_queues(self):
self._destroy_multiprocessing_queue(self.input_queue)
self._destroy_multiprocessing_queue(self.communicationQueue)
def _spawn_child(self, file_operation):
if self.child is None or not self.child.is_alive():
# self.terminationEvent = multiprocessing.Event()
self.terminationEvent = threading.Event()
self._create_multiprocessing_queues()
self._start_child_logger()
try:
self.logger.debug(u"Allocating child process to handle file "
"operation: %s" % file_operation)
self.child = WorkerChild(self.warebox,
self.input_queue,
self.communicationQueue,
self.terminationEvent,
self._send_percentage,
#self.child_logs_queue,
self.cfg,
self._worker_pool)
self.child.start()
self.logger.debug(u"Child process Started to handle file "
"operation: %s" % file_operation)
except Exception:
self._stop_child_logger()
raise
def on_operation_abort(self, file_operation):
self.logger.debug(u'Abort detected for the operation I am handling: '
'%s. Terminating child process...' % file_operation)
self.abort_operation()
def abort_operation(self):
if self.child is not None and self.child.is_alive():
try:
if self.terminationEvent is not None:
self.terminationEvent.set()
# else:
# self.child.terminate()
except:
pass
def _terminate_child(self):
self.stop_network_transfer()
if self.child is not None:
self.input_queue.put(('PoisonPill', None))
self.child.join(5)
self.child = None
def _clean_env(self):
self._terminate_child()
self._stop_child_logger()
self._destroy_multiprocessing_queues()
def _on_poison_pill(self):
self.logger.debug(u"Got poison pill.")
self.must_die.set()
self._clean_env()
def terminate(self):
'''
Signal the worker that the time has come.
'''
self.abort_operation()
def stop_network_transfer(self):
self.abort_operation()
def _termination_requested(self):
return self.must_die.wait(0.01)
class ServerSession(threading.Thread):
"""
The thread that controls the communication with the server and the
internal execution of the client.
ServerSession is the intelligent part of the client application. It
is implemented as an event-loop machine, receiving events from many
sources: the user, ServerSession itself, other application
components and the server, each with a different priority level. It
decides, for example, when to handle operations for synchronizing
data and when to answer to messages from the server, thus
implementing the communication protocol.
The logic is implemented as an object-oriented state machine, which
decides which events to handle in each state and how to do it.
ServerSession is both a container (formally a "context") for "state"
objects and the only public interface of this component. The context
gives the states a shared memory space and access to other
components.
All the logic is actually implemented in the state objects, limiting
ServerSession to execute the "current state". State objects also
decide when to switch to another state, as a reaction to an input
event. The states are implemented using object-oriented inheritance
in order to common factor the logic of several states, making
ServerSession precisely a "Hierarchical State Machine".
See the "State" design pattern from the book: Gamma et al, "Design
Patterns: Elements of Reusable Object-Oriented Software" for a
reference to the design.
"""
def __init__(self,
cfg, warebox, storage_cache,
startup_synchronization, filesystem_watcher, linker,
metadata_db, hashes_db, internal_facade, ui_controller,
lockfile_fd, auto_start, input_queue, scheduler):
"""
@param cfg:
Instance of filerockclient.config.ConfigManager.
@param warebox:
Instance of filerockclient.warebox.Warebox.
@param storage_cache:
Instance of filerockclient.databases.storage_cache.
StorageCache.
@param startup_synchronization:
Instance of filerockclient.serversession.
startup_synchronization.
@param filesystem_watcher:
Instance of any class in the filerockclient.
filesystem_watcher package.
@param linker:
Instance of filerockclient.linker.Linker.
@param metadata_db:
Instance of filerockclient.databases.metadata.
MetadataDB.
@param hashes_db:
Instance of filerockclient.databases.hashes.HashesDB.
@param internal_facade:
Instance of filerockclient.internal_facade.
InternalFacade.
@param ui_controller:
Instance of filerockclient.ui.ui_controller.
UIController.
@param lockfile_fd:
File descriptor of the lock file which ensures there
is only one instance of FileRock Client running.
Child processes have to close it to avoid stale locks.
@param auto_start:
Boolean flag telling whether ServerSession should
connect to the server when started.
@param input_queue:
Instance of filerockclient.util.multi_queue.
MultiQueue. It is expected to have the following
queues:
usercommand: Commands sent by the user
sessioncommand: ServerSession internal use commands
systemcommand: Commands sent by other client components
servermessage: Messages sent by the server.
operation: PathnameOperation objects to handle
@param scheduler:
Instance of filerockclient.util.scheduler.Scheduler.
"""
threading.Thread.__init__(self, name=self.__class__.__name__)
self.logger = logging.getLogger("FR.%s" % self.__class__.__name__)
self._input_queue = input_queue
self.warebox = warebox
self.startup_synchronization = startup_synchronization
self.filesystem_watcher = filesystem_watcher
self._internal_facade = internal_facade
self._ui_controller = ui_controller
self.metadataDB = metadata_db
self.hashesDB = hashes_db
self.auto_start = auto_start
self._scheduler = scheduler
self.storage_cache = storage_cache
self.linker = linker
self.warebox = warebox
self.cfg = cfg
self._lockfile_fd = lockfile_fd
self._started = False
self.must_die = threading.Event()
# TODO: this flag exists due to auto-disconnection. It will be removed
# and replaced by a CONNECTFORCE command as soon as ServerSession will
# stop going automatically to DisconnectedState.
self.disconnect_other_client = False
self.operation_responses = {}
self._pathname2id = {}
self.output_message_queue = Queue.Queue()
self.input_keepalive_queue = Queue.Queue()
self.current_state = None
self.reconnection_time = 1
self.num_connection_attempts = 0
self.max_connection_attempts = MAX_CONNECTION_ATTEMPTS
self._basis_lock = threading.Lock()
self.server_basis = None
self.session_id = None
self.storage_ip_address = None
self.refused_declare_count = 0
self._current_basis = None
self.id = 0
self._sync_operations = []
self.keepalive_timer = ConnectionLifeKeeper(
self._input_queue, self.input_keepalive_queue,
self.output_message_queue, True)
self.transaction = Transaction()
self.transaction_manager = TransactionManager(
self.transaction, self.storage_cache)
StateRegister.setup(self)
self.client_id = None
self.username = None
self.priv_key = None
self.host = None
self.port = None
self.server_certificate = None
self.storage_hostname = None
self.refused_declare_max = None
self.refused_declare_waiting_time = None
self.commit_threshold_seconds = None
self.commit_threshold_operations = None
self.commit_threshold_bytes = None
self.transaction_cache = None
self.integrity_manager = None
self.cryptoAdapter = None
self.temp_dir = None
self.connection_reader = None
self.connection_writer = None
self.sock = None
self.listening_operations = False
self.reload_config_info()
def reload_config_info(self):
"""
Refresh the configuration values.
Reload the configuration, get configuration values from self.cfg
and set them as attributes of self.
To be called at least once.
"""
# TODO: merge this method into the constructor, there is no reason to
# keep it separated anymore.
self.cfg.load()
self.client_id = self.cfg.get('User', 'client_id')
self.username = self.cfg.get('User', 'username')
self.priv_key = self.cfg.get('Application Paths', 'client_priv_key_file')
self.host = self.cfg.get('System', 'server_hostname')
self.port = self.cfg.getint('System', 'server_port')
self.server_certificate = self.cfg.get('Application Paths', 'server_certificate')
self.storage_hostname = self.cfg.get('System', 'storage_endpoint')
self.refused_declare_max = self.cfg.getint(
'System', 'refused_declare_max')
self.refused_declare_waiting_time = self.cfg.getint(
'System', 'refused_declare_waiting_time')
self.commit_threshold_seconds = self.cfg.getint(
'Client', 'commit_threshold_seconds')
self.commit_threshold_operations = self.cfg.getint(
'Client', 'commit_threshold_operations')
self.commit_threshold_bytes = self.cfg.getint(
'Client', 'commit_threshold_bytes')
temp = self.cfg.get('Application Paths', 'transaction_cache_db')
self.transaction_cache = TransactionCache(temp)
self.integrity_manager = IntegrityManager(None)
is_firt_startup = self._internal_facade.is_first_startup()
self.cryptoAdapter = Adapter(self.cfg,
self.warebox,
self._input_queue,
self._lockfile_fd,
enc_dir='enc',
first_startup=is_firt_startup)
self.worker_pool = WorkerPool(self.warebox,
self,
self.cfg,
self.cryptoAdapter)
self.temp_dir = self.cryptoAdapter.get_enc_dir()
self._ui_controller.update_config_info(self.cfg)
def run(self):
"""Implementation of the threading.Thread.run() method."""
self._started = True
try:
self.worker_pool.start_workers()
self.keepalive_timer.start()
self.current_state = StateRegister.get('DisconnectedState')
curr_basis = self.current_state._load_trusted_basis()
self.integrity_manager.setCurrentBasis(curr_basis)
self.logger.info(u'Current basis is: %s' % curr_basis)
self.current_state._on_entering()
self._internal_facade.set_global_status(GStatuses.NC_STOPPED)
if self.auto_start:
self._input_queue.put(Command('CONNECT'), 'sessioncommand')
self.cryptoAdapter.start()
self._scheduler.schedule_action(
self.check_encrypted_folder, name='check_encrypted_folder',
seconds=5, repeating=True)
# The event loop
self._main_loop()
except UnexpectedMessageException as e:
self.logger.critical(
u"Received an unexpected message from the Server while in "
u"state '%s': %s. Forcing termination."
% (self.current_state.__class__, str(e)))
raise
except ProtocolException as e:
self.logger.critical(
u"Detected an unrecoverable error, forcing termination: %s"
% str(e))
# Pre-emptive release, just stop before messing up the server
self.release_network_resources()
raise
except Exception as e:
self.logger.critical(
u"Forcing termination due to uncaught exception '%s': %s"
% (e.__class__, e))
self.logger.debug(
u"Last error stacktrace:\n%s" % traceback.format_exc())
# Pre-emptive release, just stop before messing up the server
self.release_network_resources()
raise
def _main_loop(self):
"""
The event loop.
A loop that contiuosly calls the current state of the state
machine, executing its logic.
It exits when self.terminate() is called.
"""
while not self.must_die.is_set():
next_state = self.current_state.do_execute()
if next_state != self.current_state:
self.current_state._on_leaving()
next_state._on_entering()
self.current_state = next_state
def check_encrypted_folder(self):
"""
Check if encryption preconditions are satisfied, try to
satisfy them otherwise.
This method is meant to be asynchronously called by a timer,
precisely to be registered into self._scheduler.
"""
if not self.cryptoAdapter.check_precondition(self._ui_controller):
self._internal_facade.terminate()
def acquire_network_resources(self):
"""
Configure a network connection to the server.
The resulting socket is handled by two instances of
ServerConnectionReader and ServerConnectionWriter, which are
created and run as well. It is possibile to send/receive
message to/from them through the queues self.output_message_queue
and self._input_queue (servermessage).
Note: this private method should be actually "protected" for
the ServerSession states, but Python doesn't have such a
protection level.
"""
try:
self.logger.debug(u"Creating a socket on %s:%s", self.host, self.port)
sock = socket.create_connection((self.host, self.port), timeout=10)
ca_chain = os.path.abspath(self.server_certificate)
self.sock = ssl.wrap_socket(
sock, cert_reqs=ssl.CERT_REQUIRED, ca_certs=ca_chain,
ssl_version=ssl.PROTOCOL_TLSv1)
self.sock.setblocking(True)
match_hostname(self.sock.getpeercert(), self.host)
except CertificateError as e:
self.logger.critical(u"SSL certificate validation failed: %s" % e)
raise ssl.SSLError(e)
except socket.error as exception:
self.logger.debug(u"Error opening SSL socket: %s" % exception)
self.logger.warning(
u"Unable to connect, re-trying in %s seconds."
% self.reconnection_time)
self.reconnection_time = stoppable_exponential_backoff_waiting(
self.reconnection_time, self.must_die, 10)
self.num_connection_attempts += 1
return False
except socket.timeout as exception:
self.logger.debug(u"Socket timeout: %s" % exception)
self.logger.warning(u"Unable to connect, re-trying in %s seconds."
% self.reconnection_time)
self.reconnection_time = stoppable_exponential_backoff_waiting(
self.reconnection_time, self.must_die, 10)
self.num_connection_attempts += 1
return False
self.connection_reader = ServerConnectionReader(
self._input_queue, self.input_keepalive_queue, self.sock)
self.connection_writer = ServerConnectionWriter(
self._input_queue, self.output_message_queue, self.sock)
self.connection_reader.start()
self.connection_writer.start()
self.reconnection_time = 1
return True
def release_network_resources(self):
"""
Close the active connection to the server, if any.
Note: this private method should be actually "protected" for
the ServerSession states, but Python doesn't have such a
protection level.
"""
try:
self.connection_reader.terminate()
except AttributeError:
pass
try:
self.connection_writer.terminate()
except AttributeError:
pass
try:
self.sock.shutdown(socket.SHUT_RDWR)
self.sock.close()
except socket.error:
# Shutdown yelds an error on already closed sockets
pass
except AttributeError:
pass
def commit(self):
"""Make the client commit the current transaction."""
self._input_queue.put(Command('USERCOMMIT'), 'usercommand')
def connect(self):
"""Make the client connect to the server."""
self._input_queue.put(Command('CONNECT'), 'usercommand')
def disconnect(self):
"""Make the client disconnect from the client, if connected."""
# TODO: this method has been temporarly replaced by PAUSE
self._input_queue.put(Command('DISCONNECT'), 'usercommand')
def signal_free_worker(self):
"""
Tell ServerSession that a worker is free to receive new tasks.
This method is meant to be called by WorkerPool.
"""
self._input_queue.put(Command('WORKERFREE'), 'systemcommand')
def signal_download_integrity_error(
self, operation, reason,
expected_etag, expected_basis,
actual_etag, computed_basis):
"""Tell ServerSession that the integrity check of a downloaded
file has failed.
This method is meant to be called by Workers.
@param operation:
Instance of PathnameOperation. Remember that it
contains also its Proof object.
@param reason:
String shortly describing the error.
@param expected_etag:
The etag the file was expected to have. It's the one
communicated by the server in its file list.
@param expected_basis:
The trusted basis. It's the one the user had
accepted when the sync started.
@param actual_etag:
The etag the file has turned to have after being
downloaded.
@param computed_basis:
The basis returned by the IntegrityManager when
computing the given proof object. Possibly None.
"""
cmd = Command('INTEGRITYERRORONDOWNLOAD')
cmd.operation = operation
cmd.proof = operation.download_info['proof']
cmd.reason = reason
cmd.expected_etag = expected_etag
cmd.expected_basis = expected_basis
cmd.actual_etag = actual_etag
cmd.computed_basis = computed_basis
self._input_queue.put(cmd, 'systemcommand')
def signal_deletelocal_integrity_error(
self, pathname, proof, reason, expected_basis, computed_basis):
"""Tell ServerSession that the integrity check of a
pathname to delete locally has failed.
This method is meant to be called by Workers.
@param pathname:
String representing the pathname.
@param reason:
String shortly describing the error.
@param expected_basis:
The trusted basis. It's the one the user had
accepted when the sync started.
@param computed_basis:
The basis returned by the IntegrityManager when
computing the given proof object. Possibly None.
"""
cmd = Command('INTEGRITYERRORONDELETELOCAL')
cmd.pathname = pathname
cmd.proof = proof
cmd.reason = reason
cmd.expected_basis = expected_basis
cmd.computed_basis = computed_basis
self._input_queue.put(cmd, 'systemcommand')
def get_current_basis(self):
"""
Return the current trusted basis.
@return The current trusted basis.
"""
with self._basis_lock:
return self._current_basis
def print_transaction(self):
"""
Print the list of operations in the current transaction.
Debug method, it prints to stdout and thus works only when the
application is attached to a console.
"""
self.transaction.print_all()
def terminate(self):
"""
Termination routine for this component.
Stops the running thread and releases any acquired resource.
"""
self.logger.debug(u"Terminating Server Session...")
if self._started:
self.must_die.set()
self.worker_pool.terminate()
self._input_queue.put(Command('TERMINATE'), 'usercommand')
self.transaction.can_be_committed.set()
self.join() if self is not threading.current_thread() else None
self.keepalive_timer.terminate()
self.release_network_resources()
self.cryptoAdapter.terminate()
self.logger.debug(u"Server Session terminanted.")
class ServerSession(threading.Thread):
"""
The thread that controls the communication with the server and the
internal execution of the client.
ServerSession is the intelligent part of the client application. It
is implemented as an event-loop machine, receiving events from many
sources: the user, ServerSession itself, other application
components and the server, each with a different priority level. It
decides, for example, when to handle operations for synchronizing
data and when to answer to messages from the server, thus
implementing the communication protocol.
The logic is implemented as an object-oriented state machine, which
decides which events to handle in each state and how to do it.
ServerSession is both a container (formally a "context") for "state"
objects and the only public interface of this component. The context
gives the states a shared memory space and access to other
components.
All the logic is actually implemented in the state objects, limiting
ServerSession to execute the "current state". State objects also
decide when to switch to another state, as a reaction to an input
event. The states are implemented using object-oriented inheritance
in order to common factor the logic of several states, making
ServerSession precisely a "Hierarchical State Machine".
See the "State" design pattern from the book: Gamma et al, "Design
Patterns: Elements of Reusable Object-Oriented Software" for a
reference to the design.
"""
def __init__(self, cfg, warebox, storage_cache, startup_synchronization,
filesystem_watcher, linker, metadata_db, hashes_db,
internal_facade, ui_controller, lockfile_fd, auto_start,
input_queue, scheduler):
"""
@param cfg:
Instance of filerockclient.config.ConfigManager.
@param warebox:
Instance of filerockclient.warebox.Warebox.
@param storage_cache:
Instance of filerockclient.databases.storage_cache.
StorageCache.
@param startup_synchronization:
Instance of filerockclient.serversession.
startup_synchronization.
@param filesystem_watcher:
Instance of any class in the filerockclient.
filesystem_watcher package.
@param linker:
Instance of filerockclient.linker.Linker.
@param metadata_db:
Instance of filerockclient.databases.metadata.
MetadataDB.
@param hashes_db:
Instance of filerockclient.databases.hashes.HashesDB.
@param internal_facade:
Instance of filerockclient.internal_facade.
InternalFacade.
@param ui_controller:
Instance of filerockclient.ui.ui_controller.
UIController.
@param lockfile_fd:
File descriptor of the lock file which ensures there
is only one instance of FileRock Client running.
Child processes have to close it to avoid stale locks.
@param auto_start:
Boolean flag telling whether ServerSession should
connect to the server when started.
@param input_queue:
Instance of filerockclient.util.multi_queue.
MultiQueue. It is expected to have the following
queues:
usercommand: Commands sent by the user
sessioncommand: ServerSession internal use commands
systemcommand: Commands sent by other client components
servermessage: Messages sent by the server.
operation: PathnameOperation objects to handle
@param scheduler:
Instance of filerockclient.util.scheduler.Scheduler.
"""
threading.Thread.__init__(self, name=self.__class__.__name__)
self.logger = logging.getLogger("FR.%s" % self.__class__.__name__)
self._input_queue = input_queue
self.warebox = warebox
self.startup_synchronization = startup_synchronization
self.filesystem_watcher = filesystem_watcher
self._internal_facade = internal_facade
self._ui_controller = ui_controller
self.metadataDB = metadata_db
self.hashesDB = hashes_db
self.auto_start = auto_start
self._scheduler = scheduler
self.storage_cache = storage_cache
self.linker = linker
self.warebox = warebox
self.cfg = cfg
self._lockfile_fd = lockfile_fd
self._started = False
self.must_die = threading.Event()
# TODO: this flag exists due to auto-disconnection. It will be removed
# and replaced by a CONNECTFORCE command as soon as ServerSession will
# stop going automatically to DisconnectedState.
self.disconnect_other_client = False
self.operation_responses = {}
self._pathname2id = {}
self.output_message_queue = Queue.Queue()
self.input_keepalive_queue = Queue.Queue()
self.current_state = None
self.reconnection_time = 1
self.num_connection_attempts = 0
self.max_connection_attempts = MAX_CONNECTION_ATTEMPTS
self._basis_lock = threading.Lock()
self.server_basis = None
self.session_id = None
self.storage_ip_address = None
self.refused_declare_count = 0
self._current_basis = None
self.id = 0
self._sync_operations = []
self.keepalive_timer = ConnectionLifeKeeper(self._input_queue,
self.input_keepalive_queue,
self.output_message_queue,
True)
self.transaction = Transaction()
self.transaction_manager = TransactionManager(self.transaction,
self.storage_cache)
StateRegister.setup(self)
self.client_id = None
self.username = None
self.priv_key = None
self.host = None
self.port = None
self.server_certificate = None
self.storage_hostname = None
self.refused_declare_max = None
self.refused_declare_waiting_time = None
self.commit_threshold_seconds = None
self.commit_threshold_operations = None
self.commit_threshold_bytes = None
self.transaction_cache = None
self.integrity_manager = None
self.cryptoAdapter = None
self.temp_dir = None
self.connection_reader = None
self.connection_writer = None
self.sock = None
self.listening_operations = False
self.reload_config_info()
def reload_config_info(self):
"""
Refresh the configuration values.
Reload the configuration, get configuration values from self.cfg
and set them as attributes of self.
To be called at least once.
"""
# TODO: merge this method into the constructor, there is no reason to
# keep it separated anymore.
self.cfg.load()
self.client_id = self.cfg.get('User', 'client_id')
self.username = self.cfg.get('User', 'username')
self.priv_key = self.cfg.get('Application Paths',
'client_priv_key_file')
self.host = self.cfg.get('System', 'server_hostname')
self.port = self.cfg.getint('System', 'server_port')
self.server_certificate = self.cfg.get('Application Paths',
'server_certificate')
self.storage_hostname = self.cfg.get('System', 'storage_endpoint')
self.refused_declare_max = self.cfg.getint('System',
'refused_declare_max')
self.refused_declare_waiting_time = self.cfg.getint(
'System', 'refused_declare_waiting_time')
self.commit_threshold_seconds = self.cfg.getint(
'Client', 'commit_threshold_seconds')
self.commit_threshold_operations = self.cfg.getint(
'Client', 'commit_threshold_operations')
self.commit_threshold_bytes = self.cfg.getint(
'Client', 'commit_threshold_bytes')
temp = self.cfg.get('Application Paths', 'transaction_cache_db')
self.transaction_cache = TransactionCache(temp)
self.integrity_manager = IntegrityManager(None)
is_firt_startup = self._internal_facade.is_first_startup()
self.cryptoAdapter = Adapter(self.cfg,
self.warebox,
self._input_queue,
self._lockfile_fd,
enc_dir='enc',
first_startup=is_firt_startup)
self.worker_pool = WorkerPool(self.warebox, self, self.cfg,
self.cryptoAdapter)
self.temp_dir = self.cryptoAdapter.get_enc_dir()
self._ui_controller.update_config_info(self.cfg)
def run(self):
"""Implementation of the threading.Thread.run() method."""
self._started = True
try:
self.worker_pool.start_workers()
self.keepalive_timer.start()
self.current_state = StateRegister.get('DisconnectedState')
curr_basis = self.current_state._load_trusted_basis()
self.integrity_manager.setCurrentBasis(curr_basis)
self.logger.info(u'Current basis is: %s' % curr_basis)
self.current_state._on_entering()
self._internal_facade.set_global_status(GStatuses.NC_STOPPED)
if self.auto_start:
self._input_queue.put(Command('CONNECT'), 'sessioncommand')
self.cryptoAdapter.start()
self._scheduler.schedule_action(self.check_encrypted_folder,
name='check_encrypted_folder',
seconds=5,
repeating=True)
# The event loop
self._main_loop()
except UnexpectedMessageException as e:
self.logger.critical(
u"Received an unexpected message from the Server while in "
u"state '%s': %s. Forcing termination." %
(self.current_state.__class__, str(e)))
raise
except ProtocolException as e:
self.logger.critical(
u"Detected an unrecoverable error, forcing termination: %s" %
str(e))
# Pre-emptive release, just stop before messing up the server
self.release_network_resources()
raise
except Exception as e:
self.logger.critical(
u"Forcing termination due to uncaught exception '%s': %s" %
(e.__class__, e))
self.logger.debug(u"Last error stacktrace:\n%s" %
traceback.format_exc())
# Pre-emptive release, just stop before messing up the server
self.release_network_resources()
raise
def _main_loop(self):
"""
The event loop.
A loop that contiuosly calls the current state of the state
machine, executing its logic.
It exits when self.terminate() is called.
"""
while not self.must_die.is_set():
next_state = self.current_state.do_execute()
if next_state != self.current_state:
self.current_state._on_leaving()
next_state._on_entering()
self.current_state = next_state
def check_encrypted_folder(self):
"""
Check if encryption preconditions are satisfied, try to
satisfy them otherwise.
This method is meant to be asynchronously called by a timer,
precisely to be registered into self._scheduler.
"""
if not self.cryptoAdapter.check_precondition(self._ui_controller):
self._internal_facade.terminate()
def acquire_network_resources(self):
"""
Configure a network connection to the server.
The resulting socket is handled by two instances of
ServerConnectionReader and ServerConnectionWriter, which are
created and run as well. It is possibile to send/receive
message to/from them through the queues self.output_message_queue
and self._input_queue (servermessage).
Note: this private method should be actually "protected" for
the ServerSession states, but Python doesn't have such a
protection level.
"""
try:
self.logger.debug(u"Creating a socket on %s:%s", self.host,
self.port)
sock = socket.create_connection((self.host, self.port), timeout=10)
ca_chain = os.path.abspath(self.server_certificate)
self.sock = ssl.wrap_socket(sock,
cert_reqs=ssl.CERT_REQUIRED,
ca_certs=ca_chain,
ssl_version=ssl.PROTOCOL_TLSv1)
self.sock.setblocking(True)
match_hostname(self.sock.getpeercert(), self.host)
except CertificateError as e:
self.logger.critical(u"SSL certificate validation failed: %s" % e)
raise ssl.SSLError(e)
except socket.error as exception:
self.logger.debug(u"Error opening SSL socket: %s" % exception)
self.logger.warning(
u"Unable to connect, re-trying in %s seconds." %
self.reconnection_time)
self.reconnection_time = stoppable_exponential_backoff_waiting(
self.reconnection_time, self.must_die, 10)
self.num_connection_attempts += 1
return False
except socket.timeout as exception:
self.logger.debug(u"Socket timeout: %s" % exception)
self.logger.warning(
u"Unable to connect, re-trying in %s seconds." %
self.reconnection_time)
self.reconnection_time = stoppable_exponential_backoff_waiting(
self.reconnection_time, self.must_die, 10)
self.num_connection_attempts += 1
return False
self.connection_reader = ServerConnectionReader(
self._input_queue, self.input_keepalive_queue, self.sock)
self.connection_writer = ServerConnectionWriter(
self._input_queue, self.output_message_queue, self.sock)
self.connection_reader.start()
self.connection_writer.start()
self.reconnection_time = 1
return True
def release_network_resources(self):
"""
Close the active connection to the server, if any.
Note: this private method should be actually "protected" for
the ServerSession states, but Python doesn't have such a
protection level.
"""
try:
self.connection_reader.terminate()
except AttributeError:
pass
try:
self.connection_writer.terminate()
except AttributeError:
pass
try:
self.sock.shutdown(socket.SHUT_RDWR)
self.sock.close()
except socket.error:
# Shutdown yelds an error on already closed sockets
pass
except AttributeError:
pass
def commit(self):
"""Make the client commit the current transaction."""
self._input_queue.put(Command('USERCOMMIT'), 'usercommand')
def connect(self):
"""Make the client connect to the server."""
self._input_queue.put(Command('CONNECT'), 'usercommand')
def disconnect(self):
"""Make the client disconnect from the client, if connected."""
# TODO: this method has been temporarly replaced by PAUSE
self._input_queue.put(Command('DISCONNECT'), 'usercommand')
def signal_free_worker(self):
"""
Tell ServerSession that a worker is free to receive new tasks.
This method is meant to be called by WorkerPool.
"""
self._input_queue.put(Command('WORKERFREE'), 'systemcommand')
def signal_download_integrity_error(self, operation, reason, expected_etag,
expected_basis, actual_etag,
computed_basis):
"""Tell ServerSession that the integrity check of a downloaded
file has failed.
This method is meant to be called by Workers.
@param operation:
Instance of PathnameOperation. Remember that it
contains also its Proof object.
@param reason:
String shortly describing the error.
@param expected_etag:
The etag the file was expected to have. It's the one
communicated by the server in its file list.
@param expected_basis:
The trusted basis. It's the one the user had
accepted when the sync started.
@param actual_etag:
The etag the file has turned to have after being
downloaded.
@param computed_basis:
The basis returned by the IntegrityManager when
computing the given proof object. Possibly None.
"""
cmd = Command('INTEGRITYERRORONDOWNLOAD')
cmd.operation = operation
cmd.proof = operation.download_info['proof']
cmd.reason = reason
cmd.expected_etag = expected_etag
cmd.expected_basis = expected_basis
cmd.actual_etag = actual_etag
cmd.computed_basis = computed_basis
self._input_queue.put(cmd, 'systemcommand')
def signal_deletelocal_integrity_error(self, pathname, proof, reason,
expected_basis, computed_basis):
"""Tell ServerSession that the integrity check of a
pathname to delete locally has failed.
This method is meant to be called by Workers.
@param pathname:
String representing the pathname.
@param reason:
String shortly describing the error.
@param expected_basis:
The trusted basis. It's the one the user had
accepted when the sync started.
@param computed_basis:
The basis returned by the IntegrityManager when
computing the given proof object. Possibly None.
"""
cmd = Command('INTEGRITYERRORONDELETELOCAL')
cmd.pathname = pathname
cmd.proof = proof
cmd.reason = reason
cmd.expected_basis = expected_basis
cmd.computed_basis = computed_basis
self._input_queue.put(cmd, 'systemcommand')
def get_current_basis(self):
"""
Return the current trusted basis.
@return The current trusted basis.
"""
with self._basis_lock:
return self._current_basis
def print_transaction(self):
"""
Print the list of operations in the current transaction.
Debug method, it prints to stdout and thus works only when the
application is attached to a console.
"""
self.transaction.print_all()
def terminate(self):
"""
Termination routine for this component.
Stops the running thread and releases any acquired resource.
"""
self.logger.debug(u"Terminating Server Session...")
if self._started:
self.must_die.set()
self.worker_pool.terminate()
self._input_queue.put(Command('TERMINATE'), 'usercommand')
self.transaction.can_be_committed.set()
self.join() if self is not threading.current_thread() else None
self.keepalive_timer.terminate()
self.release_network_resources()
self.cryptoAdapter.terminate()
self.logger.debug(u"Server Session terminanted.")