gh-90622: Do not spawn ProcessPool workers on demand via fork method. (GH-91598) (#92495)

Do not spawn ProcessPool workers on demand when they spawn via fork.

This avoids potential deadlocks in the child processes due to forking from
a multithreaded process.
(cherry picked from commit ebb37fc3fd)

Co-authored-by: Gregory P. Smith <greg@krypto.org>
This commit is contained in:
Miss Islington (bot) 2022-05-08 10:14:14 -07:00 committed by GitHub
parent 5917e71017
commit 4270b7927d
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 49 additions and 11 deletions

View File

@ -652,6 +652,10 @@ class ProcessPoolExecutor(_base.Executor):
mp_context = mp.get_context() mp_context = mp.get_context()
self._mp_context = mp_context self._mp_context = mp_context
# https://github.com/python/cpython/issues/90622
self._safe_to_dynamically_spawn_children = (
self._mp_context.get_start_method(allow_none=False) != "fork")
if initializer is not None and not callable(initializer): if initializer is not None and not callable(initializer):
raise TypeError("initializer must be a callable") raise TypeError("initializer must be a callable")
self._initializer = initializer self._initializer = initializer
@ -714,6 +718,8 @@ class ProcessPoolExecutor(_base.Executor):
def _start_executor_manager_thread(self): def _start_executor_manager_thread(self):
if self._executor_manager_thread is None: if self._executor_manager_thread is None:
# Start the processes so that their sentinels are known. # Start the processes so that their sentinels are known.
if not self._safe_to_dynamically_spawn_children: # ie, using fork.
self._launch_processes()
self._executor_manager_thread = _ExecutorManagerThread(self) self._executor_manager_thread = _ExecutorManagerThread(self)
self._executor_manager_thread.start() self._executor_manager_thread.start()
_threads_wakeups[self._executor_manager_thread] = \ _threads_wakeups[self._executor_manager_thread] = \
@ -726,6 +732,23 @@ class ProcessPoolExecutor(_base.Executor):
process_count = len(self._processes) process_count = len(self._processes)
if process_count < self._max_workers: if process_count < self._max_workers:
# Assertion disabled as this codepath is also used to replace a
# worker that unexpectedly dies, even when using the 'fork' start
# method. That means there is still a potential deadlock bug. If a
# 'fork' mp_context worker dies, we'll be forking a new one when
# we know a thread is running (self._executor_manager_thread).
#assert self._safe_to_dynamically_spawn_children or not self._executor_manager_thread, 'https://github.com/python/cpython/issues/90622'
self._spawn_process()
def _launch_processes(self):
# https://github.com/python/cpython/issues/90622
assert not self._executor_manager_thread, (
'Processes cannot be fork()ed after the thread has started, '
'deadlock in the child processes could result.')
for _ in range(len(self._processes), self._max_workers):
self._spawn_process()
def _spawn_process(self):
p = self._mp_context.Process( p = self._mp_context.Process(
target=_process_worker, target=_process_worker,
args=(self._call_queue, args=(self._call_queue,
@ -755,6 +778,7 @@ class ProcessPoolExecutor(_base.Executor):
# Wake up queue management thread # Wake up queue management thread
self._executor_manager_thread_wakeup.wakeup() self._executor_manager_thread_wakeup.wakeup()
if self._safe_to_dynamically_spawn_children:
self._adjust_process_count() self._adjust_process_count()
self._start_executor_manager_thread() self._start_executor_manager_thread()
return f return f

View File

@ -497,10 +497,16 @@ class ProcessPoolShutdownTest(ExecutorShutdownTest):
lock.acquire() lock.acquire()
mp_context = self.get_context() mp_context = self.get_context()
if mp_context.get_start_method(allow_none=False) == "fork":
# fork pre-spawns, not on demand.
expected_num_processes = self.worker_count
else:
expected_num_processes = 3
sem = mp_context.Semaphore(0) sem = mp_context.Semaphore(0)
for _ in range(3): for _ in range(3):
self.executor.submit(acquire_lock, sem) self.executor.submit(acquire_lock, sem)
self.assertEqual(len(self.executor._processes), 3) self.assertEqual(len(self.executor._processes), expected_num_processes)
for _ in range(3): for _ in range(3):
sem.release() sem.release()
processes = self.executor._processes processes = self.executor._processes
@ -1021,6 +1027,8 @@ class ProcessPoolExecutorTest(ExecutorTest):
def test_idle_process_reuse_one(self): def test_idle_process_reuse_one(self):
executor = self.executor executor = self.executor
assert executor._max_workers >= 4 assert executor._max_workers >= 4
if self.get_context().get_start_method(allow_none=False) == "fork":
raise unittest.SkipTest("Incompatible with the fork start method.")
executor.submit(mul, 21, 2).result() executor.submit(mul, 21, 2).result()
executor.submit(mul, 6, 7).result() executor.submit(mul, 6, 7).result()
executor.submit(mul, 3, 14).result() executor.submit(mul, 3, 14).result()
@ -1029,6 +1037,8 @@ class ProcessPoolExecutorTest(ExecutorTest):
def test_idle_process_reuse_multiple(self): def test_idle_process_reuse_multiple(self):
executor = self.executor executor = self.executor
assert executor._max_workers <= 5 assert executor._max_workers <= 5
if self.get_context().get_start_method(allow_none=False) == "fork":
raise unittest.SkipTest("Incompatible with the fork start method.")
executor.submit(mul, 12, 7).result() executor.submit(mul, 12, 7).result()
executor.submit(mul, 33, 25) executor.submit(mul, 33, 25)
executor.submit(mul, 25, 26).result() executor.submit(mul, 25, 26).result()

View File

@ -0,0 +1,4 @@
Worker processes for :class:`concurrent.futures.ProcessPoolExecutor` are no
longer spawned on demand (a feature added in 3.9) when the multiprocessing
context start method is ``"fork"`` as that can lead to deadlocks in the
child processes due to a fork happening while threads are running.