gh-104341: Add a Separate "Running" Lock for Each Thread (gh-104754)

Having a separate lock means Thread.join() doesn't need to wait for the thread to be cleaned up first.  It can wait for the thread's Python target to finish running.  This gives us some flexibility in how we clean up threads.

(This is a minor cleanup as part of a fix for gh-104341.)

Author: ericsnowcurrently

Lib/test/test_threading.py

@@ -747,7 +747,7 @@ def f():
         rc, out, err = assert_python_ok("-c", code)
         self.assertEqual(err, b"")
 
-    def test_tstate_lock(self):
+    def test_running_lock(self):
         # Test an implementation detail of Thread objects.
         started = _thread.allocate_lock()
         finish = _thread.allocate_lock()
@@ -757,29 +757,29 @@ def f():
             started.release()
             finish.acquire()
             time.sleep(0.01)
-        # The tstate lock is None until the thread is started
+        # The running lock is None until the thread is started
         t = threading.Thread(target=f)
-        self.assertIs(t._tstate_lock, None)
+        self.assertIs(t._running_lock, None)
         t.start()
         started.acquire()
         self.assertTrue(t.is_alive())
-        # The tstate lock can't be acquired when the thread is running
+        # The running lock can't be acquired when the thread is running
         # (or suspended).
-        tstate_lock = t._tstate_lock
-        self.assertFalse(tstate_lock.acquire(timeout=0), False)
+        running_lock = t._running_lock
+        self.assertFalse(running_lock.acquire(timeout=0), False)
         finish.release()
         # When the thread ends, the state_lock can be successfully
         # acquired.
-        self.assertTrue(tstate_lock.acquire(timeout=support.SHORT_TIMEOUT), False)
-        # But is_alive() is still True:  we hold _tstate_lock now, which
-        # prevents is_alive() from knowing the thread's end-of-life C code
+        self.assertTrue(running_lock.acquire(timeout=support.SHORT_TIMEOUT), False)
+        # But is_alive() is still True:  we hold _running_lock now, which
+        # prevents is_alive() from knowing the thread's Python code
         # is done.
         self.assertTrue(t.is_alive())
         # Let is_alive() find out the C code is done.
-        tstate_lock.release()
+        running_lock.release()
         self.assertFalse(t.is_alive())
-        # And verify the thread disposed of _tstate_lock.
-        self.assertIsNone(t._tstate_lock)
+        # And verify the thread disposed of _running_lock.
+        self.assertIsNone(t._running_lock)
         t.join()
 
     def test_repr_stopped(self):

Lib/threading.py

@@ -908,6 +908,7 @@ class is implemented.
         self._ident = None
         if _HAVE_THREAD_NATIVE_ID:
             self._native_id = None
+        self._running_lock = None
         self._tstate_lock = None
         self._started = Event()
         self._is_stopped = False
@@ -926,13 +927,17 @@ def _reset_internal_locks(self, is_alive):
             # bpo-42350: If the fork happens when the thread is already stopped
             # (ex: after threading._shutdown() has been called), _tstate_lock
             # is None. Do nothing in this case.
+            if self._running_lock is not None:
+                self._running_lock._at_fork_reinit()
+                self._running_lock.acquire()
             if self._tstate_lock is not None:
                 self._tstate_lock._at_fork_reinit()
                 self._tstate_lock.acquire()
         else:
             # The thread isn't alive after fork: it doesn't have a tstate
             # anymore.
             self._is_stopped = True
+            self._running_lock = None
             self._tstate_lock = None
 
     def __repr__(self):
@@ -1019,6 +1024,14 @@ def _set_ident(self):
         def _set_native_id(self):
             self._native_id = get_native_id()
 
+    def _set_running_lock(self):
+        """
+        Set a lock object which will be released by the interpreter when
+        the target func has finished running.
+        """
+        self._running_lock = _allocate_lock()
+        self._running_lock.acquire()
+
     def _set_tstate_lock(self):
         """
         Set a lock object which will be released by the interpreter when
@@ -1035,6 +1048,7 @@ def _set_tstate_lock(self):
     def _bootstrap_inner(self):
         try:
             self._set_ident()
+            self._set_running_lock()
             self._set_tstate_lock()
             if _HAVE_THREAD_NATIVE_ID:
                 self._set_native_id()
@@ -1054,29 +1068,29 @@ def _bootstrap_inner(self):
                 self._invoke_excepthook(self)
         finally:
             self._delete()
+            self._running_lock.release()
 
     def _stop(self):
         # After calling ._stop(), .is_alive() returns False and .join() returns
-        # immediately.  ._tstate_lock must be released before calling ._stop().
+        # immediately.  ._running_lock must be released before calling ._stop().
         #
-        # Normal case:  C code at the end of the thread's life
-        # (release_sentinel in _threadmodule.c) releases ._tstate_lock, and
-        # that's detected by our ._wait_for_tstate_lock(), called by .join()
+        # Normal case:  ._bootstrap_inner() releases ._running_lock, and
+        # that's detected by our ._wait_for_running_lock(), called by .join()
         # and .is_alive().  Any number of threads _may_ call ._stop()
         # simultaneously (for example, if multiple threads are blocked in
         # .join() calls), and they're not serialized.  That's harmless -
         # they'll just make redundant rebindings of ._is_stopped and
-        # ._tstate_lock.  Obscure:  we rebind ._tstate_lock last so that the
-        # "assert self._is_stopped" in ._wait_for_tstate_lock() always works
-        # (the assert is executed only if ._tstate_lock is None).
+        # ._running_lock.  Obscure:  we rebind ._running_lock last so that the
+        # "assert self._is_stopped" in ._wait_for_running_lock() always works
+        # (the assert is executed only if ._running_lock is None).
         #
-        # Special case:  _main_thread releases ._tstate_lock via this
+        # Special case:  _main_thread releases ._running_lock via this
         # module's _shutdown() function.
-        lock = self._tstate_lock
+        lock = self._running_lock
         if lock is not None:
             assert not lock.locked()
         self._is_stopped = True
-        self._tstate_lock = None
+        self._running_lock = None
         if not self.daemon:
             with _shutdown_locks_lock:
                 # Remove our lock and other released locks from _shutdown_locks
@@ -1123,20 +1137,17 @@ def join(self, timeout=None):
             raise RuntimeError("cannot join current thread")
 
         if timeout is None:
-            self._wait_for_tstate_lock()
+            self._wait_for_running_lock()
         else:
             # the behavior of a negative timeout isn't documented, but
             # historically .join(timeout=x) for x<0 has acted as if timeout=0
-            self._wait_for_tstate_lock(timeout=max(timeout, 0))
-
-    def _wait_for_tstate_lock(self, block=True, timeout=-1):
-        # Issue #18808: wait for the thread state to be gone.
-        # At the end of the thread's life, after all knowledge of the thread
-        # is removed from C data structures, C code releases our _tstate_lock.
-        # This method passes its arguments to _tstate_lock.acquire().
-        # If the lock is acquired, the C code is done, and self._stop() is
-        # called.  That sets ._is_stopped to True, and ._tstate_lock to None.
-        lock = self._tstate_lock
+            self._wait_for_running_lock(timeout=max(timeout, 0))
+
+    def _wait_for_running_lock(self, block=True, timeout=-1):
+        # This method passes its arguments to _running_lock.acquire().
+        # If the lock is acquired, the python code is done, and self._stop() is
+        # called.  That sets ._is_stopped to True, and ._running_lock to None.
+        lock = self._running_lock
         if lock is None:
             # already determined that the C code is done
             assert self._is_stopped
@@ -1207,7 +1218,7 @@ def is_alive(self):
         assert self._initialized, "Thread.__init__() not called"
         if self._is_stopped or not self._started.is_set():
             return False
-        self._wait_for_tstate_lock(False)
+        self._wait_for_running_lock(False)
         return not self._is_stopped
 
     @property
@@ -1417,7 +1428,7 @@ class _MainThread(Thread):
 
     def __init__(self):
         Thread.__init__(self, name="MainThread", daemon=False)
-        self._set_tstate_lock()
+        self._set_running_lock()
         self._started.set()
         self._set_ident()
         if _HAVE_THREAD_NATIVE_ID:
@@ -1558,7 +1569,7 @@ def _shutdown():
     # dubious, but some code does it.  We can't wait for C code to release
     # the main thread's tstate_lock - that won't happen until the interpreter
     # is nearly dead.  So we release it here.  Note that just calling _stop()
-    # isn't enough:  other threads may already be waiting on _tstate_lock.
+    # isn't enough:  other threads may already be waiting on _running_lock.
     if _main_thread._is_stopped:
         # _shutdown() was already called
         return
@@ -1573,12 +1584,13 @@ def _shutdown():
 
     # Main thread
     if _main_thread.ident == get_ident():
-        tlock = _main_thread._tstate_lock
-        # The main thread isn't finished yet, so its thread state lock can't
+        assert _main_thread._tstate_lock is None
+        running_lock = _main_thread._running_lock
+        # The main thread isn't finished yet, so its running lock can't
         # have been released.
-        assert tlock is not None
-        assert tlock.locked()
-        tlock.release()
+        assert running_lock is not None
+        assert running_lock.locked()
+        running_lock.release()
         _main_thread._stop()
     else:
         # bpo-1596321: _shutdown() must be called in the main thread.