pythongh-118332: Fix deadlock involving stop the world

Include/internal/pycore_lock.h

@@ -150,8 +150,10 @@ PyAPI_FUNC(void) PyEvent_Wait(PyEvent *evt);
 
 // Wait for the event to be set, or until the timeout expires. If the event is
 // already set, then this returns immediately. Returns 1 if the event was set,
-// and 0 if the timeout expired or thread was interrupted.
-PyAPI_FUNC(int) PyEvent_WaitTimed(PyEvent *evt, PyTime_t timeout_ns);
+// and 0 if the timeout expired or thread was interrupted. If `detach` is
+// true, then the thread will detach/release the GIL while waiting.
+PyAPI_FUNC(int)
+PyEvent_WaitTimed(PyEvent *evt, PyTime_t timeout_ns, int detach);
 
 // _PyRawMutex implements a word-sized mutex that that does not depend on the
 // parking lot API, and therefore can be used in the parking lot

Modules/_threadmodule.c

@@ -501,7 +501,8 @@ ThreadHandle_join(ThreadHandle *self, PyTime_t timeout_ns)
 
     // Wait until the deadline for the thread to exit.
     PyTime_t deadline = timeout_ns != -1 ? _PyDeadline_Init(timeout_ns) : 0;
-    while (!PyEvent_WaitTimed(&self->thread_is_exiting, timeout_ns)) {
+    int detach = 1;
+    while (!PyEvent_WaitTimed(&self->thread_is_exiting, timeout_ns, detach)) {
         if (deadline) {
             // _PyDeadline_Get will return a negative value if the deadline has
             // been exceeded.

Python/lock.c

@@ -277,12 +277,12 @@ _PyEvent_Notify(PyEvent *evt)
 void
 PyEvent_Wait(PyEvent *evt)
 {
-    while (!PyEvent_WaitTimed(evt, -1))
+    while (!PyEvent_WaitTimed(evt, -1, /*detach=*/1))
         ;
 }
 
 int
-PyEvent_WaitTimed(PyEvent *evt, PyTime_t timeout_ns)
+PyEvent_WaitTimed(PyEvent *evt, PyTime_t timeout_ns, int detach)
 {
     for (;;) {
         uint8_t v = _Py_atomic_load_uint8(&evt->v);
@@ -298,7 +298,7 @@ PyEvent_WaitTimed(PyEvent *evt, PyTime_t timeout_ns)
 
         uint8_t expected = _Py_HAS_PARKED;
         (void) _PyParkingLot_Park(&evt->v, &expected, sizeof(evt->v),
-                                  timeout_ns, NULL, 1);
+                                  timeout_ns, NULL, detach);
 
         return _Py_atomic_load_uint8(&evt->v) == _Py_LOCKED;
     }

Python/pystate.c

@@ -2227,7 +2227,8 @@ stop_the_world(struct _stoptheworld_state *stw)
         }
 
         PyTime_t wait_ns = 1000*1000;  // 1ms (arbitrary, may need tuning)
-        if (PyEvent_WaitTimed(&stw->stop_event, wait_ns)) {
+        int detach = 0;
+        if (PyEvent_WaitTimed(&stw->stop_event, wait_ns, detach)) {
             assert(stw->thread_countdown == 0);
             break;
         }