[3.13] gh-137400: Fix thread-safety issues when profiling all threads (gh-137518)

Include/internal/pycore_ceval.h

@@ -19,8 +19,10 @@ struct _ceval_runtime_state;
 
 // Export for '_lsprof' shared extension
 PyAPI_FUNC(int) _PyEval_SetProfile(PyThreadState *tstate, Py_tracefunc func, PyObject *arg);
+extern int _PyEval_SetProfileAllThreads(PyInterpreterState *interp, Py_tracefunc func, PyObject *arg);
 
 extern int _PyEval_SetTrace(PyThreadState *tstate, Py_tracefunc func, PyObject *arg);
+extern int _PyEval_SetTraceAllThreads(PyInterpreterState *interp, Py_tracefunc func, PyObject *arg);
 
 extern int _PyEval_SetOpcodeTrace(PyFrameObject *f, bool enable);
 

Include/internal/pycore_interp.h

@@ -262,8 +262,8 @@ struct _is {
     PyDict_WatchCallback builtins_dict_watcher;
 
     _Py_GlobalMonitors monitors;
-    bool sys_profile_initialized;
-    bool sys_trace_initialized;
+    _PyOnceFlag sys_profile_once_flag;
+    _PyOnceFlag sys_trace_once_flag;
     Py_ssize_t sys_profiling_threads; /* Count of threads with c_profilefunc set */
     Py_ssize_t sys_tracing_threads; /* Count of threads with c_tracefunc set */
     PyObject *monitoring_callables[PY_MONITORING_TOOL_IDS][_PY_MONITORING_EVENTS];

Lib/test/test_free_threading/test_monitoring.py

@@ -193,6 +193,32 @@ def during_threads(self):
         self.set = not self.set
 
 
+@threading_helper.requires_working_threading()
+class SetProfileAllThreadsMultiThreaded(InstrumentationMultiThreadedMixin, TestCase):
+    """Uses threading.setprofile_all_threads and repeatedly toggles instrumentation on and off"""
+
+    def setUp(self):
+        self.set = False
+        self.called = False
+
+    def after_test(self):
+        self.assertTrue(self.called)
+
+    def tearDown(self):
+        threading.setprofile_all_threads(None)
+
+    def trace_func(self, frame, event, arg):
+        self.called = True
+        return self.trace_func
+
+    def during_threads(self):
+        if self.set:
+            threading.setprofile_all_threads(self.trace_func)
+        else:
+            threading.setprofile_all_threads(None)
+        self.set = not self.set
+
+
 @threading_helper.requires_working_threading()
 class SetProfileAllMultiThreaded(TestCase):
     def test_profile_all_threads(self):

Misc/NEWS.d/next/Core and Builtins/2025-08-07-09-52-19.gh-issue-137400.AK1dy-.rst

@@ -0,0 +1,5 @@
+Fix a crash in the :term:`free threading` build when disabling profiling or
+tracing across all threads with :c:func:`PyEval_SetProfileAllThreads` or
+:c:func:`PyEval_SetTraceAllThreads` or their Python equivalents
+:func:`threading.settrace_all_threads` and
+:func:`threading.setprofile_all_threads`.

Python/bytecodes.c

@@ -148,7 +148,15 @@ dummy_func(
 
         tier1 inst(RESUME, (--)) {
             assert(frame == tstate->current_frame);
-            if (tstate->tracing == 0) {
+            #ifdef Py_GIL_DISABLED
+            // For thread-safety, we need to check instrumentation version
+            // even when tracing. Otherwise, another thread may concurrently
+            // re-write the bytecode while we are executing this function.
+            int check_instrumentation = 1;
+            #else
+            int check_instrumentation = (tstate->tracing == 0);
+            #endif
+            if (check_instrumentation) {
                 uintptr_t global_version =
                     _Py_atomic_load_uintptr_relaxed(&tstate->eval_breaker) &
                     ~_PY_EVAL_EVENTS_MASK;

Python/ceval.c

@@ -2334,21 +2334,10 @@ PyEval_SetProfile(Py_tracefunc func, PyObject *arg)
 void
 PyEval_SetProfileAllThreads(Py_tracefunc func, PyObject *arg)
 {
-    PyThreadState *this_tstate = _PyThreadState_GET();
-    PyInterpreterState* interp = this_tstate->interp;
-
-    _PyRuntimeState *runtime = &_PyRuntime;
-    HEAD_LOCK(runtime);
-    PyThreadState* ts = PyInterpreterState_ThreadHead(interp);
-    HEAD_UNLOCK(runtime);
-
-    while (ts) {
-        if (_PyEval_SetProfile(ts, func, arg) < 0) {
-            PyErr_FormatUnraisable("Exception ignored in PyEval_SetProfileAllThreads");
-        }
-        HEAD_LOCK(runtime);
-        ts = PyThreadState_Next(ts);
-        HEAD_UNLOCK(runtime);
+    PyInterpreterState *interp = _PyInterpreterState_GET();
+    if (_PyEval_SetProfileAllThreads(interp, func, arg) < 0) {
+        /* Log _PySys_Audit() error */
+        PyErr_FormatUnraisable("Exception ignored in PyEval_SetProfileAllThreads");
     }
 }
 
@@ -2365,21 +2354,10 @@ PyEval_SetTrace(Py_tracefunc func, PyObject *arg)
 void
 PyEval_SetTraceAllThreads(Py_tracefunc func, PyObject *arg)
 {
-    PyThreadState *this_tstate = _PyThreadState_GET();
-    PyInterpreterState* interp = this_tstate->interp;
-
-    _PyRuntimeState *runtime = &_PyRuntime;
-    HEAD_LOCK(runtime);
-    PyThreadState* ts = PyInterpreterState_ThreadHead(interp);
-    HEAD_UNLOCK(runtime);
-
-    while (ts) {
-        if (_PyEval_SetTrace(ts, func, arg) < 0) {
-            PyErr_FormatUnraisable("Exception ignored in PyEval_SetTraceAllThreads");
-        }
-        HEAD_LOCK(runtime);
-        ts = PyThreadState_Next(ts);
-        HEAD_UNLOCK(runtime);
+    PyInterpreterState *interp = _PyInterpreterState_GET();
+    if (_PyEval_SetTraceAllThreads(interp, func, arg) < 0) {
+        /* Log _PySys_Audit() error */
+        PyErr_FormatUnraisable("Exception ignored in PyEval_SetTraceAllThreads");
     }
 }
 

Python/instrumentation.c

@@ -1012,6 +1012,8 @@ set_version_raw(uintptr_t *ptr, uint32_t version)
 static void
 set_global_version(PyThreadState *tstate, uint32_t version)
 {
+    ASSERT_WORLD_STOPPED();
+
     assert((version & _PY_EVAL_EVENTS_MASK) == 0);
     PyInterpreterState *interp = tstate->interp;
     set_version_raw(&interp->ceval.instrumentation_version, version);
@@ -1908,28 +1910,27 @@ _Py_Instrument(PyCodeObject *code, PyInterpreterState *interp)
 
 
 static int
-instrument_all_executing_code_objects(PyInterpreterState *interp) {
+instrument_all_executing_code_objects(PyInterpreterState *interp)
+{
     ASSERT_WORLD_STOPPED();
 
-    _PyRuntimeState *runtime = &_PyRuntime;
-    HEAD_LOCK(runtime);
-    PyThreadState* ts = PyInterpreterState_ThreadHead(interp);
-    HEAD_UNLOCK(runtime);
-    while (ts) {
+    int err = 0;
+    HEAD_LOCK(&_PyRuntime);
+    for (PyThreadState *ts = interp->threads.head; ts != NULL; ts = ts->next) {
         _PyInterpreterFrame *frame = ts->current_frame;
         while (frame) {
             if (frame->owner != FRAME_OWNED_BY_CSTACK) {
-                if (instrument_lock_held(_PyFrame_GetCode(frame), interp)) {
-                    return -1;
+                err = instrument_lock_held(_PyFrame_GetCode(frame), interp);
+                if (err) {
+                    goto done;
                 }
             }
             frame = frame->previous;
         }
-        HEAD_LOCK(runtime);
-        ts = PyThreadState_Next(ts);
-        HEAD_UNLOCK(runtime);
     }
-    return 0;
+done:
+    HEAD_UNLOCK(&_PyRuntime);
+    return err;
 }
 
 static void
@@ -1975,6 +1976,7 @@ check_tool(PyInterpreterState *interp, int tool_id)
 int
 _PyMonitoring_SetEvents(int tool_id, _PyMonitoringEventSet events)
 {
+    ASSERT_WORLD_STOPPED();
     assert(0 <= tool_id && tool_id < PY_MONITORING_TOOL_IDS);
     PyThreadState *tstate = _PyThreadState_GET();
     PyInterpreterState *interp = tstate->interp;
@@ -1983,33 +1985,28 @@ _PyMonitoring_SetEvents(int tool_id, _PyMonitoringEventSet events)
         return -1;
     }
 
-    int res;
-    _PyEval_StopTheWorld(interp);
     uint32_t existing_events = get_events(&interp->monitors, tool_id);
     if (existing_events == events) {
-        res = 0;
-        goto done;
+        return 0;
     }
     set_events(&interp->monitors, tool_id, events);
     uint32_t new_version = global_version(interp) + MONITORING_VERSION_INCREMENT;
     if (new_version == 0) {
         PyErr_Format(PyExc_OverflowError, "events set too many times");
-        res = -1;
-        goto done;
+        return -1;
     }
     set_global_version(tstate, new_version);
 #ifdef _Py_TIER2
     _Py_Executors_InvalidateAll(interp, 1);
 #endif
-    res = instrument_all_executing_code_objects(interp);
-done:
-    _PyEval_StartTheWorld(interp);
-    return res;
+    return instrument_all_executing_code_objects(interp);
 }
 
 int
 _PyMonitoring_SetLocalEvents(PyCodeObject *code, int tool_id, _PyMonitoringEventSet events)
 {
+    ASSERT_WORLD_STOPPED();
+
     assert(0 <= tool_id && tool_id < PY_MONITORING_TOOL_IDS);
     PyInterpreterState *interp = _PyInterpreterState_GET();
     assert(events < (1 << _PY_MONITORING_LOCAL_EVENTS));
@@ -2021,26 +2018,18 @@ _PyMonitoring_SetLocalEvents(PyCodeObject *code, int tool_id, _PyMonitoringEvent
         return -1;
     }
 
-    int res;
-    _PyEval_StopTheWorld(interp);
     if (allocate_instrumentation_data(code)) {
-        res = -1;
-        goto done;
+        return -1;
     }
 
     _Py_LocalMonitors *local = &code->_co_monitoring->local_monitors;
     uint32_t existing_events = get_local_events(local, tool_id);
     if (existing_events == events) {
-        res = 0;
-        goto done;
+        return 0;
     }
     set_local_events(local, tool_id, events);
 
-    res = force_instrument_lock_held(code, interp);
-
-done:
-    _PyEval_StartTheWorld(interp);
-    return res;
+    return force_instrument_lock_held(code, interp);
 }
 
 int
@@ -2238,7 +2227,11 @@ monitoring_set_events_impl(PyObject *module, int tool_id, int event_set)
         return NULL;
     }
     event_set &= ~C_RETURN_EVENTS;
-    if (_PyMonitoring_SetEvents(tool_id, event_set)) {
+    PyInterpreterState *interp = _PyInterpreterState_GET();
+    _PyEval_StopTheWorld(interp);
+    int err = _PyMonitoring_SetEvents(tool_id, event_set);
+    _PyEval_StartTheWorld(interp);
+    if (err) {
         return NULL;
     }
     Py_RETURN_NONE;
@@ -2315,7 +2308,11 @@ monitoring_set_local_events_impl(PyObject *module, int tool_id,
         return NULL;
     }
 
-    if (_PyMonitoring_SetLocalEvents((PyCodeObject*)code, tool_id, event_set)) {
+    PyInterpreterState *interp = _PyInterpreterState_GET();
+    _PyEval_StopTheWorld(interp);
+    int err = _PyMonitoring_SetLocalEvents((PyCodeObject*)code, tool_id, event_set);
+    _PyEval_StartTheWorld(interp);
+    if (err) {
         return NULL;
     }
     Py_RETURN_NONE;