Skip to content

gh-144763: Don't detach the GIL in tracemalloc #144779

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
vstinner merged 5 commits into from
Feb 18, 2026
+19 −7
Merged

gh-144763: Don't detach the GIL in tracemalloc #144779

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
00c69d2
gh-144763: Don't detach the GIL in tracemalloc
vstinner Feb 13, 2026
a91fc22
Fix race condition in _PyTraceMalloc_Stop()
vstinner Feb 13, 2026
78ceae3
Rephase the NEWS entry using "attached thread state"
vstinner Feb 16, 2026
609f7fb
Update Python/tracemalloc.c
vstinner Feb 18, 2026
ff3241a
Fix typos
vstinner Feb 18, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions Misc/NEWS.d/next/Library/2026-02-13-11-14-18.gh-issue-144763.cDwnEE.rst
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,2 @@
Fix a race condition in :mod:`tracemalloc`: it no longer detaches the attached
thread state to acquire its internal lock. Patch by Victor Stinner.
24 changes: 17 additions & 7 deletions Python/tracemalloc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ static int _PyTraceMalloc_TraceRef(PyObject *op, PyRefTracerEvent event,
the GIL held from PyMem_RawFree(). It cannot acquire the lock because it
would introduce a deadlock in _PyThreadState_DeleteCurrent(). */
#define tables_lock _PyRuntime.tracemalloc.tables_lock
#define TABLES_LOCK() PyMutex_Lock(&tables_lock)
#define TABLES_LOCK() PyMutex_LockFlags(&tables_lock, _Py_LOCK_DONT_DETACH)
#define TABLES_UNLOCK() PyMutex_Unlock(&tables_lock)


Expand Down Expand Up @@ -224,13 +224,20 @@ tracemalloc_get_frame(_PyInterpreterFrame *pyframe, frame_t *frame)
assert(PyStackRef_CodeCheck(pyframe->f_executable));
frame->filename = &_Py_STR(anon_unknown);

int lineno = PyUnstable_InterpreterFrame_GetLine(pyframe);
int lineno = -1;
PyCodeObject *code = _PyFrame_GetCode(pyframe);
// PyUnstable_InterpreterFrame_GetLine() cannot but used, since it uses
// a critical section which can trigger a deadlock.
Copy link
Contributor

@kumaraditya303 kumaraditya303 Feb 14, 2026 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the problem is that critical sections requires an active thread state and this code can be called with detached thread state iirc.

Copy link
Member Author

@vstinner vstinner Feb 16, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

tracemalloc_get_frame() is called with an attached thread state, see the caller traceback_get_frames() which has the code:

    PyThreadState *tstate = _PyThreadState_GET();
    assert(tstate != NULL);

Example of deadlock when running #144763 (comment) reproducer on a free-threaded build:

  • Thread A: _PyCode_GetTLBC() => ... => tracemalloc_alloc(): TABLES_LOCK()
  • Thread B: _PyTraceMalloc_TraceRef() => ... => tracemalloc_get_frame() => PyUnstable_InterpreterFrame_GetLine() => PyCode_Addr2Line(): Py_BEGIN_CRITICAL_SECTION(co)

Locks:

  • Thread A is in a Py_BEGIN_CRITICAL_SECTION(co) critical section (_PyCode_GetTLBC()) and waits for TABLES_LOCK().
  • Thread B has TABLES_LOCK() lock and waits for Py_BEGIN_CRITICAL_SECTION(co) critical section.

Thread A and thread B want to use Py_BEGIN_CRITICAL_SECTION(co) on the same code object (0x20000844810).

=> deadlock :-(

Copy link
Contributor

@colesbury colesbury Feb 17, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There are probably some thread-safety issues here if the code object is being concurrently instrumented or deinstrumented. I don't see a better way of handling it, so let's leave a TODO or note.

int lasti = _PyFrame_SafeGetLasti(pyframe);
if (lasti >= 0) {
lineno = _PyCode_SafeAddr2Line(code, lasti);
}
if (lineno < 0) {
lineno = 0;
}
frame->lineno = (unsigned int)lineno;

PyObject *filename = _PyFrame_GetCode(pyframe)->co_filename;
PyObject *filename = code->co_filename;
if (filename == NULL) {
#ifdef TRACE_DEBUG
tracemalloc_error("failed to get the filename of the code object");
Expand Down Expand Up @@ -863,7 +870,8 @@ _PyTraceMalloc_Stop(void)
TABLES_LOCK();

if (!tracemalloc_config.tracing) {
goto done;
TABLES_UNLOCK();
return;
}

/* stop tracing Python memory allocations */
Expand All @@ -880,10 +888,12 @@ _PyTraceMalloc_Stop(void)
raw_free(tracemalloc_traceback);
tracemalloc_traceback = NULL;

(void)PyRefTracer_SetTracer(NULL, NULL);

done:
TABLES_UNLOCK();

// Call it after TABLES_UNLOCK() since it calls _PyEval_StopTheWorldAll()
// which would lead to a deadlock with TABLES_LOCK() which doesn't detach
// the thread state.
(void)PyRefTracer_SetTracer(NULL, NULL);
}


Expand Down
Loading