diff --git a/Include/internal/pycore_backoff.h b/Include/internal/pycore_backoff.h
index 71066f1bd9f19b..7f60eb495080ae 100644
--- a/Include/internal/pycore_backoff.h
+++ b/Include/internal/pycore_backoff.h
@@ -22,33 +22,48 @@ extern "C" {
    Another use is for the Tier 2 optimizer to decide when to create
    a new Tier 2 trace (executor). Again, exponential backoff is used.
 
-   The 16-bit counter is structured as a 12-bit unsigned 'value'
-   and a 4-bit 'backoff' field. When resetting the counter, the
+   The 16-bit counter is structured as a 13-bit unsigned 'value'
+   and a 3-bit 'backoff' field. When resetting the counter, the
    backoff field is incremented (until it reaches a limit) and the
-   value is set to a bit mask representing the value 2**backoff - 1.
-   The maximum backoff is 12 (the number of bits in the value).
+   value is set to a bit mask representing some prime value - 1.
+   New values and backoffs for each backoff are calculated once
+   at compile time and saved to value_and_backoff_next table.
+   The maximum backoff is 6, since 7 is an UNREACHABLE_BACKOFF.
 
    There is an exceptional value which must not be updated, 0xFFFF.
 */
 
-#define BACKOFF_BITS 4
-#define MAX_BACKOFF 12
-#define UNREACHABLE_BACKOFF 15
-
-static inline bool
-is_unreachable_backoff_counter(_Py_BackoffCounter counter)
-{
-    return counter.value_and_backoff == UNREACHABLE_BACKOFF;
-}
+#define BACKOFF_BITS 3
+#define BACKOFF_MASK 7
+#define MAX_BACKOFF 6
+#define UNREACHABLE_BACKOFF 7
+#define MAX_VALUE 0x1FFF
+
+#define MAKE_VALUE_AND_BACKOFF(value, backoff) \
+    ((value << BACKOFF_BITS) | backoff)
+
+// For previous backoff b we use value x such that
+// x + 1 is near to 2**(2*b+1) and x + 1 is prime.
+static const uint16_t value_and_backoff_next[] = {
+    MAKE_VALUE_AND_BACKOFF(1, 1),
+    MAKE_VALUE_AND_BACKOFF(6, 2),
+    MAKE_VALUE_AND_BACKOFF(30, 3),
+    MAKE_VALUE_AND_BACKOFF(126, 4),
+    MAKE_VALUE_AND_BACKOFF(508, 5),
+    MAKE_VALUE_AND_BACKOFF(2052, 6),
+    // We use the same backoff counter for all backoffs >= MAX_BACKOFF.
+    MAKE_VALUE_AND_BACKOFF(8190, 6),
+    MAKE_VALUE_AND_BACKOFF(8190, 6),
+};
 
 static inline _Py_BackoffCounter
 make_backoff_counter(uint16_t value, uint16_t backoff)
 {
-    assert(backoff <= 15);
-    assert(value <= 0xFFF);
-    _Py_BackoffCounter result;
-    result.value_and_backoff = (value << BACKOFF_BITS) | backoff;
-    return result;
+    assert(backoff <= UNREACHABLE_BACKOFF);
+    assert(value <= MAX_VALUE);
+    return ((_Py_BackoffCounter){
+        .value_and_backoff = MAKE_VALUE_AND_BACKOFF(value, backoff)
+    });
 }
 
 static inline _Py_BackoffCounter
@@ -62,14 +77,11 @@ forge_backoff_counter(uint16_t counter)
 static inline _Py_BackoffCounter
 restart_backoff_counter(_Py_BackoffCounter counter)
 {
-    assert(!is_unreachable_backoff_counter(counter));
-    int backoff = counter.value_and_backoff & 15;
-    if (backoff < MAX_BACKOFF) {
-        return make_backoff_counter((1 << (backoff + 1)) - 1, backoff + 1);
-    }
-    else {
-        return make_backoff_counter((1 << MAX_BACKOFF) - 1, MAX_BACKOFF);
-    }
+    uint16_t backoff = counter.value_and_backoff & BACKOFF_MASK;
+    assert(backoff <= MAX_BACKOFF);
+    return ((_Py_BackoffCounter){
+        .value_and_backoff = value_and_backoff_next[backoff]
+    });
 }
 
 static inline _Py_BackoffCounter
@@ -113,7 +125,7 @@ trigger_backoff_counter(void)
 // as we always end up tracing the loop iteration's
 // exhaustion iteration. Which aborts our current tracer.
 #define JUMP_BACKWARD_INITIAL_VALUE 4000
-#define JUMP_BACKWARD_INITIAL_BACKOFF 12
+#define JUMP_BACKWARD_INITIAL_BACKOFF 6
 static inline _Py_BackoffCounter
 initial_jump_backoff_counter(void)
 {
@@ -126,7 +138,7 @@ initial_jump_backoff_counter(void)
  * otherwise when a side exit warms up we may construct
  * a new trace before the Tier 1 code has properly re-specialized. */
 #define SIDE_EXIT_INITIAL_VALUE 4000
-#define SIDE_EXIT_INITIAL_BACKOFF 12
+#define SIDE_EXIT_INITIAL_BACKOFF 6
 
 static inline _Py_BackoffCounter
 initial_temperature_backoff_counter(void)
diff --git a/Include/internal/pycore_critical_section.h b/Include/internal/pycore_critical_section.h
index 2601de40737e85..60b6fc4a72e88f 100644
--- a/Include/internal/pycore_critical_section.h
+++ b/Include/internal/pycore_critical_section.h
@@ -32,7 +32,7 @@ extern "C" {
         const bool _should_lock_cs = PyList_CheckExact(_orig_seq);      \
         PyCriticalSection _cs;                                          \
         if (_should_lock_cs) {                                          \
-            _PyCriticalSection_Begin(&_cs, _orig_seq);                  \
+            PyCriticalSection_Begin(&_cs, _orig_seq);                  \
         }
 
 # define Py_END_CRITICAL_SECTION_SEQUENCE_FAST()                        \
@@ -77,10 +77,10 @@ _PyCriticalSection_Resume(PyThreadState *tstate);
 
 // (private) slow path for locking the mutex
 PyAPI_FUNC(void)
-_PyCriticalSection_BeginSlow(PyCriticalSection *c, PyMutex *m);
+_PyCriticalSection_BeginSlow(PyThreadState *tstate, PyCriticalSection *c, PyMutex *m);
 
 PyAPI_FUNC(void)
-_PyCriticalSection2_BeginSlow(PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2,
+_PyCriticalSection2_BeginSlow(PyThreadState *tstate, PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2,
                              int is_m1_locked);
 
 PyAPI_FUNC(void)
@@ -95,34 +95,30 @@ _PyCriticalSection_IsActive(uintptr_t tag)
 }
 
 static inline void
-_PyCriticalSection_BeginMutex(PyCriticalSection *c, PyMutex *m)
+_PyCriticalSection_BeginMutex(PyThreadState *tstate, PyCriticalSection *c, PyMutex *m)
 {
     if (PyMutex_LockFast(m)) {
-        PyThreadState *tstate = _PyThreadState_GET();
         c->_cs_mutex = m;
         c->_cs_prev = tstate->critical_section;
         tstate->critical_section = (uintptr_t)c;
     }
     else {
-        _PyCriticalSection_BeginSlow(c, m);
+        _PyCriticalSection_BeginSlow(tstate, c, m);
     }
 }
-#define PyCriticalSection_BeginMutex _PyCriticalSection_BeginMutex
 
 static inline void
-_PyCriticalSection_Begin(PyCriticalSection *c, PyObject *op)
+_PyCriticalSection_Begin(PyThreadState *tstate, PyCriticalSection *c, PyObject *op)
 {
-    _PyCriticalSection_BeginMutex(c, &op->ob_mutex);
+    _PyCriticalSection_BeginMutex(tstate, c, &op->ob_mutex);
 }
-#define PyCriticalSection_Begin _PyCriticalSection_Begin
 
 // Removes the top-most critical section from the thread's stack of critical
 // sections. If the new top-most critical section is inactive, then it is
 // resumed.
 static inline void
-_PyCriticalSection_Pop(PyCriticalSection *c)
+_PyCriticalSection_Pop(PyThreadState *tstate, PyCriticalSection *c)
 {
-    PyThreadState *tstate = _PyThreadState_GET();
     uintptr_t prev = c->_cs_prev;
     tstate->critical_section = prev;
 
@@ -132,7 +128,7 @@ _PyCriticalSection_Pop(PyCriticalSection *c)
 }
 
 static inline void
-_PyCriticalSection_End(PyCriticalSection *c)
+_PyCriticalSection_End(PyThreadState *tstate, PyCriticalSection *c)
 {
     // If the mutex is NULL, we used the fast path in
     // _PyCriticalSection_BeginSlow for locks already held in the top-most
@@ -141,18 +137,17 @@ _PyCriticalSection_End(PyCriticalSection *c)
         return;
     }
     PyMutex_Unlock(c->_cs_mutex);
-    _PyCriticalSection_Pop(c);
+    _PyCriticalSection_Pop(tstate, c);
 }
-#define PyCriticalSection_End _PyCriticalSection_End
 
 static inline void
-_PyCriticalSection2_BeginMutex(PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2)
+_PyCriticalSection2_BeginMutex(PyThreadState *tstate, PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2)
 {
     if (m1 == m2) {
         // If the two mutex arguments are the same, treat this as a critical
         // section with a single mutex.
         c->_cs_mutex2 = NULL;
-        _PyCriticalSection_BeginMutex(&c->_cs_base, m1);
+        _PyCriticalSection_BeginMutex(tstate, &c->_cs_base, m1);
         return;
     }
 
@@ -167,7 +162,6 @@ _PyCriticalSection2_BeginMutex(PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2)
 
     if (PyMutex_LockFast(m1)) {
         if (PyMutex_LockFast(m2)) {
-            PyThreadState *tstate = _PyThreadState_GET();
             c->_cs_base._cs_mutex = m1;
             c->_cs_mutex2 = m2;
             c->_cs_base._cs_prev = tstate->critical_section;
@@ -176,24 +170,22 @@ _PyCriticalSection2_BeginMutex(PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2)
             tstate->critical_section = p;
         }
         else {
-            _PyCriticalSection2_BeginSlow(c, m1, m2, 1);
+            _PyCriticalSection2_BeginSlow(tstate, c, m1, m2, 1);
         }
     }
     else {
-        _PyCriticalSection2_BeginSlow(c, m1, m2, 0);
+        _PyCriticalSection2_BeginSlow(tstate, c, m1, m2, 0);
     }
 }
-#define PyCriticalSection2_BeginMutex _PyCriticalSection2_BeginMutex
 
 static inline void
-_PyCriticalSection2_Begin(PyCriticalSection2 *c, PyObject *a, PyObject *b)
+_PyCriticalSection2_Begin(PyThreadState *tstate, PyCriticalSection2 *c, PyObject *a, PyObject *b)
 {
-    _PyCriticalSection2_BeginMutex(c, &a->ob_mutex, &b->ob_mutex);
+    _PyCriticalSection2_BeginMutex(tstate, c, &a->ob_mutex, &b->ob_mutex);
 }
-#define PyCriticalSection2_Begin _PyCriticalSection2_Begin
 
 static inline void
-_PyCriticalSection2_End(PyCriticalSection2 *c)
+_PyCriticalSection2_End(PyThreadState *tstate, PyCriticalSection2 *c)
 {
     // if mutex1 is NULL, we used the fast path in
     // _PyCriticalSection_BeginSlow for mutexes that are already held,
@@ -207,9 +199,8 @@ _PyCriticalSection2_End(PyCriticalSection2 *c)
         PyMutex_Unlock(c->_cs_mutex2);
     }
     PyMutex_Unlock(c->_cs_base._cs_mutex);
-    _PyCriticalSection_Pop(&c->_cs_base);
+    _PyCriticalSection_Pop(tstate, &c->_cs_base);
 }
-#define PyCriticalSection2_End _PyCriticalSection2_End
 
 static inline void
 _PyCriticalSection_AssertHeld(PyMutex *mutex)
@@ -251,6 +242,45 @@ _PyCriticalSection_AssertHeldObj(PyObject *op)
 
 #endif
 }
+
+#undef Py_BEGIN_CRITICAL_SECTION
+# define Py_BEGIN_CRITICAL_SECTION(op)                                  \
+    {                                                                   \
+        PyCriticalSection _py_cs;                                       \
+        PyThreadState *_cs_tstate = _PyThreadState_GET();               \
+        _PyCriticalSection_Begin(_cs_tstate, &_py_cs, _PyObject_CAST(op))
+
+#undef Py_BEGIN_CRITICAL_SECTION_MUTEX
+# define Py_BEGIN_CRITICAL_SECTION_MUTEX(mutex)                         \
+    {                                                                   \
+        PyCriticalSection _py_cs;                                       \
+        PyThreadState *_cs_tstate = _PyThreadState_GET();               \
+        _PyCriticalSection_BeginMutex(_cs_tstate, &_py_cs, mutex)
+
+#undef Py_END_CRITICAL_SECTION
+# define Py_END_CRITICAL_SECTION()                                      \
+        _PyCriticalSection_End(_cs_tstate, &_py_cs);                    \
+    }
+
+#undef Py_BEGIN_CRITICAL_SECTION2
+# define Py_BEGIN_CRITICAL_SECTION2(a, b)                               \
+    {                                                                   \
+        PyCriticalSection2 _py_cs2;                                     \
+        PyThreadState *_cs_tstate = _PyThreadState_GET();               \
+        _PyCriticalSection2_Begin(_cs_tstate, &_py_cs2, _PyObject_CAST(a), _PyObject_CAST(b))
+
+#undef Py_BEGIN_CRITICAL_SECTION2_MUTEX
+# define Py_BEGIN_CRITICAL_SECTION2_MUTEX(m1, m2)                       \
+    {                                                                   \
+        PyCriticalSection2 _py_cs2;                                     \
+        PyThreadState *_cs_tstate = _PyThreadState_GET();               \
+        _PyCriticalSection2_BeginMutex(_cs_tstate, &_py_cs2, m1, m2)
+
+#undef Py_END_CRITICAL_SECTION2
+# define Py_END_CRITICAL_SECTION2()                                     \
+        _PyCriticalSection2_End(_cs_tstate, &_py_cs2);                  \
+    }
+
 #endif /* Py_GIL_DISABLED */
 
 #ifdef __cplusplus
diff --git a/Lib/data.bin b/Lib/data.bin
deleted file mode 100644
index 1d1fab72c4aaa4..00000000000000
Binary files a/Lib/data.bin and /dev/null differ
diff --git a/Lib/test/test_free_threading/test_csv.py b/Lib/test/test_free_threading/test_csv.py
new file mode 100644
index 00000000000000..beb4510a1281b8
--- /dev/null
+++ b/Lib/test/test_free_threading/test_csv.py
@@ -0,0 +1,50 @@
+import csv
+import io
+import unittest
+
+from test.support import threading_helper
+from test.support.threading_helper import run_concurrently
+
+
+NTHREADS = 10
+
+
+@threading_helper.requires_working_threading()
+class TestCSV(unittest.TestCase):
+    def test_concurrent_reader_next(self):
+        input_rows = [f"{i},{i},{i}" for i in range(50)]
+        input_stream = io.StringIO("\n".join(input_rows))
+        reader = csv.reader(input_stream)
+        output_rows = []
+
+        def read_row():
+            for row in reader:
+                self.assertEqual(len(row), 3)
+                output_rows.append(",".join(row))
+
+        run_concurrently(worker_func=read_row, nthreads=NTHREADS)
+        self.assertSetEqual(set(input_rows), set(output_rows))
+
+    def test_concurrent_writer_writerow(self):
+        output_stream = io.StringIO()
+        writer = csv.writer(output_stream)
+        row_per_thread = 10
+        expected_rows = []
+
+        def write_row():
+            for i in range(row_per_thread):
+                writer.writerow([i, i, i])
+                expected_rows.append(f"{i},{i},{i}")
+
+        run_concurrently(worker_func=write_row, nthreads=NTHREADS)
+
+        # Rewind to the start of the stream and parse the rows
+        output_stream.seek(0)
+        output_rows = [line.strip() for line in output_stream.readlines()]
+
+        self.assertEqual(len(output_rows), NTHREADS * row_per_thread)
+        self.assertListEqual(sorted(output_rows), sorted(expected_rows))
+
+
+if __name__ == "__main__":
+    unittest.main()
diff --git a/Lib/test/test_opcache.py b/Lib/test/test_opcache.py
index c7eea75117de8c..4113b79ef5c80b 100644
--- a/Lib/test/test_opcache.py
+++ b/Lib/test/test_opcache.py
@@ -590,7 +590,7 @@ def make_deferred_ref_count_obj():
 class TestRacesDoNotCrash(TestBase):
     # Careful with these. Bigger numbers have a higher chance of catching bugs,
     # but you can also burn through a *ton* of type/dict/function versions:
-    ITEMS = 1000
+    ITEMS = 1400
     LOOPS = 4
     WRITERS = 2
 
diff --git a/Lib/test/test_perf_profiler.py b/Lib/test/test_perf_profiler.py
index e6852c93e69830..66348619073909 100644
--- a/Lib/test/test_perf_profiler.py
+++ b/Lib/test/test_perf_profiler.py
@@ -160,6 +160,16 @@ def baz():
         self.assertIn(f"py::bar_fork:{script}", child_perf_file_contents)
         self.assertIn(f"py::baz_fork:{script}", child_perf_file_contents)
 
+        # The parent's map should not contain the child's symbols.
+        self.assertNotIn(f"py::foo_fork:{script}", perf_file_contents)
+        self.assertNotIn(f"py::bar_fork:{script}", perf_file_contents)
+        self.assertNotIn(f"py::baz_fork:{script}", perf_file_contents)
+
+        # The child's map should not contain the parent's symbols.
+        self.assertNotIn(f"py::foo:{script}", child_perf_file_contents)
+        self.assertNotIn(f"py::bar:{script}", child_perf_file_contents)
+        self.assertNotIn(f"py::baz:{script}", child_perf_file_contents)
+
     @unittest.skipIf(support.check_bolt_optimized(), "fails on BOLT instrumented binaries")
     def test_sys_api(self):
         for define_eval_hook in (False, True):
diff --git a/Makefile.pre.in b/Makefile.pre.in
index 13108b1baf976a..1c793b6db7852c 100644
--- a/Makefile.pre.in
+++ b/Makefile.pre.in
@@ -2573,6 +2573,7 @@ LIBSUBDIRS=	asyncio \
 		profile \
 		profiling profiling/sampling profiling/tracing \
 		profiling/sampling/_assets \
+		profiling/sampling/live_collector \
 		profiling/sampling/_vendor/d3/7.8.5 \
 		profiling/sampling/_vendor/d3-flame-graph/4.1.3 \
 		pydoc_data \
diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2025-11-10-00-14-20.gh-issue-116738.IxliC_.rst b/Misc/NEWS.d/next/Core_and_Builtins/2025-11-10-00-14-20.gh-issue-116738.IxliC_.rst
new file mode 100644
index 00000000000000..8b08bccafd73eb
--- /dev/null
+++ b/Misc/NEWS.d/next/Core_and_Builtins/2025-11-10-00-14-20.gh-issue-116738.IxliC_.rst
@@ -0,0 +1,2 @@
+Make csv module thread-safe on the :term:`free threaded <free threading>`
+build.
diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2025-11-15-14-04-35.gh-issue-141589.VfdMDD.rst b/Misc/NEWS.d/next/Core_and_Builtins/2025-11-15-14-04-35.gh-issue-141589.VfdMDD.rst
new file mode 100644
index 00000000000000..5eb0e0c8b89f7a
--- /dev/null
+++ b/Misc/NEWS.d/next/Core_and_Builtins/2025-11-15-14-04-35.gh-issue-141589.VfdMDD.rst
@@ -0,0 +1,3 @@
+Change ``backoff counter`` to use prime numbers instead of powers of 2.
+Use only 3 bits for ``counter`` and 13 bits for ``value``.
+This allows to support values up to 8191. Patch by Mikhail Efimov.
diff --git a/Modules/_csv.c b/Modules/_csv.c
index 87be7a8f1fb136..1f41976e95fdb1 100644
--- a/Modules/_csv.c
+++ b/Modules/_csv.c
@@ -918,7 +918,7 @@ parse_reset(ReaderObj *self)
 }
 
 static PyObject *
-Reader_iternext(PyObject *op)
+Reader_iternext_lock_held(PyObject *op)
 {
     ReaderObj *self = _ReaderObj_CAST(op);
 
@@ -985,6 +985,16 @@ Reader_iternext(PyObject *op)
     return fields;
 }
 
+static PyObject *
+Reader_iternext(PyObject *op)
+{
+    PyObject *result;
+    Py_BEGIN_CRITICAL_SECTION(op);
+    result = Reader_iternext_lock_held(op);
+    Py_END_CRITICAL_SECTION();
+    return result;
+}
+
 static void
 Reader_dealloc(PyObject *op)
 {
@@ -1303,15 +1313,8 @@ join_append_lineterminator(WriterObj *self)
     return 1;
 }
 
-PyDoc_STRVAR(csv_writerow_doc,
-"writerow($self, row, /)\n"
-"--\n\n"
-"Construct and write a CSV record from an iterable of fields.\n"
-"\n"
-"Non-string elements will be converted to string.");
-
 static PyObject *
-csv_writerow(PyObject *op, PyObject *seq)
+csv_writerow_lock_held(PyObject *op, PyObject *seq)
 {
     WriterObj *self = _WriterObj_CAST(op);
     DialectObj *dialect = self->dialect;
@@ -1414,6 +1417,23 @@ csv_writerow(PyObject *op, PyObject *seq)
     return result;
 }
 
+PyDoc_STRVAR(csv_writerow_doc,
+"writerow($self, row, /)\n"
+"--\n\n"
+"Construct and write a CSV record from an iterable of fields.\n"
+"\n"
+"Non-string elements will be converted to string.");
+
+static PyObject *
+csv_writerow(PyObject *op, PyObject *seq)
+{
+    PyObject *result;
+    Py_BEGIN_CRITICAL_SECTION(op);
+    result = csv_writerow_lock_held(op, seq);
+    Py_END_CRITICAL_SECTION();
+    return result;
+}
+
 PyDoc_STRVAR(csv_writerows_doc,
 "writerows($self, rows, /)\n"
 "--\n\n"
diff --git a/Python/critical_section.c b/Python/critical_section.c
index 218b580e95176d..2c2152f5de4716 100644
--- a/Python/critical_section.c
+++ b/Python/critical_section.c
@@ -17,10 +17,9 @@ untag_critical_section(uintptr_t tag)
 #endif
 
 void
-_PyCriticalSection_BeginSlow(PyCriticalSection *c, PyMutex *m)
+_PyCriticalSection_BeginSlow(PyThreadState *tstate, PyCriticalSection *c, PyMutex *m)
 {
 #ifdef Py_GIL_DISABLED
-    PyThreadState *tstate = _PyThreadState_GET();
     // As an optimisation for locking the same object recursively, skip
     // locking if the mutex is currently locked by the top-most critical
     // section.
@@ -53,11 +52,10 @@ _PyCriticalSection_BeginSlow(PyCriticalSection *c, PyMutex *m)
 }
 
 void
-_PyCriticalSection2_BeginSlow(PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2,
+_PyCriticalSection2_BeginSlow(PyThreadState *tstate, PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2,
                               int is_m1_locked)
 {
 #ifdef Py_GIL_DISABLED
-    PyThreadState *tstate = _PyThreadState_GET();
     c->_cs_base._cs_mutex = NULL;
     c->_cs_mutex2 = NULL;
     c->_cs_base._cs_prev = tstate->critical_section;
@@ -139,7 +137,7 @@ void
 PyCriticalSection_Begin(PyCriticalSection *c, PyObject *op)
 {
 #ifdef Py_GIL_DISABLED
-    _PyCriticalSection_Begin(c, op);
+    _PyCriticalSection_Begin(_PyThreadState_GET(), c, op);
 #endif
 }
 
@@ -148,7 +146,7 @@ void
 PyCriticalSection_BeginMutex(PyCriticalSection *c, PyMutex *m)
 {
 #ifdef Py_GIL_DISABLED
-    _PyCriticalSection_BeginMutex(c, m);
+    _PyCriticalSection_BeginMutex(_PyThreadState_GET(), c, m);
 #endif
 }
 
@@ -157,7 +155,7 @@ void
 PyCriticalSection_End(PyCriticalSection *c)
 {
 #ifdef Py_GIL_DISABLED
-    _PyCriticalSection_End(c);
+    _PyCriticalSection_End(_PyThreadState_GET(), c);
 #endif
 }
 
@@ -166,7 +164,7 @@ void
 PyCriticalSection2_Begin(PyCriticalSection2 *c, PyObject *a, PyObject *b)
 {
 #ifdef Py_GIL_DISABLED
-    _PyCriticalSection2_Begin(c, a, b);
+    _PyCriticalSection2_Begin(_PyThreadState_GET(), c, a, b);
 #endif
 }
 
@@ -175,7 +173,7 @@ void
 PyCriticalSection2_BeginMutex(PyCriticalSection2 *c, PyMutex *m1, PyMutex *m2)
 {
 #ifdef Py_GIL_DISABLED
-    _PyCriticalSection2_BeginMutex(c, m1, m2);
+    _PyCriticalSection2_BeginMutex(_PyThreadState_GET(), c, m1, m2);
 #endif
 }
 
@@ -184,6 +182,6 @@ void
 PyCriticalSection2_End(PyCriticalSection2 *c)
 {
 #ifdef Py_GIL_DISABLED
-    _PyCriticalSection2_End(c);
+    _PyCriticalSection2_End(_PyThreadState_GET(), c);
 #endif
 }