gh-120321: Make gi_frame_state transitions atomic in FT build (gh-142599)

This makes generator frame state transitions atomic in the free
threading build, which avoids segfaults when trying to execute
a generator from multiple threads concurrently.

There are still a few operations that aren't thread-safe and may crash
if performed concurrently on the same generator/coroutine:

 * Accessing gi_yieldfrom/cr_await/ag_await
 * Accessing gi_frame/cr_frame/ag_frame
 * Async generator operations

Author: colesbury

Include/cpython/pyatomic.h

@@ -523,6 +523,9 @@ _Py_atomic_store_uintptr_release(uintptr_t *obj, uintptr_t value);
 static inline void
 _Py_atomic_store_ssize_release(Py_ssize_t *obj, Py_ssize_t value);
 
+static inline void
+_Py_atomic_store_int8_release(int8_t *obj, int8_t value);
+
 static inline void
 _Py_atomic_store_int_release(int *obj, int value);
 

Include/cpython/pyatomic_gcc.h

@@ -572,6 +572,10 @@ static inline void
 _Py_atomic_store_int_release(int *obj, int value)
 { __atomic_store_n(obj, value, __ATOMIC_RELEASE); }
 
+static inline void
+_Py_atomic_store_int8_release(int8_t *obj, int8_t value)
+{ __atomic_store_n(obj, value, __ATOMIC_RELEASE); }
+
 static inline void
 _Py_atomic_store_ssize_release(Py_ssize_t *obj, Py_ssize_t value)
 { __atomic_store_n(obj, value, __ATOMIC_RELEASE); }

Include/cpython/pyatomic_msc.h

@@ -1066,6 +1066,19 @@ _Py_atomic_store_int_release(int *obj, int value)
 #endif
 }
 
+static inline void
+_Py_atomic_store_int8_release(int8_t *obj, int8_t value)
+{
+#if defined(_M_X64) || defined(_M_IX86)
+    *(int8_t volatile *)obj = value;
+#elif defined(_M_ARM64)
+    _Py_atomic_ASSERT_ARG_TYPE(unsigned __int8);
+    __stlr8((unsigned __int8 volatile *)obj, (unsigned __int8)value);
+#else
+#  error "no implementation of _Py_atomic_store_int8_release"
+#endif
+}
+
 static inline void
 _Py_atomic_store_ssize_release(Py_ssize_t *obj, Py_ssize_t value)
 {

Include/cpython/pyatomic_std.h

@@ -1023,6 +1023,14 @@ _Py_atomic_store_int_release(int *obj, int value)
                           memory_order_release);
 }
 
+static inline void
+_Py_atomic_store_int8_release(int8_t *obj, int8_t value)
+{
+    _Py_USING_STD;
+    atomic_store_explicit((_Atomic(int8_t)*)obj, value,
+                          memory_order_release);
+}
+
 static inline void
 _Py_atomic_store_uint_release(unsigned int *obj, unsigned int value)
 {

Include/internal/pycore_pyatomic_ft_wrappers.h

@@ -41,6 +41,8 @@ extern "C" {
     _Py_atomic_load_uint8(&value)
 #define FT_ATOMIC_STORE_UINT8(value, new_value) \
     _Py_atomic_store_uint8(&value, new_value)
+#define FT_ATOMIC_LOAD_INT8_RELAXED(value) \
+    _Py_atomic_load_int8_relaxed(&value)
 #define FT_ATOMIC_LOAD_UINT8_RELAXED(value) \
     _Py_atomic_load_uint8_relaxed(&value)
 #define FT_ATOMIC_LOAD_UINT16_RELAXED(value) \
@@ -55,6 +57,10 @@ extern "C" {
     _Py_atomic_store_ptr_release(&value, new_value)
 #define FT_ATOMIC_STORE_UINTPTR_RELEASE(value, new_value) \
     _Py_atomic_store_uintptr_release(&value, new_value)
+#define FT_ATOMIC_STORE_INT8_RELAXED(value, new_value) \
+    _Py_atomic_store_int8_relaxed(&value, new_value)
+#define FT_ATOMIC_STORE_INT8_RELEASE(value, new_value) \
+    _Py_atomic_store_int8_release(&value, new_value)
 #define FT_ATOMIC_STORE_SSIZE_RELAXED(value, new_value) \
     _Py_atomic_store_ssize_relaxed(&value, new_value)
 #define FT_ATOMIC_STORE_SSIZE_RELEASE(value, new_value) \
@@ -134,13 +140,16 @@ extern "C" {
 #define FT_ATOMIC_LOAD_PTR_RELAXED(value) value
 #define FT_ATOMIC_LOAD_UINT8(value) value
 #define FT_ATOMIC_STORE_UINT8(value, new_value) value = new_value
+#define FT_ATOMIC_LOAD_INT8_RELAXED(value) value
 #define FT_ATOMIC_LOAD_UINT8_RELAXED(value) value
 #define FT_ATOMIC_LOAD_UINT16_RELAXED(value) value
 #define FT_ATOMIC_LOAD_UINT32_RELAXED(value) value
 #define FT_ATOMIC_LOAD_ULONG_RELAXED(value) value
 #define FT_ATOMIC_STORE_PTR_RELAXED(value, new_value) value = new_value
 #define FT_ATOMIC_STORE_PTR_RELEASE(value, new_value) value = new_value
 #define FT_ATOMIC_STORE_UINTPTR_RELEASE(value, new_value) value = new_value
+#define FT_ATOMIC_STORE_INT8_RELAXED(value, new_value) value = new_value
+#define FT_ATOMIC_STORE_INT8_RELEASE(value, new_value) value = new_value
 #define FT_ATOMIC_STORE_SSIZE_RELAXED(value, new_value) value = new_value
 #define FT_ATOMIC_STORE_SSIZE_RELEASE(value, new_value) value = new_value
 #define FT_ATOMIC_STORE_UINT8_RELAXED(value, new_value) value = new_value

Include/internal/pycore_tstate.h

@@ -82,6 +82,13 @@ typedef struct _PyThreadStateImpl {
     PyObject *asyncio_running_loop; // Strong reference
     PyObject *asyncio_running_task; // Strong reference
 
+    // Distinguishes between yield and return from PyEval_EvalFrame().
+    // See gen_send_ex2() in Objects/genobject.c
+    enum {
+        GENERATOR_RETURN = 0,
+        GENERATOR_YIELD = 1,
+    } generator_return_kind;
+
     /* Head of circular linked-list of all tasks which are instances of `asyncio.Task`
        or subclasses of it used in `asyncio.all_tasks`.
     */