gh-112075: Make instance attributes stored in inline "dict" thread safe (#114742)

Make instance attributes stored in inline "dict" thread safe on free-threaded builds

4 files changed, 35 insertions, 15 deletions

diff --git a/Include/cpython/object.h b/Include/cpython/object.h
index 2797051..a8f5782 100644
--- a/Include/cpython/object.h
+++ b/Include/cpython/object.h
@@ -493,6 +493,7 @@ do { \
 PyAPI_FUNC(void *) PyObject_GetItemData(PyObject *obj);
 
 PyAPI_FUNC(int) PyObject_VisitManagedDict(PyObject *obj, visitproc visit, void *arg);
+PyAPI_FUNC(void) _PyObject_SetManagedDict(PyObject *obj, PyObject *new_dict);
 PyAPI_FUNC(void) PyObject_ClearManagedDict(PyObject *obj);
 
 #define TYPE_MAX_WATCHERS 8
diff --git a/Include/internal/pycore_dict.h b/Include/internal/pycore_dict.h
index fba0dfc..f33026d 100644
--- a/Include/internal/pycore_dict.h
+++ b/Include/internal/pycore_dict.h
@@ -1,4 +1,3 @@
-
 #ifndef Py_INTERNAL_DICT_H
 #define Py_INTERNAL_DICT_H
 #ifdef __cplusplus
@@ -9,9 +8,10 @@ extern "C" {
 #  error "this header requires Py_BUILD_CORE define"
 #endif
 
-#include "pycore_freelist.h"      // _PyFreeListState
-#include "pycore_identifier.h"    // _Py_Identifier
-#include "pycore_object.h"        // PyManagedDictPointer
+#include "pycore_freelist.h"             // _PyFreeListState
+#include "pycore_identifier.h"           // _Py_Identifier
+#include "pycore_object.h"               // PyManagedDictPointer
+#include "pycore_pyatomic_ft_wrappers.h" // FT_ATOMIC_LOAD_SSIZE_ACQUIRE
 
 // Unsafe flavor of PyDict_GetItemWithError(): no error checking
 extern PyObject* _PyDict_GetItemWithError(PyObject *dp, PyObject *key);
@@ -249,7 +249,7 @@ _PyDict_NotifyEvent(PyInterpreterState *interp,
     return DICT_NEXT_VERSION(interp) | (mp->ma_version_tag & DICT_WATCHER_AND_MODIFICATION_MASK);
 }
 
-extern PyDictObject *_PyObject_MakeDictFromInstanceAttributes(PyObject *obj);
+extern PyDictObject *_PyObject_MaterializeManagedDict(PyObject *obj);
 
 PyAPI_FUNC(PyObject *)_PyDict_FromItems(
         PyObject *const *keys, Py_ssize_t keys_offset,
@@ -277,7 +277,6 @@ _PyDictValues_AddToInsertionOrder(PyDictValues *values, Py_ssize_t ix)
 static inline size_t
 shared_keys_usable_size(PyDictKeysObject *keys)
 {
-#ifdef Py_GIL_DISABLED
     // dk_usable will decrease for each instance that is created and each
     // value that is added.  dk_nentries will increase for each value that
     // is added.  We want to always return the right value or larger.
@@ -285,11 +284,9 @@ shared_keys_usable_size(PyDictKeysObject *keys)
     // second, and conversely here we read dk_usable first and dk_entries
     // second (to avoid the case where we read entries before the increment
     // and read usable after the decrement)
-    return (size_t)(_Py_atomic_load_ssize_acquire(&keys->dk_usable) +
-                    _Py_atomic_load_ssize_acquire(&keys->dk_nentries));
-#else
-    return (size_t)keys->dk_nentries + (size_t)keys->dk_usable;
-#endif
+    Py_ssize_t dk_usable = FT_ATOMIC_LOAD_SSIZE_ACQUIRE(keys->dk_usable);
+    Py_ssize_t dk_nentries = FT_ATOMIC_LOAD_SSIZE_ACQUIRE(keys->dk_nentries);
+    return dk_nentries + dk_usable;
 }
 
 static inline size_t
diff --git a/Include/internal/pycore_object.h b/Include/internal/pycore_object.h
index 512f7a3..88b052f 100644
--- a/Include/internal/pycore_object.h
+++ b/Include/internal/pycore_object.h
@@ -12,6 +12,7 @@ extern "C" {
 #include "pycore_gc.h"            // _PyObject_GC_IS_TRACKED()
 #include "pycore_emscripten_trampoline.h" // _PyCFunction_TrampolineCall()
 #include "pycore_interp.h"        // PyInterpreterState.gc
+#include "pycore_pyatomic_ft_wrappers.h"  // FT_ATOMIC_STORE_PTR_RELAXED
 #include "pycore_pystate.h"       // _PyInterpreterState_GET()
 
 /* Check if an object is consistent. For example, ensure that the reference
@@ -659,10 +660,10 @@ extern PyObject* _PyType_GetDocFromInternalDoc(const char *, const char *);
 extern PyObject* _PyType_GetTextSignatureFromInternalDoc(const char *, const char *, int);
 
 void _PyObject_InitInlineValues(PyObject *obj, PyTypeObject *tp);
-extern int _PyObject_StoreInstanceAttribute(PyObject *obj, PyDictValues *values,
-                                          PyObject *name, PyObject *value);
-PyObject * _PyObject_GetInstanceAttribute(PyObject *obj, PyDictValues *values,
-                                        PyObject *name);
+extern int _PyObject_StoreInstanceAttribute(PyObject *obj,
+                                            PyObject *name, PyObject *value);
+extern bool _PyObject_TryGetInstanceAttribute(PyObject *obj, PyObject *name,
+                                              PyObject **attr);
 
 #ifdef Py_GIL_DISABLED
 #  define MANAGED_DICT_OFFSET    (((Py_ssize_t)sizeof(PyObject *))*-1)
@@ -683,6 +684,13 @@ _PyObject_ManagedDictPointer(PyObject *obj)
     return (PyManagedDictPointer *)((char *)obj + MANAGED_DICT_OFFSET);
 }
 
+static inline PyDictObject *
+_PyObject_GetManagedDict(PyObject *obj)
+{
+    PyManagedDictPointer *dorv = _PyObject_ManagedDictPointer(obj);
+    return (PyDictObject *)FT_ATOMIC_LOAD_PTR_RELAXED(dorv->dict);
+}
+
 static inline PyDictValues *
 _PyObject_InlineValues(PyObject *obj)
 {
diff --git a/Include/internal/pycore_pyatomic_ft_wrappers.h b/Include/internal/pycore_pyatomic_ft_wrappers.h
index fed5d6e..bbfc462 100644
--- a/Include/internal/pycore_pyatomic_ft_wrappers.h
+++ b/Include/internal/pycore_pyatomic_ft_wrappers.h
@@ -21,7 +21,10 @@ extern "C" {
 
 #ifdef Py_GIL_DISABLED
 #define FT_ATOMIC_LOAD_PTR(value) _Py_atomic_load_ptr(&value)
+#define FT_ATOMIC_STORE_PTR(value, new_value) _Py_atomic_store_ptr(&value, new_value)
 #define FT_ATOMIC_LOAD_SSIZE(value) _Py_atomic_load_ssize(&value)
+#define FT_ATOMIC_LOAD_SSIZE_ACQUIRE(value) \
+    _Py_atomic_load_ssize_acquire(&value)
 #define FT_ATOMIC_LOAD_SSIZE_RELAXED(value) \
     _Py_atomic_load_ssize_relaxed(&value)
 #define FT_ATOMIC_STORE_PTR(value, new_value) \
@@ -30,6 +33,12 @@ extern "C" {
     _Py_atomic_load_ptr_acquire(&value)
 #define FT_ATOMIC_LOAD_UINTPTR_ACQUIRE(value) \
     _Py_atomic_load_uintptr_acquire(&value)
+#define FT_ATOMIC_LOAD_PTR_RELAXED(value) \
+    _Py_atomic_load_ptr_relaxed(&value)
+#define FT_ATOMIC_LOAD_UINT8(value) \
+    _Py_atomic_load_uint8(&value)
+#define FT_ATOMIC_STORE_UINT8(value, new_value) \
+    _Py_atomic_store_uint8(&value, new_value)
 #define FT_ATOMIC_STORE_PTR_RELAXED(value, new_value) \
     _Py_atomic_store_ptr_relaxed(&value, new_value)
 #define FT_ATOMIC_STORE_PTR_RELEASE(value, new_value) \
@@ -43,11 +52,16 @@ extern "C" {
 
 #else
 #define FT_ATOMIC_LOAD_PTR(value) value
+#define FT_ATOMIC_STORE_PTR(value, new_value) value = new_value
 #define FT_ATOMIC_LOAD_SSIZE(value) value
+#define FT_ATOMIC_LOAD_SSIZE_ACQUIRE(value) value
 #define FT_ATOMIC_LOAD_SSIZE_RELAXED(value) value
 #define FT_ATOMIC_STORE_PTR(value, new_value) value = new_value
 #define FT_ATOMIC_LOAD_PTR_ACQUIRE(value) value
 #define FT_ATOMIC_LOAD_UINTPTR_ACQUIRE(value) value
+#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_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