Add thread safety around inline dict matieralization

Author: DinoV

Include/internal/pycore_dict.h

@@ -9,9 +9,10 @@ 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);
@@ -248,7 +249,7 @@ _PyDict_NotifyEvent(PyInterpreterState *interp,
     return DICT_NEXT_VERSION(interp) | (mp->ma_version_tag & DICT_WATCHER_AND_MODIFICATION_MASK);
 }
 
-extern PyObject *_PyObject_MakeDictFromInstanceAttributes(PyObject *obj);
+extern PyObject *_PyObject_MaterializeManagedDict(PyObject *obj);
 
 PyAPI_FUNC(PyObject *)_PyDict_FromItems(
         PyObject *const *keys, Py_ssize_t keys_offset,
@@ -276,19 +277,16 @@ _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.
     // We therefore increase dk_nentries first and we decrease dk_usable
     // 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

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
@@ -621,10 +622,10 @@ 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_TryStoreInstanceAttribute(PyObject *obj,
+                                               PyObject *name, PyObject *value);
+extern int _PyObject_TryGetInstanceAttribute(PyObject *obj, PyObject *name,
+                                             PyObject **attr);
 
 #ifdef Py_GIL_DISABLED
 #  define MANAGED_DICT_OFFSET    (((Py_ssize_t)sizeof(PyObject *))*-1)
@@ -645,6 +646,20 @@ _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 void
+_PyObject_SetManagedDict(PyObject *obj, PyObject *dict)
+{
+    PyManagedDictPointer *ptr = _PyObject_ManagedDictPointer(obj);
+    FT_ATOMIC_STORE_PTR_RELEASE(ptr->dict, dict);
+}
+
 static inline PyDictValues *
 _PyObject_InlineValues(PyObject *obj)
 {

Include/internal/pycore_pyatomic_ft_wrappers.h

@@ -21,8 +21,14 @@ extern "C" {
 
 #ifdef Py_GIL_DISABLED
 #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_LOAD_PTR_RELAXED(value) \
+    _Py_atomic_load_ptr_relaxed(&value)
+#define FT_ATOMIC_LOAD_UINT8_RELAXED(value) \
+    _Py_atomic_load_uint8_relaxed(&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) \
@@ -31,7 +37,10 @@ extern "C" {
     _Py_atomic_store_ssize_relaxed(&value, new_value)
 #else
 #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_LOAD_PTR_RELAXED(value) value
+#define FT_ATOMIC_LOAD_UINT8_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_SSIZE_RELAXED(value, new_value) value = new_value