gh-131586: Avoid refcount contention in some "special" calls (#131588)

In the free threaded build, the `_PyObject_LookupSpecial()` call can lead to
reference count contention on the returned function object becuase it
doesn't use stackrefs. Refactor some of the callers to use
`_PyObject_MaybeCallSpecialNoArgs`, which uses stackrefs internally.

This fixes the scaling bottleneck in the "lookup_special" microbenchmark
in `ftscalingbench.py`. However, the are still some uses of
`_PyObject_LookupSpecial()` that need to be addressed in future PRs.

1 files changed, 20 insertions, 8 deletions

diff --git a/Objects/object.c b/Objects/object.c
index ecc5a86..974cf60 100644
--- a/Objects/object.c
+++ b/Objects/object.c
@@ -1680,14 +1680,20 @@ _PyObject_GenericGetAttrWithDict(PyObject *obj, PyObject *name,
                      Py_TYPE(name)->tp_name);
         return NULL;
     }
-    Py_INCREF(name);
 
     if (!_PyType_IsReady(tp)) {
         if (PyType_Ready(tp) < 0)
-            goto done;
+            return NULL;
     }
 
-    descr = _PyType_LookupRef(tp, name);
+    Py_INCREF(name);
+
+    PyThreadState *tstate = _PyThreadState_GET();
+    _PyCStackRef cref;
+    _PyThreadState_PushCStackRef(tstate, &cref);
+
+    _PyType_LookupStackRefAndVersion(tp, name, &cref.ref);
+    descr = PyStackRef_AsPyObjectBorrow(cref.ref);
 
     f = NULL;
     if (descr != NULL) {
@@ -1758,8 +1764,8 @@ _PyObject_GenericGetAttrWithDict(PyObject *obj, PyObject *name,
     }
 
     if (descr != NULL) {
-        res = descr;
-        descr = NULL;
+        res = PyStackRef_AsPyObjectSteal(cref.ref);
+        cref.ref = PyStackRef_NULL;
         goto done;
     }
 
@@ -1771,7 +1777,7 @@ _PyObject_GenericGetAttrWithDict(PyObject *obj, PyObject *name,
         _PyObject_SetAttributeErrorContext(obj, name);
     }
   done:
-    Py_XDECREF(descr);
+    _PyThreadState_PopCStackRef(tstate, &cref);
     Py_DECREF(name);
     return res;
 }
@@ -1805,7 +1811,13 @@ _PyObject_GenericSetAttrWithDict(PyObject *obj, PyObject *name,
 
     Py_INCREF(name);
     Py_INCREF(tp);
-    descr = _PyType_LookupRef(tp, name);
+
+    PyThreadState *tstate = _PyThreadState_GET();
+    _PyCStackRef cref;
+    _PyThreadState_PushCStackRef(tstate, &cref);
+
+    _PyType_LookupStackRefAndVersion(tp, name, &cref.ref);
+    descr = PyStackRef_AsPyObjectBorrow(cref.ref);
 
     if (descr != NULL) {
         f = Py_TYPE(descr)->tp_descr_set;
@@ -1872,7 +1884,7 @@ _PyObject_GenericSetAttrWithDict(PyObject *obj, PyObject *name,
         _PyObject_SetAttributeErrorContext(obj, name);
     }
   done:
-    Py_XDECREF(descr);
+    _PyThreadState_PopCStackRef(tstate, &cref);
     Py_DECREF(tp);
     Py_DECREF(name);
     return res;