gh-142534: Avoid TSan warnings in dictobject.c

There are places we use "relaxed" loads where C11 requires "consume" or
stronger. Unfortunately, compilers don't really implement "consume" so
fake it for our use in a way that avoids upsetting TSan.

Author: colesbury

Include/cpython/pyatomic.h

@@ -591,6 +591,17 @@ static inline void _Py_atomic_fence_release(void);
 
 // --- aliases ---------------------------------------------------------------
 
+// Compilers don't really support "consume" semantics, so we fake it. Use
+// "acquire" with TSan to support false positives. Use "relaxed" otherwise,
+// because CPUs on all platforms we support respect address dependencies without
+// extra barriers.
+// See 2.6.7 in https://www.open-std.org/jtc1/sc22/wg21/docs/papers/2020/p2055r0.pdf
+#if defined(_Py_THREAD_SANITIZER)
+# define _Py_atomic_load_ptr_consume _Py_atomic_load_ptr_acquire
+#else
+# define _Py_atomic_load_ptr_consume _Py_atomic_load_ptr_relaxed
+#endif
+
 #if SIZEOF_LONG == 8
 # define _Py_atomic_load_ulong(p) \
     _Py_atomic_load_uint64((uint64_t *)p)

Include/cpython/pyatomic_gcc.h

@@ -549,7 +549,6 @@ static inline void
 _Py_atomic_store_llong_relaxed(long long *obj, long long value)
 { __atomic_store_n(obj, value, __ATOMIC_RELAXED); }
 
-
 // --- _Py_atomic_load_ptr_acquire / _Py_atomic_store_ptr_release ------------
 
 static inline void *

Include/internal/pycore_pyatomic_ft_wrappers.h

@@ -31,6 +31,8 @@ extern "C" {
     _Py_atomic_store_ptr(&value, new_value)
 #define FT_ATOMIC_LOAD_PTR_ACQUIRE(value) \
     _Py_atomic_load_ptr_acquire(&value)
+#define FT_ATOMIC_LOAD_PTR_CONSUME(value) \
+    _Py_atomic_load_ptr_consume(&value)
 #define FT_ATOMIC_LOAD_UINTPTR_ACQUIRE(value) \
     _Py_atomic_load_uintptr_acquire(&value)
 #define FT_ATOMIC_LOAD_PTR_RELAXED(value) \

Objects/dictobject.c

@@ -1078,7 +1078,7 @@ compare_unicode_unicode(PyDictObject *mp, PyDictKeysObject *dk,
                         void *ep0, Py_ssize_t ix, PyObject *key, Py_hash_t hash)
 {
     PyDictUnicodeEntry *ep = &((PyDictUnicodeEntry *)ep0)[ix];
-    PyObject *ep_key = FT_ATOMIC_LOAD_PTR_RELAXED(ep->me_key);
+    PyObject *ep_key = FT_ATOMIC_LOAD_PTR_CONSUME(ep->me_key);
     assert(ep_key != NULL);
     assert(PyUnicode_CheckExact(ep_key));
     if (ep_key == key ||
@@ -1371,7 +1371,7 @@ compare_unicode_generic_threadsafe(PyDictObject *mp, PyDictKeysObject *dk,
                                    void *ep0, Py_ssize_t ix, PyObject *key, Py_hash_t hash)
 {
     PyDictUnicodeEntry *ep = &((PyDictUnicodeEntry *)ep0)[ix];
-    PyObject *startkey = _Py_atomic_load_ptr_relaxed(&ep->me_key);
+    PyObject *startkey = _Py_atomic_load_ptr_consume(&ep->me_key);
     assert(startkey == NULL || PyUnicode_CheckExact(ep->me_key));
     assert(!PyUnicode_CheckExact(key));
 
@@ -1414,7 +1414,7 @@ compare_unicode_unicode_threadsafe(PyDictObject *mp, PyDictKeysObject *dk,
                                    void *ep0, Py_ssize_t ix, PyObject *key, Py_hash_t hash)
 {
     PyDictUnicodeEntry *ep = &((PyDictUnicodeEntry *)ep0)[ix];
-    PyObject *startkey = _Py_atomic_load_ptr_relaxed(&ep->me_key);
+    PyObject *startkey = _Py_atomic_load_ptr_consume(&ep->me_key);
     if (startkey == key) {
         assert(PyUnicode_CheckExact(startkey));
         return 1;
@@ -1450,7 +1450,7 @@ compare_generic_threadsafe(PyDictObject *mp, PyDictKeysObject *dk,
                            void *ep0, Py_ssize_t ix, PyObject *key, Py_hash_t hash)
 {
     PyDictKeyEntry *ep = &((PyDictKeyEntry *)ep0)[ix];
-    PyObject *startkey = _Py_atomic_load_ptr_relaxed(&ep->me_key);
+    PyObject *startkey = _Py_atomic_load_ptr_consume(&ep->me_key);
     if (startkey == key) {
         return 1;
     }
@@ -5526,7 +5526,7 @@ dictiter_iternext_threadsafe(PyDictObject *d, PyObject *self,
     k = _Py_atomic_load_ptr_acquire(&d->ma_keys);
     assert(i >= 0);
     if (_PyDict_HasSplitTable(d)) {
-        PyDictValues *values = _Py_atomic_load_ptr_relaxed(&d->ma_values);
+        PyDictValues *values = _Py_atomic_load_ptr_consume(&d->ma_values);
         if (values == NULL) {
             goto concurrent_modification;
         }
@@ -7114,7 +7114,7 @@ _PyObject_TryGetInstanceAttribute(PyObject *obj, PyObject *name, PyObject **attr
     Py_BEGIN_CRITICAL_SECTION(dict);
 
     if (dict->ma_values == values && FT_ATOMIC_LOAD_UINT8(values->valid)) {
-        value = _Py_atomic_load_ptr_relaxed(&values->values[ix]);
+        value = _Py_atomic_load_ptr_consume(&values->values[ix]);
         *attr = _Py_XNewRefWithLock(value);
         success = true;
     } else {