gh-121368: Fix seq lock memory ordering in _PyType_Lookup (#121388)

The `_PySeqLock_EndRead` function needs an acquire fence to ensure that
the load of the sequence happens after any loads within the read side
critical section. The missing fence can trigger bugs on macOS arm64.

Additionally, we need a release fence in `_PySeqLock_LockWrite` to
ensure that the sequence update is visible before any modifications to
the cache entry.
This commit is contained in:
Sam Gross 2024-07-08 14:52:07 -04:00 committed by GitHub
parent 31873bea47
commit 1d3cf79a50
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
9 changed files with 51 additions and 17 deletions

View file

@ -5387,7 +5387,7 @@ _PyType_LookupRef(PyTypeObject *type, PyObject *name)
#ifdef Py_GIL_DISABLED
// synchronize-with other writing threads by doing an acquire load on the sequence
while (1) {
int sequence = _PySeqLock_BeginRead(&entry->sequence);
uint32_t sequence = _PySeqLock_BeginRead(&entry->sequence);
uint32_t entry_version = _Py_atomic_load_uint32_relaxed(&entry->version);
uint32_t type_version = _Py_atomic_load_uint32_acquire(&type->tp_version_tag);
if (entry_version == type_version &&