Avoid race due to deleted imports when cleaning up objects at interpreter shutdown (#973)

* is_shutting_down

* Update cuda_bindings/cuda/bindings/utils/__init__.py

* safe_close()

* is_finalizing

* refactor

* add note

Author: brandon-b-miller

cuda_bindings/cuda/bindings/utils/__init__.py

@@ -1,6 +1,5 @@
 # SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: LicenseRef-NVIDIA-SOFTWARE-LICENSE
-
 from typing import Any, Callable
 
 from ._ptx_utils import get_minimal_required_cuda_ver_from_ptx_ver, get_ptx_ver

cuda_core/cuda/core/experimental/_event.pyx

@@ -18,7 +18,7 @@ from cuda.core.experimental._utils.cuda_utils import (
     driver,
     handle_return,
 )
-
+import sys
 if TYPE_CHECKING:
     import cuda.bindings
     from cuda.core.experimental._device import Device
@@ -108,15 +108,20 @@ cdef class Event:
         self._ctx_handle = ctx_handle
         return self
 
-    cpdef close(self):
-        """Destroy the event."""
+    cdef _shutdown_safe_close(self, is_shutting_down=sys.is_finalizing):
+        if is_shutting_down and is_shutting_down():
+            return
         if self._handle is not None:
             err, = driver.cuEventDestroy(self._handle)
             self._handle = None
             raise_if_driver_error(err)
 
+    cpdef close(self):
+        """Destroy the event."""
+        self._shutdown_safe_close(is_shutting_down=None)
+
     def __del__(self):
-        self.close()
+        self._shutdown_safe_close()
 
     def __isub__(self, other):
         return NotImplemented

cuda_core/cuda/core/experimental/_memory.pyx

@@ -9,6 +9,7 @@ from cuda.core.experimental._utils.cuda_utils cimport (
     _check_driver_error as raise_if_driver_error,
     check_or_create_options,
 )
+import sys
 
 from dataclasses import dataclass
 from typing import TypeVar, Union, TYPE_CHECKING
@@ -69,7 +70,16 @@ cdef class Buffer:
         return self
 
     def __del__(self):
-        self.close()
+        self._shutdown_safe_close()
+
+    cdef _shutdown_safe_close(self, stream: Stream = None, is_shutting_down=sys.is_finalizing):
+        if is_shutting_down and is_shutting_down():
+            return
+        if self._ptr and self._mr is not None:
+            self._mr.deallocate(self._ptr, self._size, stream)
+            self._ptr = 0
+            self._mr = None
+            self._ptr_obj = None
 
     cpdef close(self, stream: Stream = None):
         """Deallocate this buffer asynchronously on the given stream.
@@ -83,11 +93,7 @@ cdef class Buffer:
             The stream object to use for asynchronous deallocation. If None,
             the behavior depends on the underlying memory resource.
         """
-        if self._ptr and self._mr is not None:
-            self._mr.deallocate(self._ptr, self._size, stream)
-            self._ptr = 0
-            self._mr = None
-            self._ptr_obj = None
+        self._shutdown_safe_close(stream, is_shutting_down=None)
 
     @property
     def handle(self) -> DevicePointerT:

cuda_core/cuda/core/experimental/_stream.pyx

@@ -8,6 +8,7 @@ from cuda.core.experimental._utils.cuda_utils cimport (
     _check_driver_error as raise_if_driver_error,
     check_or_create_options,
 )
+import sys
 
 import cython
 import os
@@ -186,22 +187,28 @@ cdef class Stream:
         return self
 
     def __del__(self):
-        self.close()
+        self._shutdown_safe_close()
 
-    cpdef close(self):
-        """Destroy the stream.
+    cdef _shutdown_safe_close(self, is_shutting_down=sys.is_finalizing):
+        if is_shutting_down and is_shutting_down():
+            return
 
-        Destroy the stream if we own it. Borrowed foreign stream
-        object will instead have their references released.
-
-        """
         if self._owner is None:
             if self._handle and not self._builtin:
                 handle_return(driver.cuStreamDestroy(self._handle))
         else:
             self._owner = None
         self._handle = None
 
+    cpdef close(self):
+        """Destroy the stream.
+
+        Destroy the stream if we own it. Borrowed foreign stream
+        object will instead have their references released.
+
+        """
+        self._shutdown_safe_close(is_shutting_down=None)
+
     def __cuda_stream__(self) -> tuple[int, int]:
         """Return an instance of a __cuda_stream__ protocol."""
         return (0, int(self.handle))

cuda_core/docs/source/release/0.X.Y-notes.rst

@@ -47,3 +47,4 @@ Fixes and enhancements
 - Fixed a bug in :class:`GraphBuilder.add_child` where dependencies extracted from capturing stream were passed inconsistently with num_dependencies parameter (addresses issue #843).
 - Make :class:`Buffer` creation more performant.
 - Enabled :class:`MemoryResource` subclasses to accept :class:`Device` objects, in addition to previously supported device ordinals.
+- Fixed a bug in :class:`Stream` and other classes where object cleanup would error during interpreter shutdown.