Merge pull request #1021 from mvanhorn/fix/960-atomic-store-nfs-eperm

fix: fall back to a direct write when AtomicJsonStore.save cannot use a temp file (NFS EPERM)

app/state_store.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import base64
 import collections.abc
+import errno
 import json
 import logging
 import os
@@ -17,6 +18,25 @@ STATE_SCHEMA_VERSION = 2
 _BYTES_MARKER = "__metube_bytes__"
 _DATETIME_MARKER = "__metube_datetime__"
 
+# Errnos that signal the filesystem cannot support the temp-file + rename
+# atomic-write strategy (for example an NFS-backed state dir returning EPERM on
+# mkstemp). These are safe to fall back on because they mean the atomic
+# mechanism is unavailable, not that the data write itself failed. Errors like
+# ENOSPC/EIO are deliberately excluded so a genuine storage failure surfaces
+# instead of silently truncating an existing good state file.
+_ATOMIC_UNSUPPORTED_ERRNOS = frozenset(
+    e
+    for e in (
+        errno.EPERM,
+        errno.EACCES,
+        errno.ENOSYS,
+        errno.EINVAL,
+        getattr(errno, "EOPNOTSUPP", None),
+        getattr(errno, "ENOTSUP", None),
+    )
+    if e is not None
+)
+
 
 def to_json_compatible(value: Any) -> Any:
     if value is None or isinstance(value, (bool, int, float, str)):
@@ -62,6 +82,7 @@ class AtomicJsonStore:
         self.path = path
         self.kind = kind
         self.schema_version = schema_version
+        self._direct_write_fallback_warned = False
 
     def _ensure_parent(self) -> None:
         parent = os.path.dirname(self.path)
@@ -96,6 +117,16 @@ class AtomicJsonStore:
     def save(self, data: dict[str, Any]) -> None:
         self._ensure_parent()
         payload = self._build_payload(data)
+        try:
+            self._atomic_write(payload)
+        except OSError as exc:
+            if exc.errno not in _ATOMIC_UNSUPPORTED_ERRNOS:
+                raise
+            self._warn_direct_write_fallback(exc)
+            self._direct_write(payload)
+
+    def _atomic_write(self, payload: dict[str, Any]) -> None:
+        text = self._serialize(payload)
         parent = os.path.dirname(self.path) or "."
         fd, tmp_path = tempfile.mkstemp(
             prefix=f".{os.path.basename(self.path)}.",
@@ -105,10 +136,9 @@ class AtomicJsonStore:
         )
         try:
             with os.fdopen(fd, "w", encoding="utf-8") as f:
-                json.dump(payload, f, ensure_ascii=False, separators=(",", ":"))
-                f.write("\n")
+                f.write(text)
                 f.flush()
-                os.fsync(f.fileno())
+                self._best_effort_fsync(f.fileno())
             os.replace(tmp_path, self.path)
             self._fsync_directory(parent)
         except Exception:
@@ -118,6 +148,57 @@ class AtomicJsonStore:
                 pass
             raise
 
+    def _direct_write(self, payload: dict[str, Any]) -> None:
+        # Serialize before truncating so a serialization failure never destroys
+        # the existing state file (the atomic path gets this for free via its
+        # temp file).
+        text = self._serialize(payload)
+        # Create with 0o600 so the fallback keeps the owner-only permissions the
+        # atomic path gets from mkstemp; state files can contain URLs and
+        # per-download option overrides that must not leak on shared mounts.
+        fd = os.open(self.path, os.O_WRONLY | os.O_CREAT | os.O_TRUNC, 0o600)
+        with os.fdopen(fd, "w", encoding="utf-8") as f:
+            # The 0o600 mode above only applies when the file is created; force
+            # it on rewrites too so an existing, broadly-permissioned state file
+            # is tightened to match the atomic path. Best-effort because some
+            # network filesystems reject chmod, and that must not re-crash save.
+            try:
+                os.fchmod(f.fileno(), 0o600)
+            except OSError:
+                pass
+            f.write(text)
+            f.flush()
+            self._best_effort_fsync(f.fileno())
+        # Make the new directory entry durable too, matching the atomic path.
+        self._fsync_directory(os.path.dirname(self.path) or ".")
+
+    @staticmethod
+    def _best_effort_fsync(fileno: int) -> None:
+        # Tolerate fsync being unsupported on the underlying filesystem (for
+        # example a network mount that returns EINVAL/ENOSYS), but let genuine
+        # storage failures such as ENOSPC/EIO surface so a non-durable write is
+        # never reported as success. An unsupported fsync must not by itself
+        # abandon the atomic rename path.
+        try:
+            os.fsync(fileno)
+        except OSError as exc:
+            if exc.errno not in _ATOMIC_UNSUPPORTED_ERRNOS:
+                raise
+
+    @staticmethod
+    def _serialize(payload: dict[str, Any]) -> str:
+        return json.dumps(payload, ensure_ascii=False, separators=(",", ":")) + "\n"
+
+    def _warn_direct_write_fallback(self, exc: OSError) -> None:
+        if self._direct_write_fallback_warned:
+            return
+        self._direct_write_fallback_warned = True
+        log.warning(
+            "Atomic state write failed for %s (%s); falling back to direct write",
+            self.path,
+            exc,
+        )
+
     def quarantine_invalid_file(self, exc: Exception) -> None:
         if not os.path.exists(self.path):
             return

app/tests/test_state_store.py

@@ -4,6 +4,7 @@ import os
 import tempfile
 import unittest
 from datetime import datetime
+from unittest.mock import patch
 
 from state_store import AtomicJsonStore, from_json_compatible, to_json_compatible
 
@@ -21,6 +22,135 @@ class StateStoreTests(unittest.TestCase):
             self.assertEqual(payload["schema_version"], 2)
             self.assertEqual(payload["items"][0]["info"]["title"], "hello")
 
+    def test_save_falls_back_to_direct_write_when_mkstemp_fails(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            path = os.path.join(tmp, "queue.json")
+            store = AtomicJsonStore(path, kind="persistent_queue:queue")
+
+            with self.assertLogs("state_store", level="WARNING") as logs:
+                with patch(
+                    "state_store.tempfile.mkstemp",
+                    side_effect=PermissionError(1, "Operation not permitted"),
+                ):
+                    store.save({"items": [{"key": "a"}]})
+
+            self.assertTrue(os.path.exists(path))
+            self.assertTrue(any(path in message for message in logs.output))
+            # Fallback keeps owner-only permissions, matching the atomic path.
+            self.assertEqual(os.stat(path).st_mode & 0o777, 0o600)
+            payload = store.load()
+            self.assertEqual(payload["items"], [{"key": "a"}])
+
+    def test_fallback_tightens_permissions_on_existing_file(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            path = os.path.join(tmp, "queue.json")
+            with open(path, "w", encoding="utf-8") as f:
+                f.write("{}")
+            os.chmod(path, 0o644)
+
+            store = AtomicJsonStore(path, kind="persistent_queue:queue")
+            with patch(
+                "state_store.tempfile.mkstemp",
+                side_effect=PermissionError(1, "Operation not permitted"),
+            ):
+                store.save({"items": [{"key": "a"}]})
+
+            self.assertEqual(os.stat(path).st_mode & 0o777, 0o600)
+            self.assertEqual(store.load()["items"], [{"key": "a"}])
+
+    def test_save_falls_back_to_direct_write_when_replace_fails(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            path = os.path.join(tmp, "queue.json")
+            store = AtomicJsonStore(path, kind="persistent_queue:queue")
+
+            with patch(
+                "state_store.os.replace",
+                side_effect=PermissionError(1, "Operation not permitted"),
+            ):
+                store.save({"items": [{"key": "a"}]})
+
+            self.assertTrue(os.path.exists(path))
+            payload = store.load()
+            self.assertEqual(payload["items"], [{"key": "a"}])
+            self.assertEqual([], [name for name in os.listdir(tmp) if name.endswith(".tmp")])
+
+    def test_save_reraises_when_atomic_and_direct_write_fail(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            path = os.path.join(tmp, "queue.json")
+            store = AtomicJsonStore(path, kind="persistent_queue:queue")
+
+            with patch(
+                "state_store.tempfile.mkstemp",
+                side_effect=PermissionError(1, "Operation not permitted"),
+            ):
+                with patch(
+                    "state_store.os.open",
+                    side_effect=PermissionError(13, "Permission denied"),
+                ):
+                    with self.assertRaises(PermissionError) as ctx:
+                        store.save({"items": [{"key": "a"}]})
+
+            self.assertEqual(ctx.exception.errno, 13)
+            self.assertFalse(os.path.exists(path))
+
+    def test_unsupported_fsync_keeps_atomic_path(self):
+        # fsync being unsupported (EINVAL/ENOSYS) must not by itself trigger the
+        # direct-write fallback; the atomic temp-file + rename path still runs.
+        import errno as _errno
+
+        with tempfile.TemporaryDirectory() as tmp:
+            path = os.path.join(tmp, "queue.json")
+            store = AtomicJsonStore(path, kind="persistent_queue:queue")
+
+            with patch(
+                "state_store.os.fsync",
+                side_effect=OSError(_errno.EINVAL, "Invalid argument"),
+            ):
+                with self.assertNoLogs("state_store", level="WARNING"):
+                    store.save({"items": [{"key": "a"}]})
+
+            self.assertEqual(store.load()["items"], [{"key": "a"}])
+            self.assertEqual([], [name for name in os.listdir(tmp) if name.endswith(".tmp")])
+
+    def test_save_reraises_and_preserves_state_on_non_atomic_errno(self):
+        # A storage failure such as ENOSPC is not an "atomic unavailable"
+        # signal, so it must surface instead of falling back to a direct write
+        # that would truncate the existing good state file.
+        import errno as _errno
+
+        with tempfile.TemporaryDirectory() as tmp:
+            path = os.path.join(tmp, "queue.json")
+            store = AtomicJsonStore(path, kind="persistent_queue:queue")
+            store.save({"items": [{"key": "good"}]})
+
+            with patch(
+                "state_store.tempfile.mkstemp",
+                side_effect=OSError(_errno.ENOSPC, "No space left on device"),
+            ):
+                with self.assertRaises(OSError) as ctx:
+                    store.save({"items": [{"key": "new"}]})
+
+            self.assertEqual(ctx.exception.errno, _errno.ENOSPC)
+            # Existing state is untouched.
+            self.assertEqual(store.load()["items"], [{"key": "good"}])
+
+    def test_serialization_failure_preserves_existing_state(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            path = os.path.join(tmp, "queue.json")
+            store = AtomicJsonStore(path, kind="persistent_queue:queue")
+            store.save({"items": [{"key": "good"}]})
+
+            # Even on the fallback path, a non-serializable payload must raise
+            # before the existing good state file is touched.
+            with patch(
+                "state_store.tempfile.mkstemp",
+                side_effect=PermissionError(1, "Operation not permitted"),
+            ):
+                with self.assertRaises(TypeError):
+                    store.save({"items": object()})
+
+            self.assertEqual(store.load()["items"], [{"key": "good"}])
+
     def test_invalid_file_is_quarantined(self):
         with tempfile.TemporaryDirectory() as tmp:
             path = os.path.join(tmp, "queue.json")