Bugfix: No such file or directory: '/data/homeassistant/home-assistant_v2.db-shm' (#1795)

* Do not use `tar_file.add` to recursively add backup folder

As the folders might contain files which are being removed temporarily (e.g. shared memory file of sqlite database), relying on `tar_file.add` becomes problematic as it crashes the whole backup process if a file does not exist anymore.
This becomes annoying, if the file which causes the error should be excluded by the filter.

To workaround this issue, we manually iterating over the files/directories and apply filters before passing the file or directory to the `tar_file.add` method. As per [documentation](https://docs.python.org/3/library/pathlib.html#pure-paths), pure path does not access the file system.

Fixes #779

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Remove unused import

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Applied code review suggestions

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Applied codestyle

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Remove `pathlib` util and move `is_excluded_by_filter` into `tar` utils

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Rename method

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Rename `origin_dir` to `origin_path` and apply `Path` typehint

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Codestyle

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Add comment why we add the directory even if we are iterating over all its items

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Use `atomic_contents_add` from tar utils to archive addon data

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Remove unused function `exclude_filter`

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Remove unsecure default list value

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Some more codestyle

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Lowercase method name `Path.joinpath`

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Fix codestyle and use proper variable

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Add test for `_is_excluded_by_filter`

Signed-off-by: Maximilian Bösing <2189546+boesing@users.noreply.github.com>

* Update addon.py

Co-authored-by: Pascal Vizeli <pascal.vizeli@syshack.ch>

supervisor/addons/addon.py

@@ -51,7 +51,7 @@ from ..exceptions import (
 )
 from ..utils.apparmor import adjust_profile
 from ..utils.json import read_json_file, write_json_file
-from ..utils.tar import exclude_filter, secure_path
+from ..utils.tar import secure_path, atomic_contents_add
 from .model import AddonModel, Data
 from .utils import remove_data
 from .validate import SCHEMA_ADDON_SNAPSHOT, validate_options
@@ -534,10 +534,12 @@ class Addon(AddonModel):
     async def snapshot(self, tar_file: tarfile.TarFile) -> None:
         """Snapshot state of an add-on."""
         with TemporaryDirectory(dir=self.sys_config.path_tmp) as temp:
+            temp_path = Path(temp)
+
             # store local image
             if self.need_build:
                 try:
-                    await self.instance.export_image(Path(temp, "image.tar"))
+                    await self.instance.export_image(temp_path.joinpath("image.tar"))
                 except DockerAPIError:
                     raise AddonsError() from None
 
@@ -550,14 +552,14 @@ class Addon(AddonModel):
 
             # Store local configs/state
             try:
-                write_json_file(Path(temp, "addon.json"), data)
+                write_json_file(temp_path.joinpath("addon.json"), data)
             except JsonFileError:
                 _LOGGER.error("Can't save meta for %s", self.slug)
                 raise AddonsError() from None
 
             # Store AppArmor Profile
             if self.sys_host.apparmor.exists(self.slug):
-                profile = Path(temp, "apparmor.txt")
+                profile = temp_path.joinpath("apparmor.txt")
                 try:
                     self.sys_host.apparmor.backup_profile(self.slug, profile)
                 except HostAppArmorError:
@@ -569,13 +571,15 @@ class Addon(AddonModel):
                 """Write tar inside loop."""
                 with tar_file as snapshot:
                     # Snapshot system
+
                     snapshot.add(temp, arcname=".")
 
                     # Snapshot data
-                    snapshot.add(
+                    atomic_contents_add(
+                        snapshot,
                         self.path_data,
+                        excludes=self.snapshot_exclude,
                         arcname="data",
-                        filter=exclude_filter(self.snapshot_exclude),
                     )
 
             try:

supervisor/snapshots/snapshot.py

@@ -42,7 +42,11 @@ from ..const import (
 from ..coresys import CoreSys, CoreSysAttributes
 from ..exceptions import AddonsError
 from ..utils.json import write_json_file
-from ..utils.tar import SecureTarFile, exclude_filter, secure_path
+from ..utils.tar import (
+    SecureTarFile,
+    secure_path,
+    atomic_contents_add,
+)
 from .utils import key_to_iv, password_for_validating, password_to_key, remove_folder
 from .validate import ALL_FOLDERS, SCHEMA_SNAPSHOT
 
@@ -370,10 +374,11 @@ class Snapshot(CoreSysAttributes):
             try:
                 _LOGGER.info("Snapshot folder %s", name)
                 with SecureTarFile(tar_name, "w", key=self._key) as tar_file:
-                    tar_file.add(
+                    atomic_contents_add(
+                        tar_file,
                         origin_dir,
+                        excludes=MAP_FOLDER_EXCLUDE.get(name, []),
                         arcname=".",
-                        filter=exclude_filter(MAP_FOLDER_EXCLUDE.get(name, [])),
                     )
 
                 _LOGGER.info("Snapshot folder %s done", name)

supervisor/utils/tar.py

@@ -2,9 +2,9 @@
 import hashlib
 import logging
 import os
-from pathlib import Path
+from pathlib import Path, PurePath
 import tarfile
-from typing import IO, Callable, Generator, List, Optional
+from typing import IO, Generator, List, Optional
 
 from cryptography.hazmat.backends import default_backend
 from cryptography.hazmat.primitives import padding
@@ -134,20 +134,41 @@ def secure_path(tar: tarfile.TarFile) -> Generator[tarfile.TarInfo, None, None]:
             yield member
 
 
-def exclude_filter(
+def _is_excluded_by_filter(path: PurePath, exclude_list: List[str]) -> bool:
-    exclude_list: List[str],
-) -> Callable[[tarfile.TarInfo], Optional[tarfile.TarInfo]]:
-    """Create callable filter function to check TarInfo for add."""
-
-    def my_filter(tar: tarfile.TarInfo) -> Optional[tarfile.TarInfo]:
     """Filter to filter excludes."""
-        file_path = Path(tar.name)
+
     for exclude in exclude_list:
-            if not file_path.match(exclude):
+        if not path.match(exclude):
             continue
-            _LOGGER.debug("Ignore %s because of %s", file_path, exclude)
+        _LOGGER.debug("Ignore %s because of %s", path, exclude)
+        return True
+
+    return False
+
+
+def atomic_contents_add(
+    tar_file: tarfile.TarFile,
+    origin_path: Path,
+    excludes: List[str],
+    arcname: str = ".",
+) -> None:
+    """Append directories and/or files to the TarFile if excludes wont filter."""
+
+    if _is_excluded_by_filter(origin_path, excludes):
         return None
 
-        return tar
+    # Add directory only (recursive=False) to ensure we also archive empty directories
+    tar_file.add(origin_path.as_posix(), arcname, recursive=False)
 
-    return my_filter
+    for directory_item in origin_path.iterdir():
+        if _is_excluded_by_filter(directory_item, excludes):
+            continue
+
+        arcpath = PurePath(arcname, directory_item.name).as_posix()
+        if directory_item.is_dir():
+            atomic_contents_add(tar_file, directory_item.as_posix(), excludes, arcpath)
+            continue
+
+        tar_file.add(directory_item.as_posix(), arcname=arcpath)
+
+    return None

tests/utils/test_tarfile.py

@@ -2,7 +2,8 @@
 
 import attr
 
-from supervisor.utils.tar import exclude_filter, secure_path
+from pathlib import PurePath
+from supervisor.utils.tar import secure_path, _is_excluded_by_filter
 
 
 @attr.s
@@ -33,28 +34,29 @@ def test_not_secure_path():
     assert [] == list(secure_path(test_list))
 
 
-def test_exclude_filter_good():
+def test_is_excluded_by_filter_good():
     """Test exclude filter."""
-    filter_funct = exclude_filter(["not/match", "/dev/xy"])
+    filter_list = ["not/match", "/dev/xy"]
     test_list = [
-        TarInfo("test.txt"),
+        PurePath("test.txt"),
-        TarInfo("data/xy.blob"),
+        PurePath("data/xy.blob"),
-        TarInfo("bla/blu/ble"),
+        PurePath("bla/blu/ble"),
-        TarInfo("data/../xy.blob"),
+        PurePath("data/../xy.blob"),
     ]
 
-    assert test_list == [filter_funct(result) for result in test_list]
+    for path_object in test_list:
+        assert _is_excluded_by_filter(path_object, filter_list) is False
 
 
-def test_exclude_filter_bad():
+def test_is_exclude_by_filter_bad():
     """Test exclude filter."""
-    filter_funct = exclude_filter(["*.txt", "data/*", "bla/blu/ble"])
+    filter_list = ["*.txt", "data/*", "bla/blu/ble"]
     test_list = [
-        TarInfo("test.txt"),
+        PurePath("test.txt"),
-        TarInfo("data/xy.blob"),
+        PurePath("data/xy.blob"),
-        TarInfo("bla/blu/ble"),
+        PurePath("bla/blu/ble"),
-        TarInfo("data/test_files/kk.txt"),
+        PurePath("data/test_files/kk.txt"),
     ]
 
-    for info in [filter_funct(result) for result in test_list]:
+    for path_object in test_list:
-        assert info is None
+        assert _is_excluded_by_filter(path_object, filter_list) is True