1
0
mirror of synced 2024-11-22 01:04:23 +01:00
mat2/libmat2/archive.py
jvoisin 180ea24e5a Remove pyflakes
Isn't borderline useless compared to mypy and pylint
2022-11-21 19:57:38 +01:00

461 lines
19 KiB
Python

import abc
import stat
import zipfile
import datetime
import tarfile
import tempfile
import os
import logging
import shutil
from typing import Pattern, Union, Any
from . import abstract, UnknownMemberPolicy, parser_factory
# pylint: disable=not-callable,assignment-from-no-return,too-many-branches
# An ArchiveClass is a class representing an archive,
# while an ArchiveMember is a class representing an element
# (usually a file) of an archive.
ArchiveClass = Union[zipfile.ZipFile, tarfile.TarFile]
ArchiveMember = Union[zipfile.ZipInfo, tarfile.TarInfo]
class ArchiveBasedAbstractParser(abstract.AbstractParser):
"""Base class for all archive-based formats.
Welcome to a world of frustrating complexity and tediouness:
- A lot of file formats (docx, odt, epubs, …) are archive-based,
so we need to add callbacks erverywhere to allow their respective
parsers to apply specific cleanup to the required files.
- Python has two different modules to deal with .tar and .zip files,
with similar-but-yet-o-so-different API, so we need to write
a ghetto-wrapper to avoid duplicating everything
- The combination of @staticmethod and @abstractstaticmethod is
required because for now, mypy doesn't know that
@abstractstaticmethod is, indeed, a static method.
- Mypy is too dumb (yet) to realise that a type A is valid under
the Union[A, B] constrain, hence the weird `# type: ignore`
annotations.
"""
# Tarfiles can optionally support compression
# https://docs.python.org/3/library/tarfile.html#tarfile.open
compression = ''
def __init__(self, filename):
super().__init__(filename)
# We ignore typing here because mypy is too stupid
self.archive_class = None # type: ignore
self.member_class = None # type: ignore
# Those are the files that have a format that _isn't_
# supported by mat2, but that we want to keep anyway.
self.files_to_keep = set() # type: set[Pattern]
# Those are the files that we _do not_ want to keep,
# no matter if they are supported or not.
self.files_to_omit = set() # type: set[Pattern]
# what should the parser do if it encounters an unknown file in
# the archive?
self.unknown_member_policy = UnknownMemberPolicy.ABORT # type: UnknownMemberPolicy
# The LGTM comment is to mask a false-positive,
# see https://lgtm.com/projects/g/jvoisin/mat2/
self.is_archive_valid() # lgtm [py/init-calls-subclass]
def is_archive_valid(self):
"""Raise a ValueError is the current archive isn't a valid one."""
def _specific_cleanup(self, full_path: str) -> bool:
""" This method can be used to apply specific treatment
to files present in the archive."""
# pylint: disable=unused-argument
return True # pragma: no cover
def _specific_get_meta(self, full_path: str, file_path: str) -> dict[str, Any]:
""" This method can be used to extract specific metadata
from files present in the archive."""
# pylint: disable=unused-argument
return {} # pragma: no cover
def _final_checks(self) -> bool:
""" This method is invoked after the file has been cleaned,
allowing to run final verifications.
"""
# pylint: disable=unused-argument
return True
@staticmethod
@abc.abstractmethod
def _get_all_members(archive: ArchiveClass) -> list[ArchiveMember]:
"""Return all the members of the archive."""
@staticmethod
@abc.abstractmethod
def _clean_member(member: ArchiveMember) -> ArchiveMember:
"""Remove all the metadata for a given member."""
@staticmethod
@abc.abstractmethod
def _get_member_meta(member: ArchiveMember) -> dict[str, str]:
"""Return all the metadata of a given member."""
@staticmethod
@abc.abstractmethod
def _get_member_name(member: ArchiveMember) -> str:
"""Return the name of the given member."""
@abc.abstractmethod
def _add_file_to_archive(self, archive: ArchiveClass, member: ArchiveMember,
full_path: str):
"""Add the file at full_path to the archive, via the given member."""
@staticmethod
def _set_member_permissions(member: ArchiveMember, permissions: int) -> ArchiveMember:
"""Set the permission of the archive member."""
# pylint: disable=unused-argument
return member
@staticmethod
def _get_member_compression(member: ArchiveMember):
"""Get the compression of the archive member."""
# pylint: disable=unused-argument
return None
@staticmethod
def _set_member_compression(member: ArchiveMember, compression) -> ArchiveMember:
"""Set the compression of the archive member."""
# pylint: disable=unused-argument
return member
def get_meta(self) -> dict[str, Union[str, dict]]:
meta = dict() # type: dict[str, Union[str, dict]]
with self.archive_class(self.filename) as zin:
temp_folder = tempfile.mkdtemp()
for item in self._get_all_members(zin):
local_meta = self._get_member_meta(item)
member_name = self._get_member_name(item)
if member_name[-1] == '/': # pragma: no cover
# `is_dir` is added in Python3.6
continue # don't keep empty folders
zin.extract(member=item, path=temp_folder)
full_path = os.path.join(temp_folder, member_name)
os.chmod(full_path, stat.S_IRUSR)
specific_meta = self._specific_get_meta(full_path, member_name)
local_meta = {**local_meta, **specific_meta}
member_parser, _ = parser_factory.get_parser(full_path) # type: ignore
if member_parser:
local_meta = {**local_meta, **member_parser.get_meta()}
if local_meta:
meta[member_name] = local_meta
shutil.rmtree(temp_folder)
return meta
def remove_all(self) -> bool:
# pylint: disable=too-many-branches
with self.archive_class(self.filename) as zin,\
self.archive_class(self.output_filename, 'w' + self.compression) as zout:
temp_folder = tempfile.mkdtemp()
abort = False
# Sort the items to process, to reduce fingerprinting,
# and keep them in the `items` variable.
items = list() # type: list[ArchiveMember]
for item in sorted(self._get_all_members(zin), key=self._get_member_name):
# Some fileformats do require to have the `mimetype` file
# as the first file in the archive.
if self._get_member_name(item) == 'mimetype':
items.insert(0, item)
else:
items.append(item)
# Since files order is a fingerprint factor,
# we're iterating (and thus inserting) them in lexicographic order.
for item in items:
member_name = self._get_member_name(item)
if member_name[-1] == '/': # `is_dir` is added in Python3.6
continue # don't keep empty folders
full_path = os.path.join(temp_folder, member_name)
if not os.path.abspath(full_path).startswith(temp_folder):
logging.error("%s contains a file (%s) pointing outside (%s) of its root.",
self.filename, member_name, full_path)
abort = True
break
zin.extract(member=item, path=temp_folder)
try:
original_permissions = os.stat(full_path).st_mode
except FileNotFoundError:
logging.error("Something went wrong during processing of "
"%s in %s, likely a path traversal attack.",
member_name, self.filename)
abort = True
# we're breaking instead of continuing, because this exception
# is raised in case of weird path-traversal-like atttacks.
break
os.chmod(full_path, original_permissions | stat.S_IWUSR | stat.S_IRUSR)
original_compression = self._get_member_compression(item)
if self._specific_cleanup(full_path) is False:
logging.warning("Something went wrong during deep cleaning of %s in %s",
member_name, self.filename)
abort = True
continue
if any(map(lambda r: r.search(member_name), self.files_to_keep)):
# those files aren't supported, but we want to add them anyway
pass
elif any(map(lambda r: r.search(member_name), self.files_to_omit)):
continue
else: # supported files that we want to first clean, then add
member_parser, mtype = parser_factory.get_parser(full_path) # type: ignore
if not member_parser:
if self.unknown_member_policy == UnknownMemberPolicy.OMIT:
logging.warning("In file %s, omitting unknown element %s (format: %s)",
self.filename, member_name, mtype)
continue
elif self.unknown_member_policy == UnknownMemberPolicy.KEEP:
logging.warning("In file %s, keeping unknown element %s (format: %s)",
self.filename, member_name, mtype)
else:
logging.error("In file %s, element %s's format (%s) "
"isn't supported",
self.filename, member_name, mtype)
abort = True
continue
else:
if member_parser.remove_all() is False:
logging.warning("In file %s, something went wrong \
with the cleaning of %s \
(format: %s)",
self.filename, member_name, mtype)
abort = True
continue
os.rename(member_parser.output_filename, full_path)
zinfo = self.member_class(member_name) # type: ignore
zinfo = self._set_member_permissions(zinfo, original_permissions)
zinfo = self._set_member_compression(zinfo, original_compression)
clean_zinfo = self._clean_member(zinfo)
self._add_file_to_archive(zout, clean_zinfo, full_path)
shutil.rmtree(temp_folder)
if abort:
os.remove(self.output_filename)
return False
if not self._final_checks():
return False # pragma: no cover
return True
class TarParser(ArchiveBasedAbstractParser):
mimetypes = {'application/x-tar'}
def __init__(self, filename):
super().__init__(filename)
# yes, it's tarfile.open and not tarfile.TarFile,
# as stated in the documentation:
# https://docs.python.org/3/library/tarfile.html#tarfile.TarFile
# This is required to support compressed archives.
self.archive_class = tarfile.open
self.member_class = tarfile.TarInfo
def is_archive_valid(self):
if tarfile.is_tarfile(self.filename) is False:
raise ValueError
self.__check_tarfile_safety()
def __check_tarfile_safety(self):
"""Checks if the tarfile doesn't have any "suspicious" members.
This is a rewrite of this patch: https://bugs.python.org/file47826/safetarfile-4.diff
inspired by this bug from 2014: https://bugs.python.org/issue21109
because Python's stdlib doesn't provide a way to "safely" extract
things from a tar file.
"""
names = set()
with tarfile.open(self.filename) as f:
members = f.getmembers()
for member in members:
name = member.name
if os.path.isabs(name):
raise ValueError("The archive %s contains a file with an " \
"absolute path: %s" % (self.filename, name))
elif os.path.normpath(name).startswith('../') or '/../' in name:
raise ValueError("The archive %s contains a file with an " \
"path traversal attack: %s" % (self.filename, name))
if name in names:
raise ValueError("The archive %s contains two times the same " \
"file: %s" % (self.filename, name))
else:
names.add(name)
if member.isfile():
if member.mode & stat.S_ISUID:
raise ValueError("The archive %s contains a setuid file: %s" % \
(self.filename, name))
elif member.mode & stat.S_ISGID:
raise ValueError("The archive %s contains a setgid file: %s" % \
(self.filename, name))
elif member.issym():
linkname = member.linkname
if os.path.normpath(linkname).startswith('..'):
raise ValueError('The archive %s contains a symlink pointing' \
'outside of the archive via a path traversal: %s -> %s' % \
(self.filename, name, linkname))
if os.path.isabs(linkname):
raise ValueError('The archive %s contains a symlink pointing' \
'outside of the archive: %s -> %s' % \
(self.filename, name, linkname))
elif member.isdev():
raise ValueError("The archive %s contains a non-regular " \
"file: %s" % (self.filename, name))
elif member.islnk():
raise ValueError("The archive %s contains a hardlink: %s" \
% (self.filename, name))
@staticmethod
def _clean_member(member: ArchiveMember) -> ArchiveMember:
assert isinstance(member, tarfile.TarInfo) # please mypy
member.mtime = member.uid = member.gid = 0
member.uname = member.gname = ''
return member
@staticmethod
def _get_member_meta(member: ArchiveMember) -> dict[str, str]:
assert isinstance(member, tarfile.TarInfo) # please mypy
metadata = {}
if member.mtime != 0:
metadata['mtime'] = str(datetime.datetime.fromtimestamp(member.mtime))
if member.uid != 0:
metadata['uid'] = str(member.uid)
if member.gid != 0:
metadata['gid'] = str(member.gid)
if member.uname != '':
metadata['uname'] = member.uname
if member.gname != '':
metadata['gname'] = member.gname
return metadata
def _add_file_to_archive(self, archive: ArchiveClass, member: ArchiveMember,
full_path: str):
assert isinstance(member, tarfile.TarInfo) # please mypy
assert isinstance(archive, tarfile.TarFile) # please mypy
archive.add(full_path, member.name, filter=TarParser._clean_member) # type: ignore
@staticmethod
def _get_all_members(archive: ArchiveClass) -> list[ArchiveMember]:
assert isinstance(archive, tarfile.TarFile) # please mypy
return archive.getmembers() # type: ignore
@staticmethod
def _get_member_name(member: ArchiveMember) -> str:
assert isinstance(member, tarfile.TarInfo) # please mypy
return member.name
@staticmethod
def _set_member_permissions(member: ArchiveMember, permissions: int) -> ArchiveMember:
assert isinstance(member, tarfile.TarInfo) # please mypy
member.mode = permissions
return member
class TarGzParser(TarParser):
compression = ':gz'
mimetypes = {'application/x-tar+gz'}
class TarBz2Parser(TarParser):
compression = ':bz2'
mimetypes = {'application/x-tar+bz2'}
class TarXzParser(TarParser):
compression = ':xz'
mimetypes = {'application/x-tar+xz'}
class ZipParser(ArchiveBasedAbstractParser):
mimetypes = {'application/zip'}
def __init__(self, filename):
super().__init__(filename)
self.archive_class = zipfile.ZipFile
self.member_class = zipfile.ZipInfo
def is_archive_valid(self):
try:
with zipfile.ZipFile(self.filename):
pass
except (zipfile.BadZipFile, OSError):
raise ValueError
@staticmethod
def _clean_member(member: ArchiveMember) -> ArchiveMember:
assert isinstance(member, zipfile.ZipInfo) # please mypy
member.create_system = 3 # Linux
member.comment = b''
member.date_time = (1980, 1, 1, 0, 0, 0) # this is as early as a zipfile can be
return member
@staticmethod
def _get_member_meta(member: ArchiveMember) -> dict[str, str]:
assert isinstance(member, zipfile.ZipInfo) # please mypy
metadata = {}
if member.create_system == 3: # this is Linux
pass
elif member.create_system == 2:
metadata['create_system'] = 'Windows'
else:
metadata['create_system'] = 'Weird'
if member.comment:
metadata['comment'] = member.comment # type: ignore
if member.date_time != (1980, 1, 1, 0, 0, 0):
metadata['date_time'] = str(datetime.datetime(*member.date_time))
return metadata
def _add_file_to_archive(self, archive: ArchiveClass, member: ArchiveMember,
full_path: str):
assert isinstance(archive, zipfile.ZipFile) # please mypy
assert isinstance(member, zipfile.ZipInfo) # please mypy
with open(full_path, 'rb') as f:
archive.writestr(member, f.read(),
compress_type=member.compress_type)
@staticmethod
def _get_all_members(archive: ArchiveClass) -> list[ArchiveMember]:
assert isinstance(archive, zipfile.ZipFile) # please mypy
return archive.infolist() # type: ignore
@staticmethod
def _get_member_name(member: ArchiveMember) -> str:
assert isinstance(member, zipfile.ZipInfo) # please mypy
return member.filename
@staticmethod
def _get_member_compression(member: ArchiveMember):
assert isinstance(member, zipfile.ZipInfo) # please mypy
return member.compress_type
@staticmethod
def _set_member_compression(member: ArchiveMember, compression) -> ArchiveMember:
assert isinstance(member, zipfile.ZipInfo) # please mypy
member.compress_type = compression
return member