Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

TemporaryDirectory clean-up fails with unsearchable directories #79325

Closed
lilydjwg mannequin opened this issue Nov 2, 2018 · 10 comments
Closed

TemporaryDirectory clean-up fails with unsearchable directories #79325

lilydjwg mannequin opened this issue Nov 2, 2018 · 10 comments
Assignees
Labels
3.8 only security fixes 3.9 only security fixes 3.10 only security fixes stdlib Python modules in the Lib dir type-bug An unexpected behavior, bug, or error

Comments

@lilydjwg
Copy link
Mannequin

lilydjwg mannequin commented Nov 2, 2018

BPO 35144
Nosy @lilydjwg, @serhiy-storchaka, @eryksun, @vidartf, @tirkarthi
PRs
  • bpo-26660, bpo-35144: Fix permission errors in TemporaryDirectory cleanup. #10320
  • Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.

    Show more details

    GitHub fields:

    assignee = 'https://github.com/serhiy-storchaka'
    closed_at = None
    created_at = <Date 2018-11-02.08:40:08.637>
    labels = ['3.8', 'type-bug', 'library', '3.9', '3.10']
    title = 'TemporaryDirectory clean-up fails with unsearchable directories'
    updated_at = <Date 2021-03-28.07:31:28.296>
    user = 'https://github.com/lilydjwg'

    bugs.python.org fields:

    activity = <Date 2021-03-28.07:31:28.296>
    actor = 'eryksun'
    assignee = 'serhiy.storchaka'
    closed = False
    closed_date = None
    closer = None
    components = ['Library (Lib)']
    creation = <Date 2018-11-02.08:40:08.637>
    creator = 'lilydjwg'
    dependencies = []
    files = []
    hgrepos = []
    issue_num = 35144
    keywords = ['patch']
    message_count = 9.0
    messages = ['329116', '329127', '329184', '329185', '344036', '377309', '377316', '377329', '377358']
    nosy_count = 5.0
    nosy_names = ['lilydjwg', 'serhiy.storchaka', 'eryksun', 'vidartf', 'xtreak']
    pr_nums = ['10320']
    priority = 'normal'
    resolution = None
    stage = 'patch review'
    status = 'open'
    superseder = None
    type = 'behavior'
    url = 'https://bugs.python.org/issue35144'
    versions = ['Python 3.8', 'Python 3.9', 'Python 3.10']

    Linked PRs

    @lilydjwg
    Copy link
    Mannequin Author

    lilydjwg mannequin commented Nov 2, 2018

    If the title doesn't explain clearly, here's a demo program that will fail:

    import tempfile
    import pathlib
    
    def test():
      with tempfile.TemporaryDirectory(prefix='test-bad-') as tmpdir:
        tmpdir = pathlib.Path(tmpdir)
        subdir = tmpdir / 'sub'
        subdir.mkdir()
        with open(subdir / 'file', 'w'):
          pass
        subdir.chmod(0o600)
    
    if __name__ == '__main__':
      test()

    I didn't expect this, and I didn't find an easy way to handle this except not using TemporaryDirectory at all:

    import tempfile
    import pathlib
    import shutil
    import os
    
    def rmtree_error(func, path, excinfo):
      if isinstance(excinfo[1], PermissionError):
        os.chmod(os.path.dirname(path), 0o700)
        os.unlink(path)
      print(func, path, excinfo)
    
    def test():
      tmpdir = tempfile.mkdtemp(prefix='test-good-')
      try:
        tmpdir = pathlib.Path(tmpdir)
        subdir = tmpdir / 'sub'
        subdir.mkdir()
        with open(subdir / 'file', 'w'):
          pass
        subdir.chmod(0o600)
      finally:
        shutil.rmtree(tmpdir, onerror=rmtree_error)
    
    if __name__ == '__main__':
      test()

    This works around the issue, but the dirfd is missing in the onerror callback.

    I have this issue because my program extracts tarballs to a temporary directory for examination. I expected that TemporaryDirectory cleaned up things when it could.

    What do you think? rm -rf can't remove such a directory either but this is annoying and I think Python can do better.

    @lilydjwg lilydjwg mannequin added 3.7 (EOL) end of life stdlib Python modules in the Lib dir type-bug An unexpected behavior, bug, or error labels Nov 2, 2018
    @tirkarthi
    Copy link
    Member

    Seems like a related issue : bpo-26660 . Maybe TemporaryDirectory can allow an onerror argument that is passed internally to rmtree during cleanup and state the same in the documentation that TemporaryDirectory can't cleanup read-only files?

    @serhiy-storchaka serhiy-storchaka self-assigned this Nov 2, 2018
    @terryjreedy terryjreedy changed the title TemporaryDirectory can't be cleaned up if there are unsearchable directories TemporaryDirectory clean-up fails with unsearchable directories Nov 2, 2018
    @lilydjwg
    Copy link
    Mannequin Author

    lilydjwg mannequin commented Nov 3, 2018

    Yes bpo-26660 is similar but not the same. On Windows it seems a read-only file cannot be deleted while on Linux a file resident in a non-searchable directory cannot be deleted.

    An onerror for TemporaryDirectory will work. Also I'd like to add an optional dir_fd argument for onerror.

    @serhiy-storchaka
    Copy link
    Member

    I am working on this. Left to test on Windows and analyze possible security issues.

    @serhiy-storchaka
    Copy link
    Member

    New changeset e9b51c0 by Serhiy Storchaka in branch 'master':
    bpo-26660, bpo-35144: Fix permission errors in TemporaryDirectory cleanup. (GH-10320)
    e9b51c0

    @vidartf
    Copy link
    Mannequin

    vidartf mannequin commented Sep 22, 2020

    On Python 3.8.5 on Windows using the code from the above patch I recently got a stack overflow:

    Thread 0x00002054 (most recent call first):
    File "...\lib\concurrent\futures\thread.py", line 78 in _worker
    File "...\lib\threading.py", line 870 in run
    File "...\lib\threading.py", line 932 in _bootstrap_inner
    File "...\lib\threading.py", line 890 in _bootstrap

    Thread 0x00000de4 (most recent call first):
    File "...\lib\concurrent\futures\thread.py", line 78 in _worker
    File "...\lib\threading.py", line 870 in run
    File "...\lib\threading.py", line 932 in _bootstrap_inner
    File "...\lib\threading.py", line 890 in _bootstrap

    Current thread 0x00004700 (most recent call first):
    File "...\lib\tempfile.py", line 803 in onerror
    File "...\lib\shutil.py", line 619 in _rmtree_unsafe
    File "...\lib\shutil.py", line 737 in rmtree
    File "...\lib\tempfile.py", line 814 in _rmtree
    File "...\lib\tempfile.py", line 806 in onerror
    File "...\lib\shutil.py", line 619 in _rmtree_unsafe
    File "...\lib\shutil.py", line 737 in rmtree
    ... repeating

    -------------------------------------------

    In my case, the outer exc_info from rmtree is:

    PermissionError(13, 'The process cannot access the file because it is being used by another process')

    And the inner exception from _os.unlink(path) is:

    PermissionError(13, 'Access is denied')

    I would say that expected behavior in this case would be to let the 'file is in use' error raise, instead of killing the process with an SO.

    @serhiy-storchaka
    Copy link
    Member

    Thank you Vidar! I wasn't sure about Windows, but was not able to reproduce possible failures. Your report gives a clue.

    @vidartf
    Copy link
    Mannequin

    vidartf mannequin commented Sep 22, 2020

    A somewhat easy repro:

    Create the temporary directory, add a subdir (not sure if subdir truly necessary at this point), use os.chdir() to set the cwd to that subdir. Clean up the temp dir. The cwd should prevent the deletion because it will be "in use".

    @eryksun
    Copy link
    Contributor

    eryksun commented Sep 23, 2020

    It seems to me that if path == name, then resetperms(path) and possibly a recursive call are only needed on the first call. In subsequent calls, if path == name, then we know that resetperms(path) was already called, so it shouldn't handle PermissionError. If resetperms was ineffective (e.g. in Windows, a sharing violation or custom discretionary/mandatory permissions), or if something else changed the permissions in the mean time, just give up instead of risking a RecursionError or stack overflow. For example:

        @classmethod
        def _rmtree(cls, name, first_call=True):
            resetperms_funcs = (_os.unlink, _os.rmdir, _os.scandir, _os.open)
    
            def resetperms(path):
                try:
                    _os.chflags(path, 0)
                except AttributeError:
                    pass
                _os.chmod(path, 0o700)
    
            def onerror(func, path, exc_info):
                if (issubclass(exc_info[0], PermissionError) and
                      func in resetperms_funcs and (first_call or path != name)):
                    try:
                        if path != name:
                            resetperms(_os.path.dirname(path))
                        resetperms(path)
                        try:
                            _os.unlink(path)
                        # PermissionError is raised on FreeBSD for directories
                        except (IsADirectoryError, PermissionError):
                            cls._rmtree(path, first_call=False)
                    except FileNotFoundError:
                        pass
                elif issubclass(exc_info[0], FileNotFoundError):
                    pass
                else:
                    raise
    
            _shutil.rmtree(name, onerror=onerror)

    @serhiy-storchaka
    Copy link
    Member

    I implemented this idea in #112762.

    miss-islington pushed a commit to miss-islington/cpython that referenced this issue Dec 7, 2023
    …Windows (pythonGH-112762)
    
    (cherry picked from commit b2923a6)
    
    Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
    serhiy-storchaka added a commit to serhiy-storchaka/cpython that referenced this issue Dec 7, 2023
    …nup on Windows (pythonGH-112762)
    
    (cherry picked from commit b2923a6)
    
    Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
    serhiy-storchaka added a commit that referenced this issue Dec 7, 2023
    … Windows (GH-112762) (GH-112847)
    
    (cherry picked from commit b2923a6)
    
    Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
    serhiy-storchaka added a commit that referenced this issue Dec 7, 2023
    aisk pushed a commit to aisk/cpython that referenced this issue Feb 11, 2024
    Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
    Labels
    3.8 only security fixes 3.9 only security fixes 3.10 only security fixes stdlib Python modules in the Lib dir type-bug An unexpected behavior, bug, or error
    Projects
    Status: Done
    Development

    No branches or pull requests

    3 participants