File._get_disk_file: filenames with NUL / non-UTF-8 extension bytes raise ValueError/UnicodeDecodeError instead of FileError

Abierto Apto para principiantes
#375 0 comentarios 0 reacciones 0 asignados Ver en GitHub

Los mantenedores suelen responder en 1 día

Nadie ha tomado este issue todavía.

Evaluación

Dificultad
2/5
Tiempo estimado
1-3 horas
Aptitud para principiantes
85/100
Tipo de issue
Error
Claridad
Bien especificado
Estado de actividad
Activo
Stack tecnológico
python
Área
backend

Línea de trabajo

Empieza en python_multipart/multipart.py en torno a los tres sitios en las líneas 550, 560 y 574: lee File._get_disk_file junto con el contrato de excepciones en THREAT_MODEL.md. Ejecuta el reproducer del issue con UPLOAD_KEEP_EXTENSIONS=True para confirmar el ValueError que escapa, luego convierte ValueError/UnicodeDecodeError en esos sitios en FileError y añade tests para un nombre de archivo con NUL y una extensión no UTF-8. Hecho significa que los nombres malformados lanzan FileError mientras que los nombres de archivo benignos siguen parseándose.

Escrito por el modelo de indexación a partir del texto del issue.

Descripción

Hi! Filing this as a public bug per THREAT_MODEL.md — "What does not count" → "Report these as public issues if you think they are bugs." So: exception-contract bug on malformed input, not a vulnerability claim.

What happens

File._get_disk_file guards its sinks with except OSError (multipart.py:551, :575), and computes the tempfile suffix with a decode that sits outside any try (multipart.py:560). But the underlying calls can raise non-OSError exceptions on attacker-controlled filename= bytes:

Site Call Attacker input that raises
multipart.py:550 open(path, "w+b") ValueError: embedded null byte (filename contains NUL)
multipart.py:560 self._ext.decode(sys.getfilesystemencoding()) UnicodeDecodeError (extension bytes not valid UTF-8)
multipart.py:574 tempfile.NamedTemporaryFile(suffix=suffix, ...) ValueError: open: embedded null character in path (NUL in extension)

ValueError/UnicodeDecodeError are not OSError, so the documented contract — "Malformed input raises FormParserError or one of its subclasses" (Security goals in THREAT_MODEL.md) — is bypassed and the raw exception escapes FormParser/parse_form into the calling application.

Reachability: UPLOAD_KEEP_FILENAME=True (with UPLOAD_DIR) or UPLOAD_KEEP_EXTENSIONS=True, plus a part larger than MAX_MEMORY_FILE_SIZE (default 1 MiB) so flush_to_disk runs. The filename bytes survive parse_options_header (latin-1 roundtrip) and os.path.basename/os.path.splitext untouched, including NUL and \xff.

Reproducer

"""python-multipart 0.0.34 reproducer: NUL in filename= escapes the FileError contract."""
import sys
from python_multipart.exceptions import FileError
from python_multipart.multipart import FormParser

body = (
    b"--bnd\r\n"
    b'Content-Disposition: form-data; name="f"; filename="a.tx\x00t"\r\n'
    b"Content-Type: application/octet-stream\r\n\r\n"
    + b"A" * (1024 * 1024 + 16)  # > MAX_MEMORY_FILE_SIZE (1 MiB default) forces flush_to_disk
    + b"\r\n--bnd--\r\n"
)
parser = FormParser("multipart/form-data", lambda f: None, lambda f: None,
                    boundary="bnd", config={"UPLOAD_KEEP_EXTENSIONS": True})
try:
    parser.write(body)
    parser.finalize()
except FileError as exc:
    print("contract kept ->", type(exc).__name__, ":", exc); sys.exit(1)
except Exception as exc:
    print(f"BUG: {type(exc).__name__} escaped the FileError contract: {exc}"); sys.exit(0)
print("clean parse (no disk write?)"); sys.exit(1)
$ python3 -m venv v && ./v/bin/pip install python-multipart==0.0.34 && ./v/bin/python repro.py
BUG: ValueError escaped the FileError contract: open: embedded null character in path

Verified against the 0.0.34 PyPI wheel and the 0.0.34 tag source, on Python 3.14 (host) and 3.12 (python:3.12-slim container). The same three escape variants fire with UPLOAD_KEEP_FILENAME=True + UPLOAD_DIR (open() path).

Control: the same NUL filename with MAX_MEMORY_FILE_SIZE=float("inf") (no flush → no disk write) parses cleanly — so it is specifically the disk-write path that converts the NUL byte into an escaping exception.

Suggested fix

Catch the concrete non-OSError failure modes at all three sites and convert them to FileError, e.g.:

                 tmp_file = open(path, "w+b")
-            except OSError:
+            except (OSError, ValueError):

and in the else branch, move the suffix decode inside the existing try:

-            suffix = self._ext.decode(sys.getfilesystemencoding()) if keep_extensions else None
             ...
             try:
+                suffix = self._ext.decode(sys.getfilesystemencoding()) if keep_extensions else None
                 tmp_file = cast(BufferedRandom, tempfile.NamedTemporaryFile(suffix=suffix, delete=delete_tmp, dir=dir))
-            except OSError:
+            except (OSError, ValueError, UnicodeDecodeError):

With that change the reproducer flips to contract kept -> FileError : Error creating named temporary file, and a benign filename keeps parsing cleanly (verified on a patched build). Happy to open a PR if you'd like — the patch is 6 insertions / 7 deletions against 0.0.34.

Lenguaje dominante
Python
Estrellas
539
Forks
109
Merge medio
6 h 35 min
PR fusionados (30 d)
20

Preparar el entorno

Primeros pasos

  1. Lee el issue completo y luego la guía de contribución del proyecto.
  2. Comenta en el issue que vas a ocuparte — evita que dos personas hagan lo mismo.
  3. Haz un fork del repositorio y trabaja en una rama.
  4. Abre un pull request que haga referencia al número del issue.

Más de Kludex/python-multipart

Todos los issues de Kludex/python-multipart

Issues similares

Más issues de Python

Recibe los nuevos issues en tu correo

Un resumen breve de issues de GitHub para principiantes.