Download and copy sync actions print remote file names without escape_control_chars

Ouverte Adaptée aux débutants
  • Quelque chose ne va pas ?
#631 0 commentaires 0 réactions 0 personnes assignées Voir sur GitHub

Les mainteneurs répondent en général sous 1 jour

@dreyfus92 y travaille déjà.

Depuis le 30/10/2026.

  • #636 par @dreyfus92 — ouverte

Évaluation

Difficulté
2/5
Temps estimé
1-3 heures
Accessibilité débutants
72/100
Type d'issue
Bug
Clarté
Clairement spécifiée
Activité
Active
Stack technique
python
Domaine
security

Piste de recherche

Lisez b2sdk/_internal/sync/action.py, en commençant par les appels de fin aux lignes 343 et 433 et le chemin d’erreur partagé à la ligne 60 ; comparez-les aux appels qui échappent déjà les caractères. Examinez ProgressReport._print_line dans scan/report.py pour comprendre le chemin de sortie. Le travail est terminé lorsque les sorties de téléchargement, de copie et d’erreur échappent les caractères de contrôle, et que des tests confirment que la sortie du terminal ne contient aucun caractère de contrôle brut.

Rédigé par le modèle d'indexation à partir du texte de l'issue.

Description

Description

Four of the six completion call sites in b2sdk/_internal/sync/action.py pass the remote name through escape_control_chars first, at :179 (upload), :259 (hide), :482 and :523 (delete). The download and copy sites concatenate it raw: reporter.print_completion('dnload ' + self.source_path.relative_path) at :343 and reporter.print_completion('copy ' + self.source_path.relative_path) at :433. The shared error path at :60 is unescaped too, and str(self) there embeds the raw b2_file_name via B2DownloadAction.__str__ at :346. From any of those the string reaches ProgressReport._print_line (scan/report.py:144-155), whose own docstring states the precondition the callers violate, ":param line: a string without a \r or \n in it", and which writes it straight out with self.stdout.write(line). Nothing on the inbound path strips control characters: validate_b2_file_name_as_path covers relative and absolute path components only, and RawApi.check_b2_filename (raw_api.py:991) does reject code points below 32 but has exactly one caller, the upload path at raw_api.py:1059, so it never runs on a name the SDK receives. The effect is that a file name from the endpoint is written to the user's terminal with its control characters intact and the terminal acts on them, which covers title spoofing, colour changes that persist for everything printed afterwards, screen clears and more on emulators that still implement them; all ten control characters tested reached stdout raw. Worth being clear about the limit: a well-behaved Backblaze bucket cannot hold such a name, since the documentation states that "character codes below U+0020 (decimal 32) are not allowed" and "DEL characters (127) are not allowed" and the service enforces that on upload independently of the client, so reaching this needs a B2-compatible endpoint under the attacker's control via a custom realm, an active TLS interception, or a future server-side regression. That is why I am filing it in the open rather than privately.

Reproduction

On master @ ffa02b32 with Python 3.10.19, in a container built from a fresh clone at that ref. RawSimulator is the SDK's own public test double, exported from b2sdk.v2 and b2sdk.v3, and it stands in for exactly one thing here, the endpoint's own name validation, which is the single element the real service would supply. Build with this Dockerfile, save the script as poc.py, then run docker build -t b2poc . && docker run --rm -u 1001 -v "$PWD/poc.py:/t.py:ro" b2poc python /t.py; replacing SyncReport(out, True) with SyncReport(sys.stdout, True) sends the same bytes to a real terminal:

FROM python:3.10.19-slim-bookworm
RUN apt-get update -qq && apt-get install -y -qq --no-install-recommends git >/dev/null
RUN git clone -q https://github.com/Backblaze/b2-sdk-python.git /src \
 && git -C /src checkout -q ffa02b3264bc0e7069ef2c8b25facd9a467e65a6 \
 && pip install --no-cache-dir -q /src
RUN useradd -m -u 1001 victim

import io, os, time
from b2sdk.v3 import (B2Api, B2HttpApiConfig, RawSimulator, StubAccountInfo,
                      Synchronizer, SyncReport, B2Folder, LocalFolder)

NAME = '\x1b]0;PWNED\x07\x1b[31minvoice.pdf'      # a name the endpoint returns in b2_list_file_versions
api = B2Api(StubAccountInfo(), api_config=B2HttpApiConfig(_raw_api_class=RawSimulator))
acct, key = api.session.raw_api.create_account()
api.authorize_account(application_key_id=acct, application_key=key, realm='production')
bucket = api.create_bucket('evil', 'allPublic')
bucket.upload_bytes(b'x', NAME)
bucket.upload_bytes(b'x', 'ordinary.txt')

os.makedirs('/tmp/dest', exist_ok=True)
out = io.StringIO()
Synchronizer(max_workers=1).sync_folders(source_folder=B2Folder('evil', '', api),
    dest_folder=LocalFolder('/tmp/dest'), now_millis=int(time.time() * 1000),
    reporter=SyncReport(out, True))
for line in out.getvalue().split('\n'):
    if line.strip():
        print('stdout got:', repr(line.rstrip()))

stdout got: 'dnload \x1b]0;PWNED\x07\x1b[31minvoice.pdf'
stdout got: 'dnload ordinary.txt'
Suggested remediation

Wrap the two names and the error path, matching the four sites that already do:

--- a/b2sdk/_internal/sync/action.py
+++ b/b2sdk/_internal/sync/action.py
@@ -57,7 +57,9 @@ class AbstractAction(metaclass=ABCMeta):
             self.do_report(bucket, reporter)
         except Exception as e:
             logger.exception('an exception occurred in a sync action')
-            reporter.error(str(self) + ': ' + repr(e) + ' ' + str(e))
+            reporter.error(
+                f'{escape_control_chars(str(self))}: {e!r} {escape_control_chars(str(e))}'
+            )
             raise  # Re-throw so we can identify failed actions
@@ -340,7 +342,7 @@ class B2DownloadAction(AbstractAction):
-        reporter.print_completion('dnload ' + self.source_path.relative_path)
+        reporter.print_completion(f'dnload {escape_control_chars(self.source_path.relative_path)}')
@@ -430,7 +432,7 @@ class B2CopyAction(AbstractAction):
-        reporter.print_completion('copy ' + self.source_path.relative_path)
+        reporter.print_completion(f'copy {escape_control_chars(self.source_path.relative_path)}')

repr(e) is deliberately left unescaped, written as {e!r}, because Python's repr of a string already renders control characters as escapes and it was measured not to leak ESC, whereas str(e) does leak it and so is wrapped.

Credits
  • Found by OSTIF using AI with their own harness.
  • Manually verified, fixed, and reported by Shielder.
Langage dominant
Python
Étoiles
214
Forks
68
Merge moyen
51 min
PR mergées (30 j)
6

Préparer son environnement

Par où commencer

  1. Lisez l'issue en entier, puis le guide de contribution du projet.
  2. Signalez en commentaire que vous la prenez — cela évite que deux personnes fassent le même travail.
  3. Forkez le dépôt et travaillez sur une branche.
  4. Ouvrez une pull request qui référence le numéro de l'issue.

Autres issues de Backblaze/b2-sdk-python

Toutes les issues de Backblaze/b2-sdk-python

Issues similaires

Plus d'issues Python

Recevez les nouvelles issues par e-mail

Un résumé court des issues GitHub adaptées aux débutants.