[KeyVault] ChallengeAuthPolicy keeps the first request and re-sends it in place of a later request on any 401 challenge
Maintainer antworten meist innerhalb von 1 Tag
Dieses Issue hat noch niemand übernommen.
Bewertung
- Schwierigkeit
- 3/5
- Geschätzter Aufwand
- 1-2 Tage
- Anfängerfreundlichkeit
- 76/100
Rechercherichtung
Read azure/keyvault/certificates/_shared/async_challenge_auth_policy.py and its sync twin, challenge_auth_policy.py, focusing on the request copy and challenge handling described in the issue. Add a regression test using the offline recording transport scenario, then run the relevant Key Vault certificate policy tests. Done means a later challenged GET is retried as that GET, while the saved body request is used only for its own challenge round.
Vom Indexierungsmodell aus dem Issue-Text verfasst.
Beschreibung
Package versions: azure-keyvault-certificates 4.11.2, azure-core 1.40.0, Python 3.14.6 (also read in the sync policy).
Describe the bug
The Key Vault challenge policy (azure/keyvault/certificates/_shared/async_challenge_auth_policy.py; the sync twin challenge_auth_policy.py has the same logic) keeps a copy of the first body-carrying request it sees while it has no cached challenge, and never clears it:
self._request_copy = None(async :84; sync :95)on_request: with no cached challenge and a request that has content,self._request_copy = request.http_request(async :179; sync :187), and a bodiless copy is sent to get the challengeon_challenge:if self._request_copy: request.http_request = self._request_copy(async :224-225; sync :232-233)
on_challenge runs on any later 401 that carries WWW-Authenticate, not only the first. So on a long-lived client, a 401 on an unrelated later request replaces that request with the client's very first request and re-sends it with a fresh token. For a certificate client, that is a second POST .../create for a certificate the caller did not ask about (a new certificate version), and the request the caller made is never retried.
To Reproduce (offline: a recording transport and a fake credential, no network)
"""Offline repro: azure-keyvault-certificates' async challenge policy re-sends an earlier request on a later 401.
No network: a recording transport answers from a script, and a fake credential returns a token.
"""
import asyncio
import time
from azure.core.credentials import AccessToken
from azure.core.pipeline.transport import AsyncHttpTransport
from azure.core.rest._http_response_impl_async import AsyncHttpResponseImpl
from azure.core.utils import CaseInsensitiveDict
from azure.keyvault.certificates import CertificatePolicy
from azure.keyvault.certificates.aio import CertificateClient
VAULT_URL = "https://example-vault.vault.azure.net"
CHALLENGE = 'Bearer authorization="https://login.microsoftonline.com/00000000-0000-0000-0000-000000000000", resource="https://vault.azure.net"'
class _Internal:
async def close(self):
return None
class RecordingTransport(AsyncHttpTransport):
"""401 to any request without a bearer token; 401 once to a bearer GET when reject_next_get is set; else 200/202."""
def __init__(self):
self.seen = []
self.reject_next_get = False
async def open(self): return None
async def close(self): return None
async def __aenter__(self): return self
async def __aexit__(self, *exc): return None
async def send(self, request, **kwargs):
content = request.content
body = content.decode() if isinstance(content, bytes) else (content or "")
authorized = "Authorization" in request.headers
path = request.url[len(VAULT_URL):].split("?")[0]
self.seen.append((request.method, path, "body" if body else "no body", "auth" if authorized else "no auth"))
name = path.split("/")[2]
headers = CaseInsensitiveDict({"Content-Type": "application/json"})
status = 202 if request.method == "POST" else 200
payload = ('{"id": "%s/certificates/%s/pending", "status": "inProgress", "issuer": {"name": "Unknown"}, "request_id": "r"}' % (VAULT_URL, name)).encode()
if not authorized or (request.method == "GET" and self.reject_next_get):
if authorized:
self.reject_next_get = False
status, payload = 401, b'{"error": {"code": "Unauthorized"}}'
headers["WWW-Authenticate"] = CHALLENGE
response = AsyncHttpResponseImpl(request=request, internal_response=_Internal(), block_size=None, status_code=status,
reason="mock", content_type="application/json", headers=headers, stream_download_generator=None)
response._content = payload
return response
class FakeCredential:
async def get_token(self, *scopes, **kwargs):
return AccessToken("fake-token", int(time.time()) + 3600)
async def close(self):
return None
async def main():
transport = RecordingTransport()
async with CertificateClient(vault_url=VAULT_URL, credential=FakeCredential(), transport=transport) as client:
# 1. The client's first request carries a body: the challenge policy keeps a copy of it.
await client.create_certificate(certificate_name="cert-a", policy=CertificatePolicy(issuer_name="Self", subject="CN=cert-a"))
transport.seen.clear()
# 2. Later, an unrelated GET gets one 401 with a challenge (e.g. a token the service no longer accepts).
transport.reject_next_get = True
try:
await client.get_certificate_operation(certificate_name="cert-b")
outcome = "returned"
except Exception as exc: # noqa: BLE001
outcome = f"raised {type(exc).__name__}"
print("requests sent after the 401 on GET cert-b:")
for entry in transport.seen:
print(" ", *entry)
print("get_certificate_operation(cert-b):", outcome)
asyncio.run(main())
Output:
requests sent after the 401 on GET cert-b:
GET /certificates/cert-b/pending no body auth
POST /certificates/cert-a/create body auth
get_certificate_operation(cert-b): raised HttpResponseError
Expected behavior
After the 401 on GET /certificates/cert-b/pending, the policy re-sends that GET with a fresh token. The kept copy should be used only to complete the challenge round of the request it was taken from, and then cleared (or keyed to that request).
Additional context
It only needs a 401 with a WWW-Authenticate header after the client's first request carried a body, for example a token the service stops accepting while the client lives. A process-wide client (the documented pattern) is the case where this bites. We work around it locally by priming the challenge with a bodiless request first and refusing to send a request other than the one the caller made.
- Vorherrschende Sprache
- Python
- Sterne
- 5.6k
- Forks
- 3.4k
- Ø Merge
- 2 T. 12 Std.
- Gemergte PRs (30 T.)
- 204
Entwicklungsumgebung
Startet den Dev-Container des Projekts im Browser, mit Ihrem eigenen GitHub-Konto.
- Kein Dockerfile und keine Docker-Compose-Datei
- Hat eine Pull-Request-Vorlage
- Beitragsleitfaden lesen
Erste Schritte
- Lesen Sie das ganze Issue und danach den Beitragsleitfaden des Projekts.
- Schreiben Sie ins Issue, dass Sie es übernehmen — das erspart doppelte Arbeit.
- Forken Sie das Repository und arbeiten Sie in einem Branch.
- Öffnen Sie einen Pull Request, der die Issue-Nummer nennt.
Mehr aus Azure/azure-sdk-for-python
-
[Teams Chatbot]: next-mypy install 404s should point to CFS auth or --pypi workaroundEvtl. vergeben @jianingwang123 hat das vor 4 Tagen übernommen. Offenneeds-triage
Schwierigkeit 1/5 1-3 Stunden Anfängerfreundlichkeit 72/100
Azure/azure-sdk-for-python#49886 · 1 Kommentar · 2 Reaktionen · 2 zugewiesene Personen ·
Maintainer antworten meist innerhalb von 1 Tag
-
[AI Projects] AIProjectInstrumentor: non-streamed responses.create raises AttributeError ('NonRecordingSpan' object has no attribute 'attributes') when the span is not sampledEvtl. vergeben @14H034160212 hat das vor 6 Tagen übernommen. OffenAI Projects Client customer-reported question
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 78/100
Azure/azure-sdk-for-python#49854 · 2 Kommentare · 1 Reaktion ·
Maintainer antworten meist innerhalb von 1 Tag
-
azure-eventhub migration guide: legacy checkpoint sample writes "None" metadata for partitions that were never checkpointedEvtl. vergeben @hidingmyname hat das vor 14 Tagen übernommen. OffenClient customer-reported Event Hubs needs-team-attention question Service Attention
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 78/100
Azure/azure-sdk-for-python#49723 · 2 Kommentare · 1 Reaktion ·
Maintainer antworten meist innerhalb von 1 Tag
-
Scheduled test pipelines fail at the Analyze_Test stage due to date formattingEvtl. vergeben Ein verknüpfter Pull Request ist offen oder bereits gemergt. OffenEngSys
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 78/100
Azure/azure-sdk-for-python#49614 · 1 Kommentar · 1 Reaktion ·
Maintainer antworten meist innerhalb von 1 Tag
-
[Teams Chatbot]: Clarify Python dev-feed publication from a branch versus PyPI releaseEvtl. vergeben @jianingwang123 hat das vor 29 Tagen übernommen. Offencustomer-reported question
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 62/100
Azure/azure-sdk-for-python#49548 · 1 Kommentar · 2 Reaktionen · 2 zugewiesene Personen ·
Maintainer antworten meist innerhalb von 1 Tag
Alle Issues in Azure/azure-sdk-for-python
Ähnliche Issues
-
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 72/100
Maintainer antworten meist innerhalb von 1 Tag
-
AI bug
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 68/100
AUTO-MAS-Project/AUTO-MAS#1960 · 1 Kommentar ·
Maintainer antworten meist innerhalb von 1 Tag
-
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 82/100
btclib-org/btclib-wallet#872 ·
Maintainer antworten meist innerhalb von 1 Tag
-
[Bug]: Scheduled returns scan all accepted minion keys causing extreme loadEvtl. vergeben Ein verknüpfter Pull Request ist offen oder bereits gemergt. Offenbug needs-triage
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 76/100
saltstack/salt#70789 · 1 Kommentar ·
Maintainer antworten meist innerhalb von 3 Tagen
-
Schwierigkeit 2/5 1-3 Stunden Anfängerfreundlichkeit 87/100
btclib-org/btclib#3093 ·
Maintainer antworten meist innerhalb von 1 Tag