Files
mcp-server/spark_executor/tools/fetch_url.py
T
ClaudeandClaude Fable 5 a5b9539663 refactor(fetch_url): trust the allowlist, rename to url_allowlist
Two changes:

1) Drop every check except the allowlist lookup.

  Old _validate_url_host did: scheme check, host-presence check,
  IP-literal check, empty-allowlist check, then glob match.

  New _validate_url_host does: parse host, return on glob match,
  raise on miss. That's it. The only remaining structural check is
  'the URL must have a host' (otherwise the glob has nothing to
  test against).

  Security implication: scheme (file://, gopher://, ftp://) and
  IP literals (10.0.0.1, ::1) are NO LONGER rejected by the
  validator. The allowlist is the single source of truth. If the
  user writes ['*.*.*.*'], they have opted in to 4-label hosts
  including IP literals; if they write ['ccam*'], they get ccam1-
  ccam99 and nothing else. The default ['ccam*'] / [] pattern is
  tight by construction.

  Removed: import ipaddress, the scheme/IP rejection branches, the
  'allowlist empty' explicit branch (the empty list naturally
  matches nothing).

2) Rename allowed_url_hosts -> url_allowlist.

  The previous name was a verbose double-negative ('allowed ... hosts').
  The new name is short, modern (allowlist > whitelist), and matches
  the pattern of the field (URL hosts allowed). Renamed in:
    - Connection (models.py)
    - SaveConnectionRequest, UpdateConnectionRequest, FetchUrlRequest
      (requests.py)
    - _validate_url_host, _host_matches_any_glob parameters
      (fetch_url.py)
    - save_connection / update_connection call sites and error
      messages (connections.py, fetch_url.py)
    - Route descriptions (server.py)
    - All test files
    - README

  No backward-compat alias: the field was added in 0291f36 and
  hasn't shipped, so no production migration. Local dev data
  (data/connections.json, gitignored) with allowed_url_hosts set
  will be silently dropped by Pydantic v2 (default for extra
  fields is ignore) — those connections lose their allowlist and
  fetch_url will reject everything until re-saved.

Test cleanup:
  - Removed 9 obsolete tests (scheme/IP/suffix rejection)
  - Renamed allowed_url_hosts -> url_allowlist in 11 surviving tests
  - Added 4 new tests documenting the 'allowlist is the only gate'
    model: IP literal accepted, HTTPS accepted, no-host rejected,
    empty allowlist rejected, error message mentions url_allowlist

  -1 obsolete test, net -4 from 386 -> 382 tests passing.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-09 12:09:41 +08:00

105 lines
3.6 KiB
Python

# coding=utf-8
"""
@Time :2026/7/9
@Author :tao.chen
Generic HTTP GET proxy for the agent. Lets the agent fetch URLs on the
cluster's network when it cannot reach those hosts directly. Security:
the host is checked against an explicit fnmatch glob allowlist configured
on the connection (`url_allowlist`). Empty or missing allowlist means no
URL access; the allowlist is the only gate — scheme and IP-literal checks
are intentionally NOT performed. Reuses the connection's saved auth/SSL
config so the agent doesn't need cluster credentials.
"""
import fnmatch
from urllib.parse import urlparse
import httpx
from common.logging import logger
from spark_executor.core.yarn_client import YarnClientConfig
from spark_executor.models import FetchUrlResult
from spark_executor.tools.connections import store as conn_store
_MAX_BODY_BYTES = 1_000_000 # 1 MB cap on response body
_REQUEST_TIMEOUT_SECONDS = 30
def _host_matches_any_glob(host: str, allowlist: list[str]) -> bool:
"""True if `host` matches any of the fnmatch glob patterns.
fnmatch is case-sensitive on Linux (our deployment target). `*` in a
pattern does NOT cross '.' boundaries, so 'ccam*' matches 'ccam50' but
NOT 'ccam50.evil.com' (which is what we want — domain boundary matters
for SSRF).
"""
host_labels = host.split(".")
for p in allowlist:
pat_labels = p.split(".")
if len(pat_labels) != len(host_labels):
continue
if all(fnmatch.fnmatchcase(h, pat) for h, pat in zip(host_labels, pat_labels)):
return True
return False
def _validate_url_host(url: str, allowlist: list[str] | None) -> None:
"""Reject the URL unless its host matches a glob in `allowlist`.
The only check. The url_allowlist is the single source of truth for
what fetch_url is allowed to access — no scheme, IP-literal, or
"must be the same as yarn_rm_url" guardrails. The caller is
responsible for writing a tight allowlist.
The only structural check: the URL must have a host (otherwise
the glob match has nothing to test). Anything else is the
allowlist's job.
"""
host = urlparse(url).hostname
if not host:
raise ValueError(f"URL has no host: {url!r}")
if _host_matches_any_glob(host, allowlist or []):
return
raise ValueError(
f"URL host {host!r} is not in Connection.url_allowlist {allowlist!r}. "
f"Add the host pattern to url_allowlist (or use a broader glob) "
f"and try again."
)
def fetch_url(url: str, connection_name: str) -> FetchUrlResult:
"""Proxy an HTTP GET to url using the auth/SSL settings of connection_name."""
logger.debug(f"fetch_url enter url={url} connection_name={connection_name}")
conn = conn_store.get(connection_name)
if conn is None:
raise KeyError(f"Connection not found: {connection_name}")
_validate_url_host(url, conn.url_allowlist)
config = YarnClientConfig.from_connection(conn)
resp = httpx.get(
url,
auth=config.auth_for_httpx(),
verify=config.verify_for_httpx(),
timeout=_REQUEST_TIMEOUT_SECONDS,
follow_redirects=True,
)
body = resp.text[:_MAX_BODY_BYTES]
truncated = len(resp.text) > _MAX_BODY_BYTES
logger.info(
f"fetch_url ok url={url} connection_name={connection_name} "
f"status_code={resp.status_code} "
f"content_type={resp.headers.get('content-type', '')} "
f"body_bytes={len(body)} truncated={truncated}"
)
return FetchUrlResult(
url=url,
status_code=resp.status_code,
content_type=resp.headers.get("content-type", ""),
body=body,
truncated=truncated,
)