Files
DECNET/decnet/sniffer/worker.py
anti 7f497ac552 feat(sniffer): publish decky.{id}.traffic on the bus (DEBT-031)
SnifferEngine gains an optional publish_fn hook, invoked after the
dedup + syslog write for traffic-summary events only (tls_session,
tcp_flow_timing, tcp_syn_fingerprint) — intermediate parser artifacts
like tls_client_hello stay off the bus.

The sniffer worker wires get_bus() + a thread-safe shim that marshals
sync calls from the scapy sniff thread back onto the asyncio loop via
run_coroutine_threadsafe.  Bus failure at startup degrades cleanly to
publish-off mode; publish failures at runtime never escape the sniff
thread.
2026-04-21 16:35:50 -04:00

231 lines
7.8 KiB
Python

"""
Fleet-wide MACVLAN sniffer worker.
Runs as a single host-side async background task that sniffs all TLS
traffic on the MACVLAN host interface. Maps packets to deckies by IP
and feeds fingerprint events into the existing log pipeline.
Modeled on decnet.collector.worker — same lifecycle pattern.
Fault-isolated: any exception is logged and the worker exits cleanly.
The API never depends on this worker being alive.
"""
import asyncio
import contextlib
import os
import subprocess # nosec B404 — needed for interface checks
import threading
from concurrent.futures import ThreadPoolExecutor
from pathlib import Path
from typing import Any, Callable
from decnet.bus import topics as _topics
from decnet.bus.base import BaseBus
from decnet.bus.factory import get_bus
from decnet.bus.publish import publish_safely
from decnet.logging import get_logger
from decnet.network import HOST_IPVLAN_IFACE, HOST_MACVLAN_IFACE
from decnet.sniffer.fingerprint import SnifferEngine
from decnet.sniffer.syslog import write_event
from decnet.telemetry import traced as _traced
logger = get_logger("sniffer")
_IP_MAP_REFRESH_INTERVAL: float = 60.0
def _load_ip_to_decky() -> dict[str, str]:
"""Build IP → decky-name mapping from decnet-state.json."""
from decnet.config import load_state
state = load_state()
if state is None:
return {}
config, _ = state
mapping: dict[str, str] = {}
for decky in config.deckies:
mapping[decky.ip] = decky.name
return mapping
def _make_thread_safe_publisher(
bus: BaseBus,
loop: asyncio.AbstractEventLoop,
) -> Callable[[str, str, dict[str, Any]], None]:
"""Build a sync callable that marshals bus publishes back to *loop*.
The scapy sniff loop runs in a dedicated worker thread and cannot
``await`` anything. Every call here schedules the async publish on
the event loop and returns immediately; the sniff thread is never
blocked waiting for the publish to actually land on the wire.
"""
def _publish(decky_name: str, event_type: str, payload: dict[str, Any]) -> None:
topic = _topics.decky(decky_name, _topics.DECKY_TRAFFIC)
try:
asyncio.run_coroutine_threadsafe(
publish_safely(bus, topic, payload, event_type=event_type),
loop,
)
except Exception as exc: # noqa: BLE001
logger.debug("sniffer: cross-thread bus publish failed: %s", exc)
return _publish
def _interface_exists(iface: str) -> bool:
"""Check if a network interface exists on this host."""
try:
result = subprocess.run( # nosec B603 B607 — hardcoded args
["ip", "link", "show", iface],
capture_output=True, text=True, check=False,
)
return result.returncode == 0
except Exception:
return False
@_traced("sniffer.sniff_loop")
def _sniff_loop(
interface: str,
log_path: Path,
json_path: Path,
stop_event: threading.Event,
publish_fn: Callable[[str, str, dict[str, Any]], None] | None = None,
) -> None:
"""Blocking sniff loop. Runs in a dedicated thread via asyncio.to_thread."""
try:
from scapy.sendrecv import sniff
except ImportError:
logger.error("scapy not installed — sniffer cannot start")
return
ip_map = _load_ip_to_decky()
if not ip_map:
logger.warning("sniffer: no deckies in state — nothing to sniff")
return
def _write_fn(line: str) -> None:
write_event(line, log_path, json_path)
engine = SnifferEngine(
ip_to_decky=ip_map, write_fn=_write_fn, publish_fn=publish_fn,
)
# Periodically refresh IP map in a background daemon thread
def _refresh_loop() -> None:
while not stop_event.is_set():
stop_event.wait(_IP_MAP_REFRESH_INTERVAL)
if stop_event.is_set():
break
try:
new_map = _load_ip_to_decky()
if new_map:
engine.update_ip_map(new_map)
except Exception as exc:
logger.debug("sniffer: ip map refresh failed: %s", exc)
refresh_thread = threading.Thread(target=_refresh_loop, daemon=True)
refresh_thread.start()
logger.info("sniffer: sniffing on interface=%s deckies=%d", interface, len(ip_map))
try:
sniff(
iface=interface,
filter="tcp",
prn=engine.on_packet,
store=False,
stop_filter=lambda pkt: stop_event.is_set(),
)
except Exception as exc:
logger.error("sniffer: scapy sniff exited: %s", exc)
finally:
stop_event.set()
logger.info("sniffer: sniff loop ended")
@_traced("sniffer.worker")
async def sniffer_worker(log_file: str) -> None:
"""
Async entry point — started as asyncio.create_task in the API lifespan.
Fully fault-isolated: catches all exceptions, logs them, and returns
cleanly. The API continues running regardless of sniffer state.
"""
try:
# Interface selection: explicit env override wins, otherwise probe
# both the MACVLAN and IPvlan host-side names since the driver
# choice is per-deploy (--ipvlan flag).
env_iface = os.environ.get("DECNET_SNIFFER_IFACE")
if env_iface:
interface = env_iface
elif _interface_exists(HOST_MACVLAN_IFACE):
interface = HOST_MACVLAN_IFACE
elif _interface_exists(HOST_IPVLAN_IFACE):
interface = HOST_IPVLAN_IFACE
else:
logger.warning(
"sniffer: neither %s nor %s found — sniffer disabled "
"(fleet may not be deployed yet)",
HOST_MACVLAN_IFACE, HOST_IPVLAN_IFACE,
)
return
if not _interface_exists(interface):
logger.warning(
"sniffer: interface %s not found — sniffer disabled "
"(fleet may not be deployed yet)", interface,
)
return
log_path = Path(log_file)
json_path = log_path.with_suffix(".json")
log_path.parent.mkdir(parents=True, exist_ok=True)
stop_event = threading.Event()
loop = asyncio.get_running_loop()
# Connect to the bus for decky.{id}.traffic fan-out. Failure here
# is non-fatal: the sniffer still writes syslog, it just doesn't
# push notifications to downstream consumers.
bus: BaseBus | None = None
try:
candidate = get_bus(client_name="sniffer")
await candidate.connect()
bus = candidate
except Exception as exc: # noqa: BLE001
logger.warning(
"sniffer: bus unavailable, running in publish-off mode: %s", exc,
)
publish_fn: Callable[[str, str, dict[str, Any]], None] | None = None
if bus is not None:
publish_fn = _make_thread_safe_publisher(bus, loop)
# Dedicated thread pool so the long-running sniff loop doesn't
# occupy a slot in the default asyncio executor.
sniffer_pool = ThreadPoolExecutor(
max_workers=2, thread_name_prefix="decnet-sniffer",
)
try:
await loop.run_in_executor(
sniffer_pool, _sniff_loop,
interface, log_path, json_path, stop_event, publish_fn,
)
except asyncio.CancelledError:
logger.info("sniffer: shutdown requested")
stop_event.set()
sniffer_pool.shutdown(wait=False)
raise
finally:
sniffer_pool.shutdown(wait=False)
if bus is not None:
with contextlib.suppress(Exception):
await bus.close()
except asyncio.CancelledError:
raise
except Exception as exc:
logger.error("sniffer: worker failed — API continues without sniffing: %s", exc)