Merge branch 'master' of https://github.com/commaai/openpilot into master3

This commit is contained in:
Rick Lan
2024-06-12 13:51:15 +08:00
374 changed files with 10496 additions and 6679 deletions
View File
+826
View File
@@ -0,0 +1,826 @@
#!/usr/bin/env python3
from __future__ import annotations
import base64
import bz2
import hashlib
import io
import json
import os
import queue
import random
import select
import socket
import sys
import tempfile
import threading
import time
from dataclasses import asdict, dataclass, replace
from datetime import datetime
from functools import partial
from queue import Queue
from typing import cast
from collections.abc import Callable
import requests
from jsonrpc import JSONRPCResponseManager, dispatcher
from websocket import (ABNF, WebSocket, WebSocketException, WebSocketTimeoutException,
create_connection)
import cereal.messaging as messaging
from cereal import log
from cereal.services import SERVICE_LIST
from openpilot.common.api import Api
from openpilot.common.file_helpers import CallbackReader
from openpilot.common.params import Params
from openpilot.common.realtime import set_core_affinity
from openpilot.system.hardware import HARDWARE, PC
from openpilot.system.loggerd.xattr_cache import getxattr, setxattr
from openpilot.common.swaglog import cloudlog
from openpilot.system.version import get_build_metadata
from openpilot.system.hardware.hw import Paths
ATHENA_HOST = os.getenv('ATHENA_HOST', 'wss://athena.comma.ai')
HANDLER_THREADS = int(os.getenv('HANDLER_THREADS', "4"))
LOCAL_PORT_WHITELIST = {8022}
LOG_ATTR_NAME = 'user.upload'
LOG_ATTR_VALUE_MAX_UNIX_TIME = int.to_bytes(2147483647, 4, sys.byteorder)
RECONNECT_TIMEOUT_S = 70
RETRY_DELAY = 10 # seconds
MAX_RETRY_COUNT = 30 # Try for at most 5 minutes if upload fails immediately
MAX_AGE = 31 * 24 * 3600 # seconds
WS_FRAME_SIZE = 4096
NetworkType = log.DeviceState.NetworkType
UploadFileDict = dict[str, str | int | float | bool]
UploadItemDict = dict[str, str | bool | int | float | dict[str, str]]
UploadFilesToUrlResponse = dict[str, int | list[UploadItemDict] | list[str]]
@dataclass
class UploadFile:
fn: str
url: str
headers: dict[str, str]
allow_cellular: bool
@classmethod
def from_dict(cls, d: dict) -> UploadFile:
return cls(d.get("fn", ""), d.get("url", ""), d.get("headers", {}), d.get("allow_cellular", False))
@dataclass
class UploadItem:
path: str
url: str
headers: dict[str, str]
created_at: int
id: str | None
retry_count: int = 0
current: bool = False
progress: float = 0
allow_cellular: bool = False
@classmethod
def from_dict(cls, d: dict) -> UploadItem:
return cls(d["path"], d["url"], d["headers"], d["created_at"], d["id"], d["retry_count"], d["current"],
d["progress"], d["allow_cellular"])
dispatcher["echo"] = lambda s: s
recv_queue: Queue[str] = queue.Queue()
send_queue: Queue[str] = queue.Queue()
upload_queue: Queue[UploadItem] = queue.Queue()
low_priority_send_queue: Queue[str] = queue.Queue()
log_recv_queue: Queue[str] = queue.Queue()
cancelled_uploads: set[str] = set()
cur_upload_items: dict[int, UploadItem | None] = {}
def strip_bz2_extension(fn: str) -> str:
if fn.endswith('.bz2'):
return fn[:-4]
return fn
class AbortTransferException(Exception):
pass
class UploadQueueCache:
@staticmethod
def initialize(upload_queue: Queue[UploadItem]) -> None:
try:
upload_queue_json = Params().get("AthenadUploadQueue")
if upload_queue_json is not None:
for item in json.loads(upload_queue_json):
upload_queue.put(UploadItem.from_dict(item))
except Exception:
cloudlog.exception("athena.UploadQueueCache.initialize.exception")
@staticmethod
def cache(upload_queue: Queue[UploadItem]) -> None:
try:
queue: list[UploadItem | None] = list(upload_queue.queue)
items = [asdict(i) for i in queue if i is not None and (i.id not in cancelled_uploads)]
Params().put("AthenadUploadQueue", json.dumps(items))
except Exception:
cloudlog.exception("athena.UploadQueueCache.cache.exception")
def handle_long_poll(ws: WebSocket, exit_event: threading.Event | None) -> None:
end_event = threading.Event()
threads = [
threading.Thread(target=ws_manage, args=(ws, end_event), name='ws_manage'),
threading.Thread(target=ws_recv, args=(ws, end_event), name='ws_recv'),
threading.Thread(target=ws_send, args=(ws, end_event), name='ws_send'),
threading.Thread(target=upload_handler, args=(end_event,), name='upload_handler'),
threading.Thread(target=log_handler, args=(end_event,), name='log_handler'),
threading.Thread(target=stat_handler, args=(end_event,), name='stat_handler'),
] + [
threading.Thread(target=jsonrpc_handler, args=(end_event,), name=f'worker_{x}')
for x in range(HANDLER_THREADS)
]
for thread in threads:
thread.start()
try:
while not end_event.wait(0.1):
if exit_event is not None and exit_event.is_set():
end_event.set()
except (KeyboardInterrupt, SystemExit):
end_event.set()
raise
finally:
for thread in threads:
cloudlog.debug(f"athena.joining {thread.name}")
thread.join()
def jsonrpc_handler(end_event: threading.Event) -> None:
dispatcher["startLocalProxy"] = partial(startLocalProxy, end_event)
while not end_event.is_set():
try:
data = recv_queue.get(timeout=1)
if "method" in data:
cloudlog.event("athena.jsonrpc_handler.call_method", data=data)
response = JSONRPCResponseManager.handle(data, dispatcher)
send_queue.put_nowait(response.json)
elif "id" in data and ("result" in data or "error" in data):
log_recv_queue.put_nowait(data)
else:
raise Exception("not a valid request or response")
except queue.Empty:
pass
except Exception as e:
cloudlog.exception("athena jsonrpc handler failed")
send_queue.put_nowait(json.dumps({"error": str(e)}))
def retry_upload(tid: int, end_event: threading.Event, increase_count: bool = True) -> None:
item = cur_upload_items[tid]
if item is not None and item.retry_count < MAX_RETRY_COUNT:
new_retry_count = item.retry_count + 1 if increase_count else item.retry_count
item = replace(
item,
retry_count=new_retry_count,
progress=0,
current=False
)
upload_queue.put_nowait(item)
UploadQueueCache.cache(upload_queue)
cur_upload_items[tid] = None
for _ in range(RETRY_DELAY):
time.sleep(1)
if end_event.is_set():
break
def cb(sm, item, tid, end_event: threading.Event, sz: int, cur: int) -> None:
# Abort transfer if connection changed to metered after starting upload
# or if athenad is shutting down to re-connect the websocket
sm.update(0)
metered = sm['deviceState'].networkMetered
if metered and (not item.allow_cellular):
raise AbortTransferException
if end_event.is_set():
raise AbortTransferException
cur_upload_items[tid] = replace(item, progress=cur / sz if sz else 1)
def upload_handler(end_event: threading.Event) -> None:
sm = messaging.SubMaster(['deviceState'])
tid = threading.get_ident()
while not end_event.is_set():
cur_upload_items[tid] = None
try:
cur_upload_items[tid] = item = replace(upload_queue.get(timeout=1), current=True)
if item.id in cancelled_uploads:
cancelled_uploads.remove(item.id)
continue
# Remove item if too old
age = datetime.now() - datetime.fromtimestamp(item.created_at / 1000)
if age.total_seconds() > MAX_AGE:
cloudlog.event("athena.upload_handler.expired", item=item, error=True)
continue
# Check if uploading over metered connection is allowed
sm.update(0)
metered = sm['deviceState'].networkMetered
network_type = sm['deviceState'].networkType.raw
if metered and (not item.allow_cellular):
retry_upload(tid, end_event, False)
continue
try:
fn = item.path
try:
sz = os.path.getsize(fn)
except OSError:
sz = -1
cloudlog.event("athena.upload_handler.upload_start", fn=fn, sz=sz, network_type=network_type, metered=metered, retry_count=item.retry_count)
response = _do_upload(item, partial(cb, sm, item, tid, end_event))
if response.status_code not in (200, 201, 401, 403, 412):
cloudlog.event("athena.upload_handler.retry", status_code=response.status_code, fn=fn, sz=sz, network_type=network_type, metered=metered)
retry_upload(tid, end_event)
else:
cloudlog.event("athena.upload_handler.success", fn=fn, sz=sz, network_type=network_type, metered=metered)
UploadQueueCache.cache(upload_queue)
except (requests.exceptions.Timeout, requests.exceptions.ConnectionError, requests.exceptions.SSLError):
cloudlog.event("athena.upload_handler.timeout", fn=fn, sz=sz, network_type=network_type, metered=metered)
retry_upload(tid, end_event)
except AbortTransferException:
cloudlog.event("athena.upload_handler.abort", fn=fn, sz=sz, network_type=network_type, metered=metered)
retry_upload(tid, end_event, False)
except queue.Empty:
pass
except Exception:
cloudlog.exception("athena.upload_handler.exception")
def _do_upload(upload_item: UploadItem, callback: Callable = None) -> requests.Response:
path = upload_item.path
compress = False
# If file does not exist, but does exist without the .bz2 extension we will compress on the fly
if not os.path.exists(path) and os.path.exists(strip_bz2_extension(path)):
path = strip_bz2_extension(path)
compress = True
with open(path, "rb") as f:
content = f.read()
if compress:
cloudlog.event("athena.upload_handler.compress", fn=path, fn_orig=upload_item.path)
content = bz2.compress(content)
with io.BytesIO(content) as data:
return requests.put(upload_item.url,
data=CallbackReader(data, callback, len(content)) if callback else data,
headers={**upload_item.headers, 'Content-Length': str(len(content))},
timeout=30)
# security: user should be able to request any message from their car
@dispatcher.add_method
def getMessage(service: str, timeout: int = 1000) -> dict:
if service is None or service not in SERVICE_LIST:
raise Exception("invalid service")
socket = messaging.sub_sock(service, timeout=timeout)
ret = messaging.recv_one(socket)
if ret is None:
raise TimeoutError
# this is because capnp._DynamicStructReader doesn't have typing information
return cast(dict, ret.to_dict())
@dispatcher.add_method
def getVersion() -> dict[str, str]:
build_metadata = get_build_metadata()
return {
"version": build_metadata.openpilot.version,
"remote": build_metadata.openpilot.git_normalized_origin,
"branch": build_metadata.channel,
"commit": build_metadata.openpilot.git_commit,
}
@dispatcher.add_method
def setNavDestination(latitude: int = 0, longitude: int = 0, place_name: str = None, place_details: str = None) -> dict[str, int]:
destination = {
"latitude": latitude,
"longitude": longitude,
"place_name": place_name,
"place_details": place_details,
}
Params().put("NavDestination", json.dumps(destination))
return {"success": 1}
def scan_dir(path: str, prefix: str) -> list[str]:
files = []
# only walk directories that match the prefix
# (glob and friends traverse entire dir tree)
with os.scandir(path) as i:
for e in i:
rel_path = os.path.relpath(e.path, Paths.log_root())
if e.is_dir(follow_symlinks=False):
# add trailing slash
rel_path = os.path.join(rel_path, '')
# if prefix is a partial dir name, current dir will start with prefix
# if prefix is a partial file name, prefix with start with dir name
if rel_path.startswith(prefix) or prefix.startswith(rel_path):
files.extend(scan_dir(e.path, prefix))
else:
if rel_path.startswith(prefix):
files.append(rel_path)
return files
@dispatcher.add_method
def listDataDirectory(prefix='') -> list[str]:
return scan_dir(Paths.log_root(), prefix)
@dispatcher.add_method
def uploadFileToUrl(fn: str, url: str, headers: dict[str, str]) -> UploadFilesToUrlResponse:
# this is because mypy doesn't understand that the decorator doesn't change the return type
response: UploadFilesToUrlResponse = uploadFilesToUrls([{
"fn": fn,
"url": url,
"headers": headers,
}])
return response
@dispatcher.add_method
def uploadFilesToUrls(files_data: list[UploadFileDict]) -> UploadFilesToUrlResponse:
files = map(UploadFile.from_dict, files_data)
items: list[UploadItemDict] = []
failed: list[str] = []
for file in files:
if len(file.fn) == 0 or file.fn[0] == '/' or '..' in file.fn or len(file.url) == 0:
failed.append(file.fn)
continue
path = os.path.join(Paths.log_root(), file.fn)
if not os.path.exists(path) and not os.path.exists(strip_bz2_extension(path)):
failed.append(file.fn)
continue
# Skip item if already in queue
url = file.url.split('?')[0]
if any(url == item['url'].split('?')[0] for item in listUploadQueue()):
continue
item = UploadItem(
path=path,
url=file.url,
headers=file.headers,
created_at=int(time.time() * 1000),
id=None,
allow_cellular=file.allow_cellular,
)
upload_id = hashlib.sha1(str(item).encode()).hexdigest()
item = replace(item, id=upload_id)
upload_queue.put_nowait(item)
items.append(asdict(item))
UploadQueueCache.cache(upload_queue)
resp: UploadFilesToUrlResponse = {"enqueued": len(items), "items": items}
if failed:
resp["failed"] = failed
return resp
@dispatcher.add_method
def listUploadQueue() -> list[UploadItemDict]:
items = list(upload_queue.queue) + list(cur_upload_items.values())
return [asdict(i) for i in items if (i is not None) and (i.id not in cancelled_uploads)]
@dispatcher.add_method
def cancelUpload(upload_id: str | list[str]) -> dict[str, int | str]:
if not isinstance(upload_id, list):
upload_id = [upload_id]
uploading_ids = {item.id for item in list(upload_queue.queue)}
cancelled_ids = uploading_ids.intersection(upload_id)
if len(cancelled_ids) == 0:
return {"success": 0, "error": "not found"}
cancelled_uploads.update(cancelled_ids)
return {"success": 1}
@dispatcher.add_method
def setRouteViewed(route: str) -> dict[str, int | str]:
# maintain a list of the last 10 routes viewed in connect
params = Params()
r = params.get("AthenadRecentlyViewedRoutes", encoding="utf8")
routes = [] if r is None else r.split(",")
routes.append(route)
# remove duplicates
routes = list(dict.fromkeys(routes))
params.put("AthenadRecentlyViewedRoutes", ",".join(routes[-10:]))
return {"success": 1}
def startLocalProxy(global_end_event: threading.Event, remote_ws_uri: str, local_port: int) -> dict[str, int]:
try:
if local_port not in LOCAL_PORT_WHITELIST:
raise Exception("Requested local port not whitelisted")
cloudlog.debug("athena.startLocalProxy.starting")
dongle_id = Params().get("DongleId").decode('utf8')
identity_token = Api(dongle_id).get_token()
ws = create_connection(remote_ws_uri,
cookie="jwt=" + identity_token,
enable_multithread=True)
# Set TOS to keep connection responsive while under load.
# DSCP of 36/HDD_LINUX_AC_VI with the minimum delay flag
ws.sock.setsockopt(socket.IPPROTO_IP, socket.IP_TOS, 0x90)
ssock, csock = socket.socketpair()
local_sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
local_sock.connect(('127.0.0.1', local_port))
local_sock.setblocking(False)
proxy_end_event = threading.Event()
threads = [
threading.Thread(target=ws_proxy_recv, args=(ws, local_sock, ssock, proxy_end_event, global_end_event)),
threading.Thread(target=ws_proxy_send, args=(ws, local_sock, csock, proxy_end_event))
]
for thread in threads:
thread.start()
cloudlog.debug("athena.startLocalProxy.started")
return {"success": 1}
except Exception as e:
cloudlog.exception("athenad.startLocalProxy.exception")
raise e
@dispatcher.add_method
def getPublicKey() -> str | None:
if not os.path.isfile(Paths.persist_root() + '/comma/id_rsa.pub'):
return None
with open(Paths.persist_root() + '/comma/id_rsa.pub') as f:
return f.read()
@dispatcher.add_method
def getSshAuthorizedKeys() -> str:
return Params().get("GithubSshKeys", encoding='utf8') or ''
@dispatcher.add_method
def getGithubUsername() -> str:
return Params().get("GithubUsername", encoding='utf8') or ''
@dispatcher.add_method
def getSimInfo():
return HARDWARE.get_sim_info()
@dispatcher.add_method
def getNetworkType():
return HARDWARE.get_network_type()
@dispatcher.add_method
def getNetworkMetered() -> bool:
network_type = HARDWARE.get_network_type()
return HARDWARE.get_network_metered(network_type)
@dispatcher.add_method
def getNetworks():
return HARDWARE.get_networks()
@dispatcher.add_method
def takeSnapshot() -> str | dict[str, str] | None:
from openpilot.system.camerad.snapshot.snapshot import jpeg_write, snapshot
ret = snapshot()
if ret is not None:
def b64jpeg(x):
if x is not None:
f = io.BytesIO()
jpeg_write(f, x)
return base64.b64encode(f.getvalue()).decode("utf-8")
else:
return None
return {'jpegBack': b64jpeg(ret[0]),
'jpegFront': b64jpeg(ret[1])}
else:
raise Exception("not available while camerad is started")
def get_logs_to_send_sorted() -> list[str]:
# TODO: scan once then use inotify to detect file creation/deletion
curr_time = int(time.time())
logs = []
for log_entry in os.listdir(Paths.swaglog_root()):
log_path = os.path.join(Paths.swaglog_root(), log_entry)
time_sent = 0
try:
value = getxattr(log_path, LOG_ATTR_NAME)
if value is not None:
time_sent = int.from_bytes(value, sys.byteorder)
except (ValueError, TypeError):
pass
# assume send failed and we lost the response if sent more than one hour ago
if not time_sent or curr_time - time_sent > 3600:
logs.append(log_entry)
# excluding most recent (active) log file
return sorted(logs)[:-1]
def log_handler(end_event: threading.Event) -> None:
if PC:
return
log_files = []
last_scan = 0.
while not end_event.is_set():
try:
curr_scan = time.monotonic()
if curr_scan - last_scan > 10:
log_files = get_logs_to_send_sorted()
last_scan = curr_scan
# send one log
curr_log = None
if len(log_files) > 0:
log_entry = log_files.pop() # newest log file
cloudlog.debug(f"athena.log_handler.forward_request {log_entry}")
try:
curr_time = int(time.time())
log_path = os.path.join(Paths.swaglog_root(), log_entry)
setxattr(log_path, LOG_ATTR_NAME, int.to_bytes(curr_time, 4, sys.byteorder))
with open(log_path) as f:
jsonrpc = {
"method": "forwardLogs",
"params": {
"logs": f.read()
},
"jsonrpc": "2.0",
"id": log_entry
}
low_priority_send_queue.put_nowait(json.dumps(jsonrpc))
curr_log = log_entry
except OSError:
pass # file could be deleted by log rotation
# wait for response up to ~100 seconds
# always read queue at least once to process any old responses that arrive
for _ in range(100):
if end_event.is_set():
break
try:
log_resp = json.loads(log_recv_queue.get(timeout=1))
log_entry = log_resp.get("id")
log_success = "result" in log_resp and log_resp["result"].get("success")
cloudlog.debug(f"athena.log_handler.forward_response {log_entry} {log_success}")
if log_entry and log_success:
log_path = os.path.join(Paths.swaglog_root(), log_entry)
try:
setxattr(log_path, LOG_ATTR_NAME, LOG_ATTR_VALUE_MAX_UNIX_TIME)
except OSError:
pass # file could be deleted by log rotation
if curr_log == log_entry:
break
except queue.Empty:
if curr_log is None:
break
except Exception:
cloudlog.exception("athena.log_handler.exception")
def stat_handler(end_event: threading.Event) -> None:
STATS_DIR = Paths.stats_root()
while not end_event.is_set():
last_scan = 0.
curr_scan = time.monotonic()
try:
if curr_scan - last_scan > 10:
stat_filenames = list(filter(lambda name: not name.startswith(tempfile.gettempprefix()), os.listdir(STATS_DIR)))
if len(stat_filenames) > 0:
stat_path = os.path.join(STATS_DIR, stat_filenames[0])
with open(stat_path) as f:
jsonrpc = {
"method": "storeStats",
"params": {
"stats": f.read()
},
"jsonrpc": "2.0",
"id": stat_filenames[0]
}
low_priority_send_queue.put_nowait(json.dumps(jsonrpc))
os.remove(stat_path)
last_scan = curr_scan
except Exception:
cloudlog.exception("athena.stat_handler.exception")
time.sleep(0.1)
def ws_proxy_recv(ws: WebSocket, local_sock: socket.socket, ssock: socket.socket, end_event: threading.Event, global_end_event: threading.Event) -> None:
while not (end_event.is_set() or global_end_event.is_set()):
try:
r = select.select((ws.sock,), (), (), 30)
if r[0]:
data = ws.recv()
if isinstance(data, str):
data = data.encode("utf-8")
local_sock.sendall(data)
except WebSocketTimeoutException:
pass
except Exception:
cloudlog.exception("athenad.ws_proxy_recv.exception")
break
cloudlog.debug("athena.ws_proxy_recv closing sockets")
ssock.close()
local_sock.close()
ws.close()
cloudlog.debug("athena.ws_proxy_recv done closing sockets")
end_event.set()
def ws_proxy_send(ws: WebSocket, local_sock: socket.socket, signal_sock: socket.socket, end_event: threading.Event) -> None:
while not end_event.is_set():
try:
r, _, _ = select.select((local_sock, signal_sock), (), ())
if r:
if r[0].fileno() == signal_sock.fileno():
# got end signal from ws_proxy_recv
end_event.set()
break
data = local_sock.recv(4096)
if not data:
# local_sock is dead
end_event.set()
break
ws.send(data, ABNF.OPCODE_BINARY)
except Exception:
cloudlog.exception("athenad.ws_proxy_send.exception")
end_event.set()
cloudlog.debug("athena.ws_proxy_send closing sockets")
signal_sock.close()
cloudlog.debug("athena.ws_proxy_send done closing sockets")
def ws_recv(ws: WebSocket, end_event: threading.Event) -> None:
last_ping = int(time.monotonic() * 1e9)
while not end_event.is_set():
try:
opcode, data = ws.recv_data(control_frame=True)
if opcode in (ABNF.OPCODE_TEXT, ABNF.OPCODE_BINARY):
if opcode == ABNF.OPCODE_TEXT:
data = data.decode("utf-8")
recv_queue.put_nowait(data)
elif opcode == ABNF.OPCODE_PING:
last_ping = int(time.monotonic() * 1e9)
Params().put("LastAthenaPingTime", str(last_ping))
except WebSocketTimeoutException:
ns_since_last_ping = int(time.monotonic() * 1e9) - last_ping
if ns_since_last_ping > RECONNECT_TIMEOUT_S * 1e9:
cloudlog.exception("athenad.ws_recv.timeout")
end_event.set()
except Exception:
cloudlog.exception("athenad.ws_recv.exception")
end_event.set()
def ws_send(ws: WebSocket, end_event: threading.Event) -> None:
while not end_event.is_set():
try:
try:
data = send_queue.get_nowait()
except queue.Empty:
data = low_priority_send_queue.get(timeout=1)
for i in range(0, len(data), WS_FRAME_SIZE):
frame = data[i:i+WS_FRAME_SIZE]
last = i + WS_FRAME_SIZE >= len(data)
opcode = ABNF.OPCODE_TEXT if i == 0 else ABNF.OPCODE_CONT
ws.send_frame(ABNF.create_frame(frame, opcode, last))
except queue.Empty:
pass
except Exception:
cloudlog.exception("athenad.ws_send.exception")
end_event.set()
def ws_manage(ws: WebSocket, end_event: threading.Event) -> None:
params = Params()
onroad_prev = None
sock = ws.sock
while True:
onroad = params.get_bool("IsOnroad")
if onroad != onroad_prev:
onroad_prev = onroad
if sock is not None:
# While not sending data, onroad, we can expect to time out in 7 + (7 * 2) = 21s
# offroad, we can expect to time out in 30 + (10 * 3) = 60s
# FIXME: TCP_USER_TIMEOUT is effectively 2x for some reason (32s), so it's mostly unused
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_USER_TIMEOUT, 16000 if onroad else 0)
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPIDLE, 7 if onroad else 30)
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPINTVL, 7 if onroad else 10)
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPCNT, 2 if onroad else 3)
if end_event.wait(5):
break
def backoff(retries: int) -> int:
return random.randrange(0, min(128, int(2 ** retries)))
def main(exit_event: threading.Event = None):
try:
set_core_affinity([0, 1, 2, 3])
except Exception:
cloudlog.exception("failed to set core affinity")
params = Params()
dongle_id = params.get("DongleId", encoding='utf-8')
UploadQueueCache.initialize(upload_queue)
ws_uri = ATHENA_HOST + "/ws/v2/" + dongle_id
api = Api(dongle_id)
conn_start = None
conn_retries = 0
while exit_event is None or not exit_event.is_set():
try:
if conn_start is None:
conn_start = time.monotonic()
cloudlog.event("athenad.main.connecting_ws", ws_uri=ws_uri, retries=conn_retries)
ws = create_connection(ws_uri,
cookie="jwt=" + api.get_token(),
enable_multithread=True,
timeout=30.0)
cloudlog.event("athenad.main.connected_ws", ws_uri=ws_uri, retries=conn_retries,
duration=time.monotonic() - conn_start)
conn_start = None
conn_retries = 0
cur_upload_items.clear()
handle_long_poll(ws, exit_event)
except (KeyboardInterrupt, SystemExit):
break
except (ConnectionError, TimeoutError, WebSocketException):
conn_retries += 1
params.remove("LastAthenaPingTime")
except Exception:
cloudlog.exception("athenad.main.exception")
conn_retries += 1
params.remove("LastAthenaPingTime")
time.sleep(backoff(conn_retries))
if __name__ == "__main__":
main()
+43
View File
@@ -0,0 +1,43 @@
#!/usr/bin/env python3
import time
from multiprocessing import Process
from openpilot.common.params import Params
from openpilot.system.manager.process import launcher
from openpilot.common.swaglog import cloudlog
from openpilot.system.hardware import HARDWARE
from openpilot.system.version import get_build_metadata
ATHENA_MGR_PID_PARAM = "AthenadPid"
def main():
params = Params()
dongle_id = params.get("DongleId").decode('utf-8')
build_metadata = get_build_metadata()
cloudlog.bind_global(dongle_id=dongle_id,
version=build_metadata.openpilot.version,
origin=build_metadata.openpilot.git_normalized_origin,
branch=build_metadata.channel,
commit=build_metadata.openpilot.git_commit,
dirty=build_metadata.openpilot.is_dirty,
device=HARDWARE.get_device_type())
try:
while 1:
cloudlog.info("starting athena daemon")
proc = Process(name='athenad', target=launcher, args=('system.athena.athenad', 'athenad'))
proc.start()
proc.join()
cloudlog.event("athenad exited", exitcode=proc.exitcode)
time.sleep(5)
except Exception:
cloudlog.exception("manage_athenad.exception")
finally:
params.remove(ATHENA_MGR_PID_PARAM)
if __name__ == '__main__':
main()
+99
View File
@@ -0,0 +1,99 @@
#!/usr/bin/env python3
import time
import json
import jwt
from pathlib import Path
from datetime import datetime, timedelta, UTC
from openpilot.common.api import api_get
from openpilot.common.params import Params
from openpilot.common.spinner import Spinner
from openpilot.selfdrive.controls.lib.alertmanager import set_offroad_alert
from openpilot.system.hardware import HARDWARE, PC
from openpilot.system.hardware.hw import Paths
from openpilot.common.swaglog import cloudlog
UNREGISTERED_DONGLE_ID = "UnregisteredDevice"
def is_registered_device() -> bool:
dongle = Params().get("DongleId", encoding='utf-8')
return dongle not in (None, UNREGISTERED_DONGLE_ID)
def register(show_spinner=False) -> str | None:
params = Params()
IMEI = params.get("IMEI", encoding='utf8')
HardwareSerial = params.get("HardwareSerial", encoding='utf8')
dongle_id: str | None = params.get("DongleId", encoding='utf8')
needs_registration = None in (IMEI, HardwareSerial, dongle_id)
pubkey = Path(Paths.persist_root()+"/comma/id_rsa.pub")
if not pubkey.is_file():
dongle_id = UNREGISTERED_DONGLE_ID
cloudlog.warning(f"missing public key: {pubkey}")
elif needs_registration:
if show_spinner:
spinner = Spinner()
spinner.update("registering device")
# Create registration token, in the future, this key will make JWTs directly
with open(Paths.persist_root()+"/comma/id_rsa.pub") as f1, open(Paths.persist_root()+"/comma/id_rsa") as f2:
public_key = f1.read()
private_key = f2.read()
# Block until we get the imei
serial = HARDWARE.get_serial()
start_time = time.monotonic()
imei1: str | None = None
imei2: str | None = None
while imei1 is None and imei2 is None:
try:
imei1, imei2 = HARDWARE.get_imei(0), HARDWARE.get_imei(1)
except Exception:
cloudlog.exception("Error getting imei, trying again...")
time.sleep(1)
if time.monotonic() - start_time > 60 and show_spinner:
spinner.update(f"registering device - serial: {serial}, IMEI: ({imei1}, {imei2})")
params.put("IMEI", imei1)
params.put("HardwareSerial", serial)
backoff = 0
start_time = time.monotonic()
while True:
try:
register_token = jwt.encode({'register': True, 'exp': datetime.now(UTC).replace(tzinfo=None) + timedelta(hours=1)}, private_key, algorithm='RS256')
cloudlog.info("getting pilotauth")
resp = api_get("v2/pilotauth/", method='POST', timeout=15,
imei=imei1, imei2=imei2, serial=serial, public_key=public_key, register_token=register_token)
if resp.status_code in (402, 403):
cloudlog.info(f"Unable to register device, got {resp.status_code}")
dongle_id = UNREGISTERED_DONGLE_ID
else:
dongleauth = json.loads(resp.text)
dongle_id = dongleauth["dongle_id"]
break
except Exception:
cloudlog.exception("failed to authenticate")
backoff = min(backoff + 1, 15)
time.sleep(backoff)
if time.monotonic() - start_time > 60 and show_spinner:
spinner.update(f"registering device - serial: {serial}, IMEI: ({imei1}, {imei2})")
if show_spinner:
spinner.close()
if dongle_id:
params.put("DongleId", dongle_id)
set_offroad_alert("Offroad_UnofficialHardware", (dongle_id == UNREGISTERED_DONGLE_ID) and not PC)
return dongle_id
if __name__ == "__main__":
print(register())
View File
+70
View File
@@ -0,0 +1,70 @@
import http.server
import socket
class MockResponse:
def __init__(self, json, status_code):
self.json = json
self.text = json
self.status_code = status_code
class EchoSocket:
def __init__(self, port):
self.socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
self.socket.bind(('127.0.0.1', port))
self.socket.listen(1)
def run(self):
conn, _ = self.socket.accept()
conn.settimeout(5.0)
try:
while True:
data = conn.recv(4096)
if data:
print(f'EchoSocket got {data}')
conn.sendall(data)
else:
break
finally:
conn.shutdown(0)
conn.close()
self.socket.shutdown(0)
self.socket.close()
class MockApi:
def __init__(self, dongle_id):
pass
def get_token(self):
return "fake-token"
class MockWebsocket:
sock = socket.socket()
def __init__(self, recv_queue, send_queue):
self.recv_queue = recv_queue
self.send_queue = send_queue
def recv(self):
data = self.recv_queue.get()
if isinstance(data, Exception):
raise data
return data
def send(self, data, opcode):
self.send_queue.put_nowait((data, opcode))
def close(self):
pass
class HTTPRequestHandler(http.server.SimpleHTTPRequestHandler):
def do_PUT(self):
length = int(self.headers['Content-Length'])
self.rfile.read(length)
self.send_response(201, "Created")
self.end_headers()
+423
View File
@@ -0,0 +1,423 @@
import pytest
from functools import wraps
import json
import multiprocessing
import os
import requests
import shutil
import time
import threading
import queue
from dataclasses import asdict, replace
from datetime import datetime, timedelta
from websocket import ABNF
from websocket._exceptions import WebSocketConnectionClosedException
from cereal import messaging
from openpilot.common.params import Params
from openpilot.common.timeout import Timeout
from openpilot.system.athena import athenad
from openpilot.system.athena.athenad import MAX_RETRY_COUNT, dispatcher
from openpilot.system.athena.tests.helpers import HTTPRequestHandler, MockWebsocket, MockApi, EchoSocket
from openpilot.selfdrive.test.helpers import http_server_context
from openpilot.system.hardware.hw import Paths
def seed_athena_server(host, port):
with Timeout(2, 'HTTP Server seeding failed'):
while True:
try:
requests.put(f'http://{host}:{port}/qlog.bz2', data='', timeout=10)
break
except requests.exceptions.ConnectionError:
time.sleep(0.1)
def with_upload_handler(func):
@wraps(func)
def wrapper(*args, **kwargs):
end_event = threading.Event()
thread = threading.Thread(target=athenad.upload_handler, args=(end_event,))
thread.start()
try:
return func(*args, **kwargs)
finally:
end_event.set()
thread.join()
return wrapper
@pytest.fixture
def mock_create_connection(mocker):
return mocker.patch('openpilot.system.athena.athenad.create_connection')
@pytest.fixture
def host():
with http_server_context(handler=HTTPRequestHandler, setup=seed_athena_server) as (host, port):
yield f"http://{host}:{port}"
class TestAthenadMethods:
@classmethod
def setup_class(cls):
cls.SOCKET_PORT = 45454
athenad.Api = MockApi
athenad.LOCAL_PORT_WHITELIST = {cls.SOCKET_PORT}
def setup_method(self):
self.default_params = {
"DongleId": "0000000000000000",
"GithubSshKeys": b"ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQC307aE+nuHzTAgaJhzSf5v7ZZQW9gaperjhCmyPyl4PzY7T1mDGenTlVTN7yoVFZ9UfO9oMQqo0n1OwDIiqbIFxqnhrHU0cYfj88rI85m5BEKlNu5RdaVTj1tcbaPpQc5kZEolaI1nDDjzV0lwS7jo5VYDHseiJHlik3HH1SgtdtsuamGR2T80q1SyW+5rHoMOJG73IH2553NnWuikKiuikGHUYBd00K1ilVAK2xSiMWJp55tQfZ0ecr9QjEsJ+J/efL4HqGNXhffxvypCXvbUYAFSddOwXUPo5BTKevpxMtH+2YrkpSjocWA04VnTYFiPG6U4ItKmbLOTFZtPzoez private", # noqa: E501
"GithubUsername": b"commaci",
"AthenadUploadQueue": '[]',
}
self.params = Params()
for k, v in self.default_params.items():
self.params.put(k, v)
self.params.put_bool("GsmMetered", True)
athenad.upload_queue = queue.Queue()
athenad.cur_upload_items.clear()
athenad.cancelled_uploads.clear()
for i in os.listdir(Paths.log_root()):
p = os.path.join(Paths.log_root(), i)
if os.path.isdir(p):
shutil.rmtree(p)
else:
os.unlink(p)
# *** test helpers ***
@staticmethod
def _wait_for_upload():
now = time.time()
while time.time() - now < 5:
if athenad.upload_queue.qsize() == 0:
break
@staticmethod
def _create_file(file: str, parent: str = None, data: bytes = b'') -> str:
fn = os.path.join(Paths.log_root() if parent is None else parent, file)
os.makedirs(os.path.dirname(fn), exist_ok=True)
with open(fn, 'wb') as f:
f.write(data)
return fn
# *** test cases ***
def test_echo(self):
assert dispatcher["echo"]("bob") == "bob"
def test_get_message(self):
with pytest.raises(TimeoutError) as _:
dispatcher["getMessage"]("controlsState")
end_event = multiprocessing.Event()
pub_sock = messaging.pub_sock("deviceState")
def send_deviceState():
while not end_event.is_set():
msg = messaging.new_message('deviceState')
pub_sock.send(msg.to_bytes())
time.sleep(0.01)
p = multiprocessing.Process(target=send_deviceState)
p.start()
time.sleep(0.1)
try:
deviceState = dispatcher["getMessage"]("deviceState")
assert deviceState['deviceState']
finally:
end_event.set()
p.join()
def test_list_data_directory(self):
route = '2021-03-29--13-32-47'
segments = [0, 1, 2, 3, 11]
filenames = ['qlog', 'qcamera.ts', 'rlog', 'fcamera.hevc', 'ecamera.hevc', 'dcamera.hevc']
files = [f'{route}--{s}/{f}' for s in segments for f in filenames]
for file in files:
self._create_file(file)
resp = dispatcher["listDataDirectory"]()
assert resp, 'list empty!'
assert len(resp) == len(files)
resp = dispatcher["listDataDirectory"](f'{route}--123')
assert len(resp) == 0
prefix = f'{route}'
expected = list(filter(lambda f: f.startswith(prefix), files))
resp = dispatcher["listDataDirectory"](prefix)
assert resp, 'list empty!'
assert len(resp) == len(expected)
prefix = f'{route}--1'
expected = list(filter(lambda f: f.startswith(prefix), files))
resp = dispatcher["listDataDirectory"](prefix)
assert resp, 'list empty!'
assert len(resp) == len(expected)
prefix = f'{route}--1/'
expected = list(filter(lambda f: f.startswith(prefix), files))
resp = dispatcher["listDataDirectory"](prefix)
assert resp, 'list empty!'
assert len(resp) == len(expected)
prefix = f'{route}--1/q'
expected = list(filter(lambda f: f.startswith(prefix), files))
resp = dispatcher["listDataDirectory"](prefix)
assert resp, 'list empty!'
assert len(resp) == len(expected)
def test_strip_bz2_extension(self):
fn = self._create_file('qlog.bz2')
if fn.endswith('.bz2'):
assert athenad.strip_bz2_extension(fn) == fn[:-4]
@pytest.mark.parametrize("compress", [True, False])
def test_do_upload(self, host, compress):
# random bytes to ensure rather large object post-compression
fn = self._create_file('qlog', data=os.urandom(10000 * 1024))
upload_fn = fn + ('.bz2' if compress else '')
item = athenad.UploadItem(path=upload_fn, url="http://localhost:1238", headers={}, created_at=int(time.time()*1000), id='')
with pytest.raises(requests.exceptions.ConnectionError):
athenad._do_upload(item)
item = athenad.UploadItem(path=upload_fn, url=f"{host}/qlog.bz2", headers={}, created_at=int(time.time()*1000), id='')
resp = athenad._do_upload(item)
assert resp.status_code == 201
def test_upload_file_to_url(self, host):
fn = self._create_file('qlog.bz2')
resp = dispatcher["uploadFileToUrl"]("qlog.bz2", f"{host}/qlog.bz2", {})
assert resp['enqueued'] == 1
assert 'failed' not in resp
assert {"path": fn, "url": f"{host}/qlog.bz2", "headers": {}}.items() <= resp['items'][0].items()
assert resp['items'][0].get('id') is not None
assert athenad.upload_queue.qsize() == 1
def test_upload_file_to_url_duplicate(self, host):
self._create_file('qlog.bz2')
url1 = f"{host}/qlog.bz2?sig=sig1"
dispatcher["uploadFileToUrl"]("qlog.bz2", url1, {})
# Upload same file again, but with different signature
url2 = f"{host}/qlog.bz2?sig=sig2"
resp = dispatcher["uploadFileToUrl"]("qlog.bz2", url2, {})
assert resp == {'enqueued': 0, 'items': []}
def test_upload_file_to_url_does_not_exist(self, host):
not_exists_resp = dispatcher["uploadFileToUrl"]("does_not_exist.bz2", "http://localhost:1238", {})
assert not_exists_resp == {'enqueued': 0, 'items': [], 'failed': ['does_not_exist.bz2']}
@with_upload_handler
def test_upload_handler(self, host):
fn = self._create_file('qlog.bz2')
item = athenad.UploadItem(path=fn, url=f"{host}/qlog.bz2", headers={}, created_at=int(time.time()*1000), id='', allow_cellular=True)
athenad.upload_queue.put_nowait(item)
self._wait_for_upload()
time.sleep(0.1)
# TODO: verify that upload actually succeeded
# TODO: also check that end_event and metered network raises AbortTransferException
assert athenad.upload_queue.qsize() == 0
@pytest.mark.parametrize("status,retry", [(500,True), (412,False)])
@with_upload_handler
def test_upload_handler_retry(self, mocker, host, status, retry):
mock_put = mocker.patch('requests.put')
mock_put.return_value.status_code = status
fn = self._create_file('qlog.bz2')
item = athenad.UploadItem(path=fn, url=f"{host}/qlog.bz2", headers={}, created_at=int(time.time()*1000), id='', allow_cellular=True)
athenad.upload_queue.put_nowait(item)
self._wait_for_upload()
time.sleep(0.1)
assert athenad.upload_queue.qsize() == (1 if retry else 0)
if retry:
assert athenad.upload_queue.get().retry_count == 1
@with_upload_handler
def test_upload_handler_timeout(self):
"""When an upload times out or fails to connect it should be placed back in the queue"""
fn = self._create_file('qlog.bz2')
item = athenad.UploadItem(path=fn, url="http://localhost:44444/qlog.bz2", headers={}, created_at=int(time.time()*1000), id='', allow_cellular=True)
item_no_retry = replace(item, retry_count=MAX_RETRY_COUNT)
athenad.upload_queue.put_nowait(item_no_retry)
self._wait_for_upload()
time.sleep(0.1)
# Check that upload with retry count exceeded is not put back
assert athenad.upload_queue.qsize() == 0
athenad.upload_queue.put_nowait(item)
self._wait_for_upload()
time.sleep(0.1)
# Check that upload item was put back in the queue with incremented retry count
assert athenad.upload_queue.qsize() == 1
assert athenad.upload_queue.get().retry_count == 1
@with_upload_handler
def test_cancel_upload(self):
item = athenad.UploadItem(path="qlog.bz2", url="http://localhost:44444/qlog.bz2", headers={},
created_at=int(time.time()*1000), id='id', allow_cellular=True)
athenad.upload_queue.put_nowait(item)
dispatcher["cancelUpload"](item.id)
assert item.id in athenad.cancelled_uploads
self._wait_for_upload()
time.sleep(0.1)
assert athenad.upload_queue.qsize() == 0
assert len(athenad.cancelled_uploads) == 0
@with_upload_handler
def test_cancel_expiry(self):
t_future = datetime.now() - timedelta(days=40)
ts = int(t_future.strftime("%s")) * 1000
# Item that would time out if actually uploaded
fn = self._create_file('qlog.bz2')
item = athenad.UploadItem(path=fn, url="http://localhost:44444/qlog.bz2", headers={}, created_at=ts, id='', allow_cellular=True)
athenad.upload_queue.put_nowait(item)
self._wait_for_upload()
time.sleep(0.1)
assert athenad.upload_queue.qsize() == 0
def test_list_upload_queue_empty(self):
items = dispatcher["listUploadQueue"]()
assert len(items) == 0
@with_upload_handler
def test_list_upload_queue_current(self, host: str):
fn = self._create_file('qlog.bz2')
item = athenad.UploadItem(path=fn, url=f"{host}/qlog.bz2", headers={}, created_at=int(time.time()*1000), id='', allow_cellular=True)
athenad.upload_queue.put_nowait(item)
self._wait_for_upload()
items = dispatcher["listUploadQueue"]()
assert len(items) == 1
assert items[0]['current']
def test_list_upload_queue(self):
item = athenad.UploadItem(path="qlog.bz2", url="http://localhost:44444/qlog.bz2", headers={},
created_at=int(time.time()*1000), id='id', allow_cellular=True)
athenad.upload_queue.put_nowait(item)
items = dispatcher["listUploadQueue"]()
assert len(items) == 1
assert items[0] == asdict(item)
assert not items[0]['current']
athenad.cancelled_uploads.add(item.id)
items = dispatcher["listUploadQueue"]()
assert len(items) == 0
def test_upload_queue_persistence(self):
item1 = athenad.UploadItem(path="_", url="_", headers={}, created_at=int(time.time()), id='id1')
item2 = athenad.UploadItem(path="_", url="_", headers={}, created_at=int(time.time()), id='id2')
athenad.upload_queue.put_nowait(item1)
athenad.upload_queue.put_nowait(item2)
# Ensure cancelled items are not persisted
athenad.cancelled_uploads.add(item2.id)
# serialize item
athenad.UploadQueueCache.cache(athenad.upload_queue)
# deserialize item
athenad.upload_queue.queue.clear()
athenad.UploadQueueCache.initialize(athenad.upload_queue)
assert athenad.upload_queue.qsize() == 1
assert asdict(athenad.upload_queue.queue[-1]) == asdict(item1)
def test_start_local_proxy(self, mock_create_connection):
end_event = threading.Event()
ws_recv = queue.Queue()
ws_send = queue.Queue()
mock_ws = MockWebsocket(ws_recv, ws_send)
mock_create_connection.return_value = mock_ws
echo_socket = EchoSocket(self.SOCKET_PORT)
socket_thread = threading.Thread(target=echo_socket.run)
socket_thread.start()
athenad.startLocalProxy(end_event, 'ws://localhost:1234', self.SOCKET_PORT)
ws_recv.put_nowait(b'ping')
try:
recv = ws_send.get(timeout=5)
assert recv == (b'ping', ABNF.OPCODE_BINARY), recv
finally:
# signal websocket close to athenad.ws_proxy_recv
ws_recv.put_nowait(WebSocketConnectionClosedException())
socket_thread.join()
def test_get_ssh_authorized_keys(self):
keys = dispatcher["getSshAuthorizedKeys"]()
assert keys == self.default_params["GithubSshKeys"].decode('utf-8')
def test_get_github_username(self):
keys = dispatcher["getGithubUsername"]()
assert keys == self.default_params["GithubUsername"].decode('utf-8')
def test_get_version(self):
resp = dispatcher["getVersion"]()
keys = ["version", "remote", "branch", "commit"]
assert list(resp.keys()) == keys
for k in keys:
assert isinstance(resp[k], str), f"{k} is not a string"
assert len(resp[k]) > 0, f"{k} has no value"
def test_jsonrpc_handler(self):
end_event = threading.Event()
thread = threading.Thread(target=athenad.jsonrpc_handler, args=(end_event,))
thread.daemon = True
thread.start()
try:
# with params
athenad.recv_queue.put_nowait(json.dumps({"method": "echo", "params": ["hello"], "jsonrpc": "2.0", "id": 0}))
resp = athenad.send_queue.get(timeout=3)
assert json.loads(resp) == {'result': 'hello', 'id': 0, 'jsonrpc': '2.0'}
# without params
athenad.recv_queue.put_nowait(json.dumps({"method": "getNetworkType", "jsonrpc": "2.0", "id": 0}))
resp = athenad.send_queue.get(timeout=3)
assert json.loads(resp) == {'result': 1, 'id': 0, 'jsonrpc': '2.0'}
# log forwarding
athenad.recv_queue.put_nowait(json.dumps({'result': {'success': 1}, 'id': 0, 'jsonrpc': '2.0'}))
resp = athenad.log_recv_queue.get(timeout=3)
assert json.loads(resp) == {'result': {'success': 1}, 'id': 0, 'jsonrpc': '2.0'}
finally:
end_event.set()
thread.join()
def test_get_logs_to_send_sorted(self):
fl = list()
for i in range(10):
file = f'swaglog.{i:010}'
self._create_file(file, Paths.swaglog_root())
fl.append(file)
# ensure the list is all logs except most recent
sl = athenad.get_logs_to_send_sorted()
assert sl == fl[:-1]
+102
View File
@@ -0,0 +1,102 @@
import pytest
import subprocess
import threading
import time
from typing import cast
from openpilot.common.params import Params
from openpilot.common.timeout import Timeout
from openpilot.system.athena import athenad
from openpilot.system.manager.helpers import write_onroad_params
from openpilot.system.hardware import TICI
TIMEOUT_TOLERANCE = 20 # seconds
def wifi_radio(on: bool) -> None:
if not TICI:
return
print(f"wifi {'on' if on else 'off'}")
subprocess.run(["nmcli", "radio", "wifi", "on" if on else "off"], check=True)
class TestAthenadPing:
params: Params
dongle_id: str
athenad: threading.Thread
exit_event: threading.Event
def _get_ping_time(self) -> str | None:
return cast(str | None, self.params.get("LastAthenaPingTime", encoding="utf-8"))
def _clear_ping_time(self) -> None:
self.params.remove("LastAthenaPingTime")
def _received_ping(self) -> bool:
return self._get_ping_time() is not None
@classmethod
def teardown_class(cls) -> None:
wifi_radio(True)
def setup_method(self) -> None:
self.params = Params()
self.dongle_id = self.params.get("DongleId", encoding="utf-8")
wifi_radio(True)
self._clear_ping_time()
self.exit_event = threading.Event()
self.athenad = threading.Thread(target=athenad.main, args=(self.exit_event,))
def teardown_method(self) -> None:
if self.athenad.is_alive():
self.exit_event.set()
self.athenad.join()
def assertTimeout(self, reconnect_time: float, subtests, mocker) -> None:
self.athenad.start()
mock_create_connection = mocker.patch('openpilot.system.athena.athenad.create_connection',
new_callable=lambda: mocker.MagicMock(wraps=athenad.create_connection))
time.sleep(1)
mock_create_connection.assert_called_once()
mock_create_connection.reset_mock()
# check normal behaviour, server pings on connection
with subtests.test("Wi-Fi: receives ping"), Timeout(70, "no ping received"):
while not self._received_ping():
time.sleep(0.1)
print("ping received")
mock_create_connection.assert_not_called()
# websocket should attempt reconnect after short time
with subtests.test("LTE: attempt reconnect"):
wifi_radio(False)
print("waiting for reconnect attempt")
start_time = time.monotonic()
with Timeout(reconnect_time, "no reconnect attempt"):
while not mock_create_connection.called:
time.sleep(0.1)
print(f"reconnect attempt after {time.monotonic() - start_time:.2f}s")
self._clear_ping_time()
# check ping received after reconnect
with subtests.test("LTE: receives ping"), Timeout(70, "no ping received"):
while not self._received_ping():
time.sleep(0.1)
print("ping received")
@pytest.mark.skipif(not TICI, reason="only run on desk")
def test_offroad(self, subtests, mocker) -> None:
write_onroad_params(False, self.params)
self.assertTimeout(60 + TIMEOUT_TOLERANCE, subtests, mocker) # based using TCP keepalive settings
@pytest.mark.skipif(not TICI, reason="only run on desk")
def test_onroad(self, subtests, mocker) -> None:
write_onroad_params(True, self.params)
self.assertTimeout(21 + TIMEOUT_TOLERANCE, subtests, mocker)
+74
View File
@@ -0,0 +1,74 @@
import json
from Crypto.PublicKey import RSA
from pathlib import Path
from openpilot.common.params import Params
from openpilot.system.athena.registration import register, UNREGISTERED_DONGLE_ID
from openpilot.system.athena.tests.helpers import MockResponse
from openpilot.system.hardware.hw import Paths
class TestRegistration:
def setup_method(self):
# clear params and setup key paths
self.params = Params()
self.params.clear_all()
persist_dir = Path(Paths.persist_root()) / "comma"
persist_dir.mkdir(parents=True, exist_ok=True)
self.priv_key = persist_dir / "id_rsa"
self.pub_key = persist_dir / "id_rsa.pub"
def _generate_keys(self):
self.pub_key.touch()
k = RSA.generate(2048)
with open(self.priv_key, "wb") as f:
f.write(k.export_key())
with open(self.pub_key, "wb") as f:
f.write(k.publickey().export_key())
def test_valid_cache(self, mocker):
# if all params are written, return the cached dongle id
self.params.put("IMEI", "imei")
self.params.put("HardwareSerial", "serial")
self._generate_keys()
m = mocker.patch("openpilot.system.athena.registration.api_get", autospec=True)
dongle = "DONGLE_ID_123"
self.params.put("DongleId", dongle)
assert register() == dongle
assert not m.called
def test_no_keys(self, mocker):
# missing pubkey
m = mocker.patch("openpilot.system.athena.registration.api_get", autospec=True)
dongle = register()
assert m.call_count == 0
assert dongle == UNREGISTERED_DONGLE_ID
assert self.params.get("DongleId", encoding='utf-8') == dongle
def test_missing_cache(self, mocker):
# keys exist but no dongle id
self._generate_keys()
m = mocker.patch("openpilot.system.athena.registration.api_get", autospec=True)
dongle = "DONGLE_ID_123"
m.return_value = MockResponse(json.dumps({'dongle_id': dongle}), 200)
assert register() == dongle
assert m.call_count == 1
# call again, shouldn't hit the API this time
assert register() == dongle
assert m.call_count == 1
assert self.params.get("DongleId", encoding='utf-8') == dongle
def test_unregistered(self, mocker):
# keys exist, but unregistered
self._generate_keys()
m = mocker.patch("openpilot.system.athena.registration.api_get", autospec=True)
m.return_value = MockResponse(None, 402)
dongle = register()
assert m.call_count == 1
assert dongle == UNREGISTERED_DONGLE_ID
assert self.params.get("DongleId", encoding='utf-8') == dongle
+2 -2
View File
@@ -1,6 +1,6 @@
Import('env', 'arch', 'cereal', 'messaging', 'common', 'gpucommon', 'visionipc')
Import('env', 'arch', 'messaging', 'common', 'gpucommon', 'visionipc')
libs = ['m', 'pthread', common, 'jpeg', 'OpenCL', 'yuv', cereal, messaging, 'zmq', 'capnp', 'kj', visionipc, gpucommon, 'atomic']
libs = ['m', 'pthread', common, 'jpeg', 'OpenCL', 'yuv', messaging, visionipc, gpucommon, 'atomic']
camera_obj = env.Object(['cameras/camera_qcom2.cc', 'cameras/camera_common.cc', 'cameras/camera_util.cc',
'sensors/ar0231.cc', 'sensors/ox03c10.cc', 'sensors/os04c10.cc'])
+1 -1
View File
@@ -5,7 +5,7 @@
#include <thread>
#include "cereal/messaging/messaging.h"
#include "cereal/visionipc/visionipc_server.h"
#include "msgq/visionipc/visionipc_server.h"
#include "common/queue.h"
#include "common/util.h"
+3 -3
View File
@@ -6,12 +6,12 @@ import numpy as np
from PIL import Image
import cereal.messaging as messaging
from cereal.visionipc import VisionIpcClient, VisionStreamType
from msgq.visionipc import VisionIpcClient, VisionStreamType
from openpilot.common.params import Params
from openpilot.common.realtime import DT_MDL
from openpilot.system.hardware import PC
from openpilot.selfdrive.controls.lib.alertmanager import set_offroad_alert
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
VISION_STREAMS = {
@@ -83,7 +83,7 @@ def snapshot():
front_camera_allowed = params.get_bool("RecordFront")
params.put_bool("IsTakingSnapshot", True)
set_offroad_alert("Offroad_IsTakingSnapshot", True)
time.sleep(2.0) # Give thermald time to read the param, or if just started give camerad time to start
time.sleep(2.0) # Give hardwared time to read the param, or if just started give camerad time to start
# Check if camerad is already started
try:
+1 -2
View File
@@ -1,4 +1,3 @@
#!/usr/bin/env python3
import pytest
import time
import numpy as np
@@ -8,7 +7,7 @@ from collections import defaultdict
import cereal.messaging as messaging
from cereal import log
from cereal.services import SERVICE_LIST
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
TEST_TIMESPAN = 30
LAG_FRAME_TOLERANCE = {log.FrameData.ImageSensor.ar0231: 0.5, # ARs use synced pulses for frame starts
+3 -8
View File
@@ -1,6 +1,4 @@
#!/usr/bin/env python3
import time
import unittest
import numpy as np
from openpilot.selfdrive.test.helpers import with_processes, phone_only
@@ -9,9 +7,9 @@ from openpilot.system.camerad.snapshot.snapshot import get_snapshots
TEST_TIME = 45
REPEAT = 5
class TestCamerad(unittest.TestCase):
class TestCamerad:
@classmethod
def setUpClass(cls):
def setup_class(cls):
pass
def _numpy_rgb2gray(self, im):
@@ -49,7 +47,4 @@ class TestCamerad(unittest.TestCase):
passed += int(res)
time.sleep(2)
self.assertGreaterEqual(passed, REPEAT)
if __name__ == "__main__":
unittest.main()
assert passed >= REPEAT
+1 -1
View File
@@ -5,7 +5,7 @@
#include <map>
#include <string>
#include "cereal/messaging/messaging.h"
#include "cereal/gen/cpp/log.capnp.h"
// no-op base hw class
class HardwareNone {
+38
View File
@@ -0,0 +1,38 @@
#!/usr/bin/env python3
from abc import ABC, abstractmethod
from openpilot.common.realtime import DT_HW
from openpilot.common.numpy_fast import interp
from openpilot.common.swaglog import cloudlog
from openpilot.selfdrive.controls.lib.pid import PIDController
class BaseFanController(ABC):
@abstractmethod
def update(self, cur_temp: float, ignition: bool) -> int:
pass
class TiciFanController(BaseFanController):
def __init__(self) -> None:
super().__init__()
cloudlog.info("Setting up TICI fan handler")
self.last_ignition = False
self.controller = PIDController(k_p=0, k_i=4e-3, k_f=1, rate=(1 / DT_HW))
def update(self, cur_temp: float, ignition: bool) -> int:
self.controller.neg_limit = -(100 if ignition else 30)
self.controller.pos_limit = -(30 if ignition else 0)
if ignition != self.last_ignition:
self.controller.reset()
error = 70 - cur_temp
fan_pwr_out = -int(self.controller.update(
error=error,
feedforward=interp(cur_temp, [60.0, 100.0], [0, -100])
))
self.last_ignition = ignition
return fan_pwr_out
+474
View File
@@ -0,0 +1,474 @@
#!/usr/bin/env python3
import os
import json
import queue
import threading
import time
from collections import OrderedDict, namedtuple
from pathlib import Path
import psutil
import cereal.messaging as messaging
from cereal import log
from cereal.services import SERVICE_LIST
from openpilot.common.dict_helpers import strip_deprecated_keys
from openpilot.common.filter_simple import FirstOrderFilter
from openpilot.common.params import Params
from openpilot.common.realtime import DT_HW
from openpilot.selfdrive.controls.lib.alertmanager import set_offroad_alert
from openpilot.system.hardware import HARDWARE, TICI, AGNOS
from openpilot.system.loggerd.config import get_available_percent
from openpilot.system.statsd import statlog
from openpilot.common.swaglog import cloudlog
from openpilot.system.hardware.power_monitoring import PowerMonitoring
from openpilot.system.hardware.fan_controller import TiciFanController
from openpilot.system.version import terms_version, training_version
ThermalStatus = log.DeviceState.ThermalStatus
NetworkType = log.DeviceState.NetworkType
NetworkStrength = log.DeviceState.NetworkStrength
CURRENT_TAU = 15. # 15s time constant
TEMP_TAU = 5. # 5s time constant
DISCONNECT_TIMEOUT = 5. # wait 5 seconds before going offroad after disconnect so you get an alert
PANDA_STATES_TIMEOUT = round(1000 / SERVICE_LIST['pandaStates'].frequency * 1.5) # 1.5x the expected pandaState frequency
ThermalBand = namedtuple("ThermalBand", ['min_temp', 'max_temp'])
HardwareState = namedtuple("HardwareState", ['network_type', 'network_info', 'network_strength', 'network_stats',
'network_metered', 'nvme_temps', 'modem_temps'])
# List of thermal bands. We will stay within this region as long as we are within the bounds.
# When exiting the bounds, we'll jump to the lower or higher band. Bands are ordered in the dict.
THERMAL_BANDS = OrderedDict({
ThermalStatus.green: ThermalBand(None, 80.0),
ThermalStatus.yellow: ThermalBand(75.0, 96.0),
ThermalStatus.red: ThermalBand(88.0, 107.),
ThermalStatus.danger: ThermalBand(94.0, None),
})
# Override to highest thermal band when offroad and above this temp
OFFROAD_DANGER_TEMP = 75
prev_offroad_states: dict[str, tuple[bool, str | None]] = {}
tz_by_type: dict[str, int] | None = None
def populate_tz_by_type():
global tz_by_type
tz_by_type = {}
for n in os.listdir("/sys/devices/virtual/thermal"):
if not n.startswith("thermal_zone"):
continue
with open(os.path.join("/sys/devices/virtual/thermal", n, "type")) as f:
tz_by_type[f.read().strip()] = int(n.removeprefix("thermal_zone"))
def read_tz(x):
if x is None:
return 0
if isinstance(x, str):
if tz_by_type is None:
populate_tz_by_type()
x = tz_by_type[x]
try:
with open(f"/sys/devices/virtual/thermal/thermal_zone{x}/temp") as f:
return int(f.read())
except FileNotFoundError:
return 0
def read_thermal(thermal_config):
dat = messaging.new_message('deviceState', valid=True)
dat.deviceState.cpuTempC = [read_tz(z) / thermal_config.cpu[1] for z in thermal_config.cpu[0]]
dat.deviceState.gpuTempC = [read_tz(z) / thermal_config.gpu[1] for z in thermal_config.gpu[0]]
dat.deviceState.memoryTempC = read_tz(thermal_config.mem[0]) / thermal_config.mem[1]
dat.deviceState.pmicTempC = [read_tz(z) / thermal_config.pmic[1] for z in thermal_config.pmic[0]]
return dat
def set_offroad_alert_if_changed(offroad_alert: str, show_alert: bool, extra_text: str | None=None):
if prev_offroad_states.get(offroad_alert, None) == (show_alert, extra_text):
return
prev_offroad_states[offroad_alert] = (show_alert, extra_text)
set_offroad_alert(offroad_alert, show_alert, extra_text)
def hw_state_thread(end_event, hw_queue):
"""Handles non critical hardware state, and sends over queue"""
count = 0
prev_hw_state = None
modem_version = None
modem_nv = None
modem_configured = False
modem_restarted = False
modem_missing_count = 0
while not end_event.is_set():
# these are expensive calls. update every 10s
if (count % int(10. / DT_HW)) == 0:
try:
network_type = HARDWARE.get_network_type()
modem_temps = HARDWARE.get_modem_temperatures()
if len(modem_temps) == 0 and prev_hw_state is not None:
modem_temps = prev_hw_state.modem_temps
# Log modem version once
if AGNOS and ((modem_version is None) or (modem_nv is None)):
modem_version = HARDWARE.get_modem_version()
modem_nv = HARDWARE.get_modem_nv()
if (modem_version is not None) and (modem_nv is not None):
cloudlog.event("modem version", version=modem_version, nv=modem_nv)
else:
if not modem_restarted:
# TODO: we may be able to remove this with a MM update
# ModemManager's probing on startup can fail
# rarely, restart the service to probe again.
modem_missing_count += 1
if modem_missing_count > 3:
modem_restarted = True
cloudlog.event("restarting ModemManager")
os.system("sudo systemctl restart --no-block ModemManager")
tx, rx = HARDWARE.get_modem_data_usage()
hw_state = HardwareState(
network_type=network_type,
network_info=HARDWARE.get_network_info(),
network_strength=HARDWARE.get_network_strength(network_type),
network_stats={'wwanTx': tx, 'wwanRx': rx},
network_metered=HARDWARE.get_network_metered(network_type),
nvme_temps=HARDWARE.get_nvme_temperatures(),
modem_temps=modem_temps,
)
try:
hw_queue.put_nowait(hw_state)
except queue.Full:
pass
# TODO: remove this once the config is in AGNOS
if not modem_configured and len(HARDWARE.get_sim_info().get('sim_id', '')) > 0:
cloudlog.warning("configuring modem")
HARDWARE.configure_modem()
modem_configured = True
prev_hw_state = hw_state
except Exception:
cloudlog.exception("Error getting hardware state")
count += 1
time.sleep(DT_HW)
def hardware_thread(end_event, hw_queue) -> None:
pm = messaging.PubMaster(['deviceState'])
sm = messaging.SubMaster(["peripheralState", "gpsLocationExternal", "controlsState", "pandaStates"], poll="pandaStates")
count = 0
onroad_conditions: dict[str, bool] = {
"ignition": False,
}
startup_conditions: dict[str, bool] = {}
startup_conditions_prev: dict[str, bool] = {}
off_ts: float | None = None
started_ts: float | None = None
started_seen = False
startup_blocked_ts: float | None = None
thermal_status = ThermalStatus.yellow
last_hw_state = HardwareState(
network_type=NetworkType.none,
network_info=None,
network_metered=False,
network_strength=NetworkStrength.unknown,
network_stats={'wwanTx': -1, 'wwanRx': -1},
nvme_temps=[],
modem_temps=[],
)
all_temp_filter = FirstOrderFilter(0., TEMP_TAU, DT_HW, initialized=False)
offroad_temp_filter = FirstOrderFilter(0., TEMP_TAU, DT_HW, initialized=False)
should_start_prev = False
in_car = False
engaged_prev = False
params = Params()
power_monitor = PowerMonitoring()
HARDWARE.initialize_hardware()
thermal_config = HARDWARE.get_thermal_config()
fan_controller = None
while not end_event.is_set():
sm.update(PANDA_STATES_TIMEOUT)
pandaStates = sm['pandaStates']
peripheralState = sm['peripheralState']
peripheral_panda_present = peripheralState.pandaType != log.PandaState.PandaType.unknown
if sm.updated['pandaStates'] and len(pandaStates) > 0:
# Set ignition based on any panda connected
onroad_conditions["ignition"] = any(ps.ignitionLine or ps.ignitionCan for ps in pandaStates if ps.pandaType != log.PandaState.PandaType.unknown)
pandaState = pandaStates[0]
in_car = pandaState.harnessStatus != log.PandaState.HarnessStatus.notConnected
# Setup fan handler on first connect to panda
if fan_controller is None and peripheral_panda_present:
if TICI:
fan_controller = TiciFanController()
elif (time.monotonic() - sm.recv_time['pandaStates']) > DISCONNECT_TIMEOUT:
if onroad_conditions["ignition"]:
onroad_conditions["ignition"] = False
cloudlog.error("panda timed out onroad")
# Run at 2Hz, plus rising edge of ignition
ign_edge = started_ts is None and onroad_conditions["ignition"]
if (sm.frame % round(SERVICE_LIST['pandaStates'].frequency * DT_HW) != 0) and not ign_edge:
continue
msg = read_thermal(thermal_config)
msg.deviceState.deviceType = HARDWARE.get_device_type()
try:
last_hw_state = hw_queue.get_nowait()
except queue.Empty:
pass
msg.deviceState.freeSpacePercent = get_available_percent(default=100.0)
msg.deviceState.memoryUsagePercent = int(round(psutil.virtual_memory().percent))
msg.deviceState.gpuUsagePercent = int(round(HARDWARE.get_gpu_usage_percent()))
online_cpu_usage = [int(round(n)) for n in psutil.cpu_percent(percpu=True)]
offline_cpu_usage = [0., ] * (len(msg.deviceState.cpuTempC) - len(online_cpu_usage))
msg.deviceState.cpuUsagePercent = online_cpu_usage + offline_cpu_usage
msg.deviceState.networkType = last_hw_state.network_type
msg.deviceState.networkMetered = last_hw_state.network_metered
msg.deviceState.networkStrength = last_hw_state.network_strength
msg.deviceState.networkStats = last_hw_state.network_stats
if last_hw_state.network_info is not None:
msg.deviceState.networkInfo = last_hw_state.network_info
msg.deviceState.nvmeTempC = last_hw_state.nvme_temps
msg.deviceState.modemTempC = last_hw_state.modem_temps
msg.deviceState.screenBrightnessPercent = HARDWARE.get_screen_brightness()
# this subset is only used for offroad
temp_sources = [
msg.deviceState.memoryTempC,
max(msg.deviceState.cpuTempC),
max(msg.deviceState.gpuTempC),
]
offroad_comp_temp = offroad_temp_filter.update(max(temp_sources))
# this drives the thermal status while onroad
temp_sources.append(max(msg.deviceState.pmicTempC))
all_comp_temp = all_temp_filter.update(max(temp_sources))
msg.deviceState.maxTempC = all_comp_temp
if fan_controller is not None:
msg.deviceState.fanSpeedPercentDesired = fan_controller.update(all_comp_temp, onroad_conditions["ignition"])
is_offroad_for_5_min = (started_ts is None) and ((not started_seen) or (off_ts is None) or (time.monotonic() - off_ts > 60 * 5))
if is_offroad_for_5_min and offroad_comp_temp > OFFROAD_DANGER_TEMP:
# if device is offroad and already hot without the extra onroad load,
# we want to cool down first before increasing load
thermal_status = ThermalStatus.danger
else:
current_band = THERMAL_BANDS[thermal_status]
band_idx = list(THERMAL_BANDS.keys()).index(thermal_status)
if current_band.min_temp is not None and all_comp_temp < current_band.min_temp:
thermal_status = list(THERMAL_BANDS.keys())[band_idx - 1]
elif current_band.max_temp is not None and all_comp_temp > current_band.max_temp:
thermal_status = list(THERMAL_BANDS.keys())[band_idx + 1]
# **** starting logic ****
startup_conditions["up_to_date"] = params.get("Offroad_ConnectivityNeeded") is None or params.get_bool("DisableUpdates") or params.get_bool("SnoozeUpdate")
startup_conditions["not_uninstalling"] = not params.get_bool("DoUninstall")
startup_conditions["accepted_terms"] = params.get("HasAcceptedTerms") == terms_version
# with 2% left, we killall, otherwise the phone will take a long time to boot
startup_conditions["free_space"] = msg.deviceState.freeSpacePercent > 2
startup_conditions["completed_training"] = params.get("CompletedTrainingVersion") == training_version
startup_conditions["not_driver_view"] = not params.get_bool("IsDriverViewEnabled")
startup_conditions["not_taking_snapshot"] = not params.get_bool("IsTakingSnapshot")
# must be at an engageable thermal band to go onroad
startup_conditions["device_temp_engageable"] = thermal_status < ThermalStatus.red
# ensure device is fully booted
startup_conditions["device_booted"] = startup_conditions.get("device_booted", False) or HARDWARE.booted()
# if the temperature enters the danger zone, go offroad to cool down
onroad_conditions["device_temp_good"] = thermal_status < ThermalStatus.danger
extra_text = f"{offroad_comp_temp:.1f}C"
show_alert = (not onroad_conditions["device_temp_good"] or not startup_conditions["device_temp_engageable"]) and onroad_conditions["ignition"]
set_offroad_alert_if_changed("Offroad_TemperatureTooHigh", show_alert, extra_text=extra_text)
# TODO: this should move to TICI.initialize_hardware, but we currently can't import params there
if TICI:
if not os.path.isfile("/persist/comma/living-in-the-moment"):
if not Path("/data/media").is_mount():
set_offroad_alert_if_changed("Offroad_StorageMissing", True)
else:
# check for bad NVMe
try:
with open("/sys/block/nvme0n1/device/model") as f:
model = f.read().strip()
if not model.startswith("Samsung SSD 980") and params.get("Offroad_BadNvme") is None:
set_offroad_alert_if_changed("Offroad_BadNvme", True)
cloudlog.event("Unsupported NVMe", model=model, error=True)
except Exception:
pass
# Handle offroad/onroad transition
should_start = all(onroad_conditions.values())
if started_ts is None:
should_start = should_start and all(startup_conditions.values())
if should_start != should_start_prev or (count == 0):
params.put_bool("IsEngaged", False)
engaged_prev = False
HARDWARE.set_power_save(not should_start)
if sm.updated['controlsState']:
engaged = sm['controlsState'].enabled
if engaged != engaged_prev:
params.put_bool("IsEngaged", engaged)
engaged_prev = engaged
try:
with open('/dev/kmsg', 'w') as kmsg:
kmsg.write(f"<3>[hardware] engaged: {engaged}\n")
except Exception:
pass
if should_start:
off_ts = None
if started_ts is None:
started_ts = time.monotonic()
started_seen = True
if startup_blocked_ts is not None:
cloudlog.event("Startup after block", block_duration=(time.monotonic() - startup_blocked_ts),
startup_conditions=startup_conditions, onroad_conditions=onroad_conditions,
startup_conditions_prev=startup_conditions_prev, error=True)
startup_blocked_ts = None
else:
if onroad_conditions["ignition"] and (startup_conditions != startup_conditions_prev):
cloudlog.event("Startup blocked", startup_conditions=startup_conditions, onroad_conditions=onroad_conditions, error=True)
startup_conditions_prev = startup_conditions.copy()
startup_blocked_ts = time.monotonic()
started_ts = None
if off_ts is None:
off_ts = time.monotonic()
# Offroad power monitoring
voltage = None if peripheralState.pandaType == log.PandaState.PandaType.unknown else peripheralState.voltage
power_monitor.calculate(voltage, onroad_conditions["ignition"])
msg.deviceState.offroadPowerUsageUwh = power_monitor.get_power_used()
msg.deviceState.carBatteryCapacityUwh = max(0, power_monitor.get_car_battery_capacity())
current_power_draw = HARDWARE.get_current_power_draw()
statlog.sample("power_draw", current_power_draw)
msg.deviceState.powerDrawW = current_power_draw
som_power_draw = HARDWARE.get_som_power_draw()
statlog.sample("som_power_draw", som_power_draw)
msg.deviceState.somPowerDrawW = som_power_draw
# Check if we need to shut down
if power_monitor.should_shutdown(onroad_conditions["ignition"], in_car, off_ts, started_seen):
cloudlog.warning(f"shutting device down, offroad since {off_ts}")
params.put_bool("DoShutdown", True)
msg.deviceState.started = started_ts is not None
msg.deviceState.startedMonoTime = int(1e9*(started_ts or 0))
last_ping = params.get("LastAthenaPingTime")
if last_ping is not None:
msg.deviceState.lastAthenaPingTime = int(last_ping)
msg.deviceState.thermalStatus = thermal_status
pm.send("deviceState", msg)
# Log to statsd
statlog.gauge("free_space_percent", msg.deviceState.freeSpacePercent)
statlog.gauge("gpu_usage_percent", msg.deviceState.gpuUsagePercent)
statlog.gauge("memory_usage_percent", msg.deviceState.memoryUsagePercent)
for i, usage in enumerate(msg.deviceState.cpuUsagePercent):
statlog.gauge(f"cpu{i}_usage_percent", usage)
for i, temp in enumerate(msg.deviceState.cpuTempC):
statlog.gauge(f"cpu{i}_temperature", temp)
for i, temp in enumerate(msg.deviceState.gpuTempC):
statlog.gauge(f"gpu{i}_temperature", temp)
statlog.gauge("memory_temperature", msg.deviceState.memoryTempC)
for i, temp in enumerate(msg.deviceState.pmicTempC):
statlog.gauge(f"pmic{i}_temperature", temp)
for i, temp in enumerate(last_hw_state.nvme_temps):
statlog.gauge(f"nvme_temperature{i}", temp)
for i, temp in enumerate(last_hw_state.modem_temps):
statlog.gauge(f"modem_temperature{i}", temp)
statlog.gauge("fan_speed_percent_desired", msg.deviceState.fanSpeedPercentDesired)
statlog.gauge("screen_brightness_percent", msg.deviceState.screenBrightnessPercent)
# report to server once every 10 minutes
rising_edge_started = should_start and not should_start_prev
if rising_edge_started or (count % int(600. / DT_HW)) == 0:
dat = {
'count': count,
'pandaStates': [strip_deprecated_keys(p.to_dict()) for p in pandaStates],
'peripheralState': strip_deprecated_keys(peripheralState.to_dict()),
'location': (strip_deprecated_keys(sm["gpsLocationExternal"].to_dict()) if sm.alive["gpsLocationExternal"] else None),
'deviceState': strip_deprecated_keys(msg.to_dict())
}
cloudlog.event("STATUS_PACKET", **dat)
# save last one before going onroad
if rising_edge_started:
try:
params.put("LastOffroadStatusPacket", json.dumps(dat))
except Exception:
cloudlog.exception("failed to save offroad status")
params.put_bool_nonblocking("NetworkMetered", msg.deviceState.networkMetered)
count += 1
should_start_prev = should_start
def main():
hw_queue = queue.Queue(maxsize=1)
end_event = threading.Event()
threads = [
threading.Thread(target=hw_state_thread, args=(end_event, hw_queue)),
threading.Thread(target=hardware_thread, args=(end_event, hw_queue)),
]
for t in threads:
t.start()
try:
while True:
time.sleep(1)
if not all(t.is_alive() for t in threads):
break
finally:
end_event.set()
for t in threads:
t.join()
if __name__ == "__main__":
main()
+128
View File
@@ -0,0 +1,128 @@
import time
import threading
from openpilot.common.params import Params
from openpilot.system.hardware import HARDWARE
from openpilot.common.swaglog import cloudlog
from openpilot.system.statsd import statlog
CAR_VOLTAGE_LOW_PASS_K = 0.011 # LPF gain for 45s tau (dt/tau / (dt/tau + 1))
# While driving, a battery charges completely in about 30-60 minutes
CAR_BATTERY_CAPACITY_uWh = 30e6
CAR_CHARGING_RATE_W = 45
VBATT_PAUSE_CHARGING = 11.8 # Lower limit on the LPF car battery voltage
MAX_TIME_OFFROAD_S = 30*3600
MIN_ON_TIME_S = 3600
DELAY_SHUTDOWN_TIME_S = 300 # Wait at least DELAY_SHUTDOWN_TIME_S seconds after offroad_time to shutdown.
VOLTAGE_SHUTDOWN_MIN_OFFROAD_TIME_S = 60
class PowerMonitoring:
def __init__(self):
self.params = Params()
self.last_measurement_time = None # Used for integration delta
self.last_save_time = 0 # Used for saving current value in a param
self.power_used_uWh = 0 # Integrated power usage in uWh since going into offroad
self.next_pulsed_measurement_time = None
self.car_voltage_mV = 12e3 # Low-passed version of peripheralState voltage
self.car_voltage_instant_mV = 12e3 # Last value of peripheralState voltage
self.integration_lock = threading.Lock()
car_battery_capacity_uWh = self.params.get("CarBatteryCapacity")
if car_battery_capacity_uWh is None:
car_battery_capacity_uWh = 0
# Reset capacity if it's low
self.car_battery_capacity_uWh = max((CAR_BATTERY_CAPACITY_uWh / 10), int(car_battery_capacity_uWh))
# Calculation tick
def calculate(self, voltage: int | None, ignition: bool):
try:
now = time.monotonic()
# If peripheralState is None, we're probably not in a car, so we don't care
if voltage is None:
with self.integration_lock:
self.last_measurement_time = None
self.next_pulsed_measurement_time = None
self.power_used_uWh = 0
return
# Low-pass battery voltage
self.car_voltage_instant_mV = voltage
self.car_voltage_mV = ((voltage * CAR_VOLTAGE_LOW_PASS_K) + (self.car_voltage_mV * (1 - CAR_VOLTAGE_LOW_PASS_K)))
statlog.gauge("car_voltage", self.car_voltage_mV / 1e3)
# Cap the car battery power and save it in a param every 10-ish seconds
self.car_battery_capacity_uWh = max(self.car_battery_capacity_uWh, 0)
self.car_battery_capacity_uWh = min(self.car_battery_capacity_uWh, CAR_BATTERY_CAPACITY_uWh)
if now - self.last_save_time >= 10:
self.params.put_nonblocking("CarBatteryCapacity", str(int(self.car_battery_capacity_uWh)))
self.last_save_time = now
# First measurement, set integration time
with self.integration_lock:
if self.last_measurement_time is None:
self.last_measurement_time = now
return
if ignition:
# If there is ignition, we integrate the charging rate of the car
with self.integration_lock:
self.power_used_uWh = 0
integration_time_h = (now - self.last_measurement_time) / 3600
if integration_time_h < 0:
raise ValueError(f"Negative integration time: {integration_time_h}h")
self.car_battery_capacity_uWh += (CAR_CHARGING_RATE_W * 1e6 * integration_time_h)
self.last_measurement_time = now
else:
# Get current power draw somehow
current_power = HARDWARE.get_current_power_draw()
# Do the integration
self._perform_integration(now, current_power)
except Exception:
cloudlog.exception("Power monitoring calculation failed")
def _perform_integration(self, t: float, current_power: float) -> None:
with self.integration_lock:
try:
if self.last_measurement_time:
integration_time_h = (t - self.last_measurement_time) / 3600
power_used = (current_power * 1000000) * integration_time_h
if power_used < 0:
raise ValueError(f"Negative power used! Integration time: {integration_time_h} h Current Power: {power_used} uWh")
self.power_used_uWh += power_used
self.car_battery_capacity_uWh -= power_used
self.last_measurement_time = t
except Exception:
cloudlog.exception("Integration failed")
# Get the power usage
def get_power_used(self) -> int:
return int(self.power_used_uWh)
def get_car_battery_capacity(self) -> int:
return int(self.car_battery_capacity_uWh)
# See if we need to shutdown
def should_shutdown(self, ignition: bool, in_car: bool, offroad_timestamp: float | None, started_seen: bool):
if offroad_timestamp is None:
return False
now = time.monotonic()
should_shutdown = False
offroad_time = (now - offroad_timestamp)
low_voltage_shutdown = (self.car_voltage_mV < (VBATT_PAUSE_CHARGING * 1e3) and
offroad_time > VOLTAGE_SHUTDOWN_MIN_OFFROAD_TIME_S)
should_shutdown |= offroad_time > MAX_TIME_OFFROAD_S
should_shutdown |= low_voltage_shutdown
should_shutdown |= (self.car_battery_capacity_uWh <= 0)
should_shutdown &= not ignition
should_shutdown &= (not self.params.get_bool("DisablePowerDown"))
should_shutdown &= in_car
should_shutdown &= offroad_time > DELAY_SHUTDOWN_TIME_S
should_shutdown |= self.params.get_bool("ForcePowerDown")
should_shutdown &= started_seen or (now > MIN_ON_TIME_S)
return should_shutdown
View File
@@ -0,0 +1,50 @@
import pytest
from openpilot.system.hardware.fan_controller import TiciFanController
ALL_CONTROLLERS = [TiciFanController]
def patched_controller(mocker, controller_class):
mocker.patch("os.system", new=mocker.Mock())
return controller_class()
class TestFanController:
def wind_up(self, controller, ignition=True):
for _ in range(1000):
controller.update(100, ignition)
def wind_down(self, controller, ignition=False):
for _ in range(1000):
controller.update(10, ignition)
@pytest.mark.parametrize("controller_class", ALL_CONTROLLERS)
def test_hot_onroad(self, mocker, controller_class):
controller = patched_controller(mocker, controller_class)
self.wind_up(controller)
assert controller.update(100, True) >= 70
@pytest.mark.parametrize("controller_class", ALL_CONTROLLERS)
def test_offroad_limits(self, mocker, controller_class):
controller = patched_controller(mocker, controller_class)
self.wind_up(controller)
assert controller.update(100, False) <= 30
@pytest.mark.parametrize("controller_class", ALL_CONTROLLERS)
def test_no_fan_wear(self, mocker, controller_class):
controller = patched_controller(mocker, controller_class)
self.wind_down(controller)
assert controller.update(10, False) == 0
@pytest.mark.parametrize("controller_class", ALL_CONTROLLERS)
def test_limited(self, mocker, controller_class):
controller = patched_controller(mocker, controller_class)
self.wind_up(controller, True)
assert controller.update(100, True) == 100
@pytest.mark.parametrize("controller_class", ALL_CONTROLLERS)
def test_windup_speed(self, mocker, controller_class):
controller = patched_controller(mocker, controller_class)
self.wind_down(controller, True)
for _ in range(10):
controller.update(90, True)
assert controller.update(90, True) >= 60
@@ -0,0 +1,199 @@
import pytest
from openpilot.common.params import Params
from openpilot.system.hardware.power_monitoring import PowerMonitoring, CAR_BATTERY_CAPACITY_uWh, \
CAR_CHARGING_RATE_W, VBATT_PAUSE_CHARGING, DELAY_SHUTDOWN_TIME_S
# Create fake time
ssb = 0.
def mock_time_monotonic():
global ssb
ssb += 1.
return ssb
TEST_DURATION_S = 50
GOOD_VOLTAGE = 12 * 1e3
VOLTAGE_BELOW_PAUSE_CHARGING = (VBATT_PAUSE_CHARGING - 1) * 1e3
def pm_patch(mocker, name, value, constant=False):
if constant:
mocker.patch(f"openpilot.system.hardware.power_monitoring.{name}", value)
else:
mocker.patch(f"openpilot.system.hardware.power_monitoring.{name}", return_value=value)
@pytest.fixture(autouse=True)
def mock_time(mocker):
mocker.patch("time.monotonic", mock_time_monotonic)
class TestPowerMonitoring:
def setup_method(self):
self.params = Params()
# Test to see that it doesn't do anything when pandaState is None
def test_panda_state_present(self):
pm = PowerMonitoring()
for _ in range(10):
pm.calculate(None, None)
assert pm.get_power_used() == 0
assert pm.get_car_battery_capacity() == (CAR_BATTERY_CAPACITY_uWh / 10)
# Test to see that it doesn't integrate offroad when ignition is True
def test_offroad_ignition(self):
pm = PowerMonitoring()
for _ in range(10):
pm.calculate(GOOD_VOLTAGE, True)
assert pm.get_power_used() == 0
# Test to see that it integrates with discharging battery
def test_offroad_integration_discharging(self, mocker):
POWER_DRAW = 4
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
for _ in range(TEST_DURATION_S + 1):
pm.calculate(GOOD_VOLTAGE, False)
expected_power_usage = ((TEST_DURATION_S/3600) * POWER_DRAW * 1e6)
assert abs(pm.get_power_used() - expected_power_usage) < 10
# Test to check positive integration of car_battery_capacity
def test_car_battery_integration_onroad(self, mocker):
POWER_DRAW = 4
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = 0
for _ in range(TEST_DURATION_S + 1):
pm.calculate(GOOD_VOLTAGE, True)
expected_capacity = ((TEST_DURATION_S/3600) * CAR_CHARGING_RATE_W * 1e6)
assert abs(pm.get_car_battery_capacity() - expected_capacity) < 10
# Test to check positive integration upper limit
def test_car_battery_integration_upper_limit(self, mocker):
POWER_DRAW = 4
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = CAR_BATTERY_CAPACITY_uWh - 1000
for _ in range(TEST_DURATION_S + 1):
pm.calculate(GOOD_VOLTAGE, True)
estimated_capacity = CAR_BATTERY_CAPACITY_uWh + (CAR_CHARGING_RATE_W / 3600 * 1e6)
assert abs(pm.get_car_battery_capacity() - estimated_capacity) < 10
# Test to check negative integration of car_battery_capacity
def test_car_battery_integration_offroad(self, mocker):
POWER_DRAW = 4
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = CAR_BATTERY_CAPACITY_uWh
for _ in range(TEST_DURATION_S + 1):
pm.calculate(GOOD_VOLTAGE, False)
expected_capacity = CAR_BATTERY_CAPACITY_uWh - ((TEST_DURATION_S/3600) * POWER_DRAW * 1e6)
assert abs(pm.get_car_battery_capacity() - expected_capacity) < 10
# Test to check negative integration lower limit
def test_car_battery_integration_lower_limit(self, mocker):
POWER_DRAW = 4
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = 1000
for _ in range(TEST_DURATION_S + 1):
pm.calculate(GOOD_VOLTAGE, False)
estimated_capacity = 0 - ((1/3600) * POWER_DRAW * 1e6)
assert abs(pm.get_car_battery_capacity() - estimated_capacity) < 10
# Test to check policy of stopping charging after MAX_TIME_OFFROAD_S
def test_max_time_offroad(self, mocker):
MOCKED_MAX_OFFROAD_TIME = 3600
POWER_DRAW = 0 # To stop shutting down for other reasons
pm_patch(mocker, "MAX_TIME_OFFROAD_S", MOCKED_MAX_OFFROAD_TIME, constant=True)
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = CAR_BATTERY_CAPACITY_uWh
start_time = ssb
ignition = False
while ssb <= start_time + MOCKED_MAX_OFFROAD_TIME:
pm.calculate(GOOD_VOLTAGE, ignition)
if (ssb - start_time) % 1000 == 0 and ssb < start_time + MOCKED_MAX_OFFROAD_TIME:
assert not pm.should_shutdown(ignition, True, start_time, False)
assert pm.should_shutdown(ignition, True, start_time, False)
def test_car_voltage(self, mocker):
POWER_DRAW = 0 # To stop shutting down for other reasons
TEST_TIME = 350
VOLTAGE_SHUTDOWN_MIN_OFFROAD_TIME_S = 50
pm_patch(mocker, "VOLTAGE_SHUTDOWN_MIN_OFFROAD_TIME_S", VOLTAGE_SHUTDOWN_MIN_OFFROAD_TIME_S, constant=True)
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = CAR_BATTERY_CAPACITY_uWh
ignition = False
start_time = ssb
for i in range(TEST_TIME):
pm.calculate(VOLTAGE_BELOW_PAUSE_CHARGING, ignition)
if i % 10 == 0:
assert pm.should_shutdown(ignition, True, start_time, True) == \
(pm.car_voltage_mV < VBATT_PAUSE_CHARGING * 1e3 and \
(ssb - start_time) > VOLTAGE_SHUTDOWN_MIN_OFFROAD_TIME_S and \
(ssb - start_time) > DELAY_SHUTDOWN_TIME_S)
assert pm.should_shutdown(ignition, True, start_time, True)
# Test to check policy of not stopping charging when DisablePowerDown is set
def test_disable_power_down(self, mocker):
POWER_DRAW = 0 # To stop shutting down for other reasons
TEST_TIME = 100
self.params.put_bool("DisablePowerDown", True)
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = CAR_BATTERY_CAPACITY_uWh
ignition = False
for i in range(TEST_TIME):
pm.calculate(VOLTAGE_BELOW_PAUSE_CHARGING, ignition)
if i % 10 == 0:
assert not pm.should_shutdown(ignition, True, ssb, False)
assert not pm.should_shutdown(ignition, True, ssb, False)
# Test to check policy of not stopping charging when ignition
def test_ignition(self, mocker):
POWER_DRAW = 0 # To stop shutting down for other reasons
TEST_TIME = 100
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = CAR_BATTERY_CAPACITY_uWh
ignition = True
for i in range(TEST_TIME):
pm.calculate(VOLTAGE_BELOW_PAUSE_CHARGING, ignition)
if i % 10 == 0:
assert not pm.should_shutdown(ignition, True, ssb, False)
assert not pm.should_shutdown(ignition, True, ssb, False)
# Test to check policy of not stopping charging when harness is not connected
def test_harness_connection(self, mocker):
POWER_DRAW = 0 # To stop shutting down for other reasons
TEST_TIME = 100
pm_patch(mocker, "HARDWARE.get_current_power_draw", POWER_DRAW)
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = CAR_BATTERY_CAPACITY_uWh
ignition = False
for i in range(TEST_TIME):
pm.calculate(VOLTAGE_BELOW_PAUSE_CHARGING, ignition)
if i % 10 == 0:
assert not pm.should_shutdown(ignition, False, ssb, False)
assert not pm.should_shutdown(ignition, False, ssb, False)
def test_delay_shutdown_time(self):
pm = PowerMonitoring()
pm.car_battery_capacity_uWh = 0
ignition = False
in_car = True
offroad_timestamp = ssb
started_seen = True
pm.calculate(VOLTAGE_BELOW_PAUSE_CHARGING, ignition)
while ssb < offroad_timestamp + DELAY_SHUTDOWN_TIME_S:
assert not pm.should_shutdown(ignition, in_car,
offroad_timestamp,
started_seen), \
f"Should not shutdown before {DELAY_SHUTDOWN_TIME_S} seconds offroad time"
assert pm.should_shutdown(ignition, in_car,
offroad_timestamp,
started_seen), \
f"Should shutdown after {DELAY_SHUTDOWN_TIME_S} seconds offroad time"
+17 -9
View File
@@ -379,23 +379,19 @@ class Tici(HardwareBase):
# *** CPU config ***
# offline big cluster, leave core 4 online for boardd
for i in range(5, 8):
# offline big cluster, leave core 4 online for pandad
for i in range(4, 8):
val = '0' if powersave_enabled else '1'
sudo_write(val, f'/sys/devices/system/cpu/cpu{i}/online')
for n in ('0', '4'):
if powersave_enabled and n == '4':
continue
gov = 'ondemand' if powersave_enabled else 'performance'
sudo_write(gov, f'/sys/devices/system/cpu/cpufreq/policy{n}/scaling_governor')
# *** IRQ config ***
# boardd core
affine_irq(4, "spi_geni") # SPI
affine_irq(4, "xhci-hcd:usb3") # aux panda USB (or potentially anything else on USB)
if "tici" in self.get_device_type():
affine_irq(4, "xhci-hcd:usb1") # internal panda USB (also modem)
# GPU
affine_irq(5, "kgsl-3d0")
@@ -416,7 +412,7 @@ class Tici(HardwareBase):
if self.amplifier is not None:
self.amplifier.initialize_configuration(self.get_device_type())
# Allow thermald to write engagement status to kmsg
# Allow hardwared to write engagement status to kmsg
os.system("sudo chmod a+w /dev/kmsg")
# Ensure fan gpio is enabled so fan runs until shutdown, also turned on at boot by the ABL
@@ -453,6 +449,18 @@ class Tici(HardwareBase):
sudo_write("N", "/sys/kernel/debug/msm_vidc/clock_scaling")
sudo_write("Y", "/sys/kernel/debug/msm_vidc/disable_thermal_mitigation")
# pandad core
affine_irq(3, "spi_geni") # SPI
if "tici" in self.get_device_type():
affine_irq(3, "xhci-hcd:usb3") # aux panda USB (or potentially anything else on USB)
affine_irq(3, "xhci-hcd:usb1") # internal panda USB (also modem)
try:
pid = subprocess.check_output(["pgrep", "-f", "spi0"], encoding='utf8').strip()
subprocess.call(["sudo", "chrt", "-f", "-p", "1", pid])
subprocess.call(["sudo", "taskset", "-pc", "3", pid])
except subprocess.CalledProcessException as e:
print(str(e))
def configure_modem(self):
sim_id = self.get_sim_info().get('sim_id', '')
+2 -8
View File
@@ -1,14 +1,12 @@
#!/usr/bin/env python3
import json
import os
import unittest
import requests
TEST_DIR = os.path.join(os.path.dirname(os.path.abspath(__file__)))
MANIFEST = os.path.join(TEST_DIR, "../agnos.json")
class TestAgnosUpdater(unittest.TestCase):
class TestAgnosUpdater:
def test_manifest(self):
with open(MANIFEST) as f:
@@ -17,10 +15,6 @@ class TestAgnosUpdater(unittest.TestCase):
for img in m:
r = requests.head(img['url'], timeout=10)
r.raise_for_status()
self.assertEqual(r.headers['Content-Type'], "application/x-xz")
assert r.headers['Content-Type'] == "application/x-xz"
if not img['sparse']:
assert img['hash'] == img['hash_raw']
if __name__ == "__main__":
unittest.main()
+7 -12
View File
@@ -1,7 +1,6 @@
#!/usr/bin/env python3
import pytest
import time
import random
import unittest
import subprocess
from panda import Panda
@@ -10,14 +9,14 @@ from openpilot.system.hardware.tici.hardware import Tici
from openpilot.system.hardware.tici.amplifier import Amplifier
class TestAmplifier(unittest.TestCase):
class TestAmplifier:
@classmethod
def setUpClass(cls):
def setup_class(cls):
if not TICI:
raise unittest.SkipTest
pytest.skip()
def setUp(self):
def setup_method(self):
# clear dmesg
subprocess.check_call("sudo dmesg -C", shell=True)
@@ -25,7 +24,7 @@ class TestAmplifier(unittest.TestCase):
Panda.wait_for_panda(None, 30)
self.panda = Panda()
def tearDown(self):
def teardown_method(self):
HARDWARE.reset_internal_panda()
def _check_for_i2c_errors(self, expected):
@@ -68,8 +67,4 @@ class TestAmplifier(unittest.TestCase):
if self._check_for_i2c_errors(True):
break
else:
self.fail("didn't hit any i2c errors")
if __name__ == "__main__":
unittest.main()
pytest.fail("didn't hit any i2c errors")
+8 -10
View File
@@ -1,7 +1,5 @@
#!/usr/bin/env python3
import pytest
import time
import unittest
import numpy as np
from openpilot.system.hardware.tici.hardware import Tici
@@ -10,19 +8,19 @@ HARDWARE = Tici()
@pytest.mark.tici
class TestHardware(unittest.TestCase):
class TestHardware:
def test_power_save_time(self):
ts = []
ts = {True: [], False: []}
for _ in range(5):
for on in (True, False):
st = time.monotonic()
HARDWARE.set_power_save(on)
ts.append(time.monotonic() - st)
ts[on].append(time.monotonic() - st)
assert 0.1 < np.mean(ts) < 0.25
assert max(ts) < 0.3
# disabling power save is the main time-critical one
assert 0.1 < np.mean(ts[False]) < 0.15
assert max(ts[False]) < 0.2
if __name__ == "__main__":
unittest.main()
assert 0.1 < np.mean(ts[True]) < 0.35
assert max(ts[True]) < 0.4
+9 -15
View File
@@ -1,7 +1,5 @@
#!/usr/bin/env python3
from collections import defaultdict, deque
import pytest
import unittest
import time
import numpy as np
from dataclasses import dataclass
@@ -12,8 +10,8 @@ from cereal.services import SERVICE_LIST
from openpilot.common.mock import mock_messages
from openpilot.selfdrive.car.car_helpers import write_car_param
from openpilot.system.hardware.tici.power_monitor import get_power
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.selfdrive.manager.manager import manager_cleanup
from openpilot.system.manager.process_config import managed_processes
from openpilot.system.manager.manager import manager_cleanup
SAMPLE_TIME = 8 # seconds to sample power
MAX_WARMUP_TIME = 30 # seconds to wait for SAMPLE_TIME consecutive valid samples
@@ -40,15 +38,15 @@ PROCS = [
@pytest.mark.tici
class TestPowerDraw(unittest.TestCase):
class TestPowerDraw:
def setUp(self):
def setup_method(self):
write_car_param()
# wait a bit for power save to disable
time.sleep(5)
def tearDown(self):
def teardown_method(self):
manager_cleanup()
def get_expected_messages(self, proc):
@@ -97,7 +95,7 @@ class TestPowerDraw(unittest.TestCase):
return now, msg_counts, time.monotonic() - start_time - SAMPLE_TIME
@mock_messages(['liveLocationKalman'])
def test_camera_procs(self):
def test_camera_procs(self, subtests):
baseline = get_power()
prev = baseline
@@ -122,12 +120,8 @@ class TestPowerDraw(unittest.TestCase):
expected = proc.power
msgs_received = sum(msg_counts[msg] for msg in proc.msgs)
tab.append([proc.name, round(expected, 2), round(cur, 2), self.get_expected_messages(proc), msgs_received, round(warmup_time[proc.name], 2)])
with self.subTest(proc=proc.name):
self.assertTrue(self.valid_msg_count(proc, msg_counts), f"expected {self.get_expected_messages(proc)} msgs, got {msgs_received} msgs")
self.assertTrue(self.valid_power_draw(proc, cur), f"expected {expected:.2f}W, got {cur:.2f}W")
with subtests.test(proc=proc.name):
assert self.valid_msg_count(proc, msg_counts), f"expected {self.get_expected_messages(proc)} msgs, got {msgs_received} msgs"
assert self.valid_power_draw(proc, cur), f"expected {expected:.2f}W, got {cur:.2f}W"
print(tabulate(tab))
print(f"Baseline {baseline:.2f}W\n")
if __name__ == "__main__":
unittest.main()
+2 -2
View File
@@ -1,3 +1,3 @@
Import('env', 'cereal', 'messaging', 'common')
Import('env', 'messaging', 'common')
env.Program('logcatd', 'logcatd_systemd.cc', LIBS=[cereal, messaging, common, 'zmq', 'capnp', 'kj', 'systemd', 'json11'])
env.Program('logcatd', 'logcatd_systemd.cc', LIBS=[messaging, common, 'systemd', 'json11'])
+4 -5
View File
@@ -1,9 +1,8 @@
Import('env', 'arch', 'cereal', 'messaging', 'common', 'visionipc')
Import('env', 'arch', 'messaging', 'common', 'visionipc')
libs = [common, cereal, messaging, visionipc,
'zmq', 'capnp', 'kj', 'z',
'avformat', 'avcodec', 'swscale', 'avutil',
'yuv', 'OpenCL', 'pthread']
libs = [common, messaging, visionipc,
'z', 'avformat', 'avcodec', 'swscale',
'avutil', 'yuv', 'OpenCL', 'pthread']
src = ['logger.cc', 'video_writer.cc', 'encoder/encoder.cc', 'encoder/v4l_encoder.cc']
if arch != "larch64":
+18 -2
View File
@@ -6,7 +6,12 @@ VideoEncoder::VideoEncoder(const EncoderInfo &encoder_info, int in_width, int in
out_width = encoder_info.frame_width > 0 ? encoder_info.frame_width : in_width;
out_height = encoder_info.frame_height > 0 ? encoder_info.frame_height : in_height;
pm.reset(new PubMaster({encoder_info.publish_name}));
std::vector pubs = {encoder_info.publish_name};
if (encoder_info.thumbnail_name != NULL) {
pubs.push_back(encoder_info.thumbnail_name);
}
pm.reset(new PubMaster(pubs));
}
void VideoEncoder::publisher_publish(VideoEncoder *e, int segment_num, uint32_t idx, VisionIpcBufExtra &extra,
@@ -40,4 +45,15 @@ void VideoEncoder::publisher_publish(VideoEncoder *e, int segment_num, uint32_t
kj::ArrayOutputStream output_stream(kj::ArrayPtr<capnp::byte>(e->msg_cache.data(), bytes_size));
capnp::writeMessage(output_stream, msg);
e->pm->send(e->encoder_info.publish_name, e->msg_cache.data(), bytes_size);
}
// Publish keyframe thumbnail
if ((flags & V4L2_BUF_FLAG_KEYFRAME) && e->encoder_info.thumbnail_name != NULL) {
MessageBuilder tm;
auto thumbnail = tm.initEvent().initThumbnail();
thumbnail.setFrameId(extra.frame_id);
thumbnail.setTimestampEof(extra.timestamp_eof);
thumbnail.setThumbnail(dat);
thumbnail.setEncoding(cereal::Thumbnail::Encoding::KEYFRAME);
pm->send(e->encoder_info.thumbnail_name, tm);
}
}
+3 -1
View File
@@ -7,7 +7,7 @@
#include <vector>
#include "cereal/messaging/messaging.h"
#include "cereal/visionipc/visionipc.h"
#include "msgq/visionipc/visionipc.h"
#include "common/queue.h"
#include "system/camerad/cameras/camera_common.h"
#include "system/loggerd/loggerd.h"
@@ -25,6 +25,8 @@ public:
void publisher_publish(VideoEncoder *e, int segment_num, uint32_t idx, VisionIpcBufExtra &extra, unsigned int flags, kj::ArrayPtr<capnp::byte> header, kj::ArrayPtr<capnp::byte> dat);
protected:
void publish_thumbnail(uint32_t frame_id, uint64_t timestamp_eof, kj::ArrayPtr<capnp::byte> dat);
int in_width, in_height;
int out_width, out_height;
const EncoderInfo encoder_info;
+1 -1
View File
@@ -215,7 +215,7 @@ void loggerd_thread() {
const bool encoder = util::ends_with(it.name, "EncodeData");
const bool livestream_encoder = util::starts_with(it.name, "livestream");
if (!it.should_log && (!encoder || livestream_encoder)) continue;
LOGD("logging %s (on port %d)", it.name.c_str(), it.port);
LOGD("logging %s", it.name.c_str());
SubSocket * sock = SubSocket::create(ctx.get(), it.name);
assert(sock != NULL);
+3 -1
View File
@@ -4,7 +4,7 @@
#include "cereal/messaging/messaging.h"
#include "cereal/services.h"
#include "cereal/visionipc/visionipc_client.h"
#include "msgq/visionipc/visionipc_client.h"
#include "system/camerad/cameras/camera_common.h"
#include "system/hardware/hw.h"
#include "common/params.h"
@@ -33,6 +33,7 @@ constexpr char PRESERVE_ATTR_VALUE = '1';
class EncoderInfo {
public:
const char *publish_name;
const char *thumbnail_name = NULL;
const char *filename = NULL;
bool record = true;
int frame_width = -1;
@@ -76,6 +77,7 @@ const EncoderInfo main_driver_encoder_info = {
const EncoderInfo stream_road_encoder_info = {
.publish_name = "livestreamRoadEncodeData",
//.thumbnail_name = "thumbnail",
.encode_type = cereal::EncodeIndex::Type::QCAMERA_H264,
.record = false,
.bitrate = LIVESTREAM_BITRATE,
+5 -6
View File
@@ -1,6 +1,5 @@
import os
import random
import unittest
from pathlib import Path
@@ -29,12 +28,12 @@ def create_random_file(file_path: Path, size_mb: float, lock: bool = False, uplo
if upload_xattr is not None:
setxattr(str(file_path), uploader.UPLOAD_ATTR_NAME, upload_xattr)
class MockResponse():
class MockResponse:
def __init__(self, text, status_code):
self.text = text
self.status_code = status_code
class MockApi():
class MockApi:
def __init__(self, dongle_id):
pass
@@ -44,7 +43,7 @@ class MockApi():
def get_token(self):
return "fake-token"
class MockApiIgnore():
class MockApiIgnore:
def __init__(self, dongle_id):
pass
@@ -54,7 +53,7 @@ class MockApiIgnore():
def get_token(self):
return "fake-token"
class UploaderTestCase(unittest.TestCase):
class UploaderTestCase:
f_type = "UNKNOWN"
root: Path
@@ -66,7 +65,7 @@ class UploaderTestCase(unittest.TestCase):
def set_ignore(self):
uploader.Api = MockApiIgnore
def setUp(self):
def setup_method(self):
uploader.Api = MockApi
uploader.fake_upload = True
uploader.force_wifi = True
+5 -11
View File
@@ -1,7 +1,5 @@
#!/usr/bin/env python3
import time
import threading
import unittest
from collections import namedtuple
from pathlib import Path
from collections.abc import Sequence
@@ -17,9 +15,9 @@ class TestDeleter(UploaderTestCase):
def fake_statvfs(self, d):
return self.fake_stats
def setUp(self):
def setup_method(self):
self.f_type = "fcamera.hevc"
super().setUp()
super().setup_method()
self.fake_stats = Stats(f_bavail=0, f_blocks=10, f_frsize=4096)
deleter.os.statvfs = self.fake_statvfs
@@ -64,7 +62,7 @@ class TestDeleter(UploaderTestCase):
finally:
self.join_thread()
self.assertEqual(deleted_order, f_paths, "Files not deleted in expected order")
assert deleted_order == f_paths, "Files not deleted in expected order"
def test_delete_order(self):
self.assertDeleteOrder([
@@ -105,7 +103,7 @@ class TestDeleter(UploaderTestCase):
time.sleep(0.01)
self.join_thread()
self.assertTrue(f_path.exists(), "File deleted with available space")
assert f_path.exists(), "File deleted with available space"
def test_no_delete_with_lock_file(self):
f_path = self.make_file_with_data(self.seg_dir, self.f_type, lock=True)
@@ -116,8 +114,4 @@ class TestDeleter(UploaderTestCase):
time.sleep(0.01)
self.join_thread()
self.assertTrue(f_path.exists(), "File deleted when locked")
if __name__ == "__main__":
unittest.main()
assert f_path.exists(), "File deleted when locked"
+19 -25
View File
@@ -1,4 +1,3 @@
#!/usr/bin/env python3
import math
import os
import pytest
@@ -6,7 +5,6 @@ import random
import shutil
import subprocess
import time
import unittest
from pathlib import Path
from parameterized import parameterized
@@ -15,7 +13,7 @@ from tqdm import trange
from openpilot.common.params import Params
from openpilot.common.timeout import Timeout
from openpilot.system.hardware import TICI
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
from openpilot.tools.lib.logreader import LogReader
from openpilot.system.hardware.hw import Paths
@@ -29,18 +27,18 @@ CAMERAS = [
]
# we check frame count, so we don't have to be too strict on size
FILE_SIZE_TOLERANCE = 0.5
FILE_SIZE_TOLERANCE = 0.7
@pytest.mark.tici # TODO: all of loggerd should work on PC
class TestEncoder(unittest.TestCase):
class TestEncoder:
def setUp(self):
def setup_method(self):
self._clear_logs()
os.environ["LOGGERD_TEST"] = "1"
os.environ["LOGGERD_SEGMENT_LENGTH"] = str(SEGMENT_LENGTH)
def tearDown(self):
def teardown_method(self):
self._clear_logs()
def _clear_logs(self):
@@ -85,7 +83,7 @@ class TestEncoder(unittest.TestCase):
file_path = f"{route_prefix_path}--{i}/{camera}"
# check file exists
self.assertTrue(os.path.exists(file_path), f"segment #{i}: '{file_path}' missing")
assert os.path.exists(file_path), f"segment #{i}: '{file_path}' missing"
# TODO: this ffprobe call is really slow
# check frame count
@@ -98,13 +96,13 @@ class TestEncoder(unittest.TestCase):
frame_count = int(probe.split('\n')[0].strip())
counts.append(frame_count)
self.assertEqual(frame_count, expected_frames,
f"segment #{i}: {camera} failed frame count check: expected {expected_frames}, got {frame_count}")
assert frame_count == expected_frames, \
f"segment #{i}: {camera} failed frame count check: expected {expected_frames}, got {frame_count}"
# sanity check file size
file_size = os.path.getsize(file_path)
self.assertTrue(math.isclose(file_size, size, rel_tol=FILE_SIZE_TOLERANCE),
f"{file_path} size {file_size} isn't close to target size {size}")
assert math.isclose(file_size, size, rel_tol=FILE_SIZE_TOLERANCE), \
f"{file_path} size {file_size} isn't close to target size {size}"
# Check encodeIdx
if encode_idx_name is not None:
@@ -118,24 +116,24 @@ class TestEncoder(unittest.TestCase):
frame_idxs = [m.frameId for m in encode_msgs]
# Check frame count
self.assertEqual(frame_count, len(segment_idxs))
self.assertEqual(frame_count, len(encode_idxs))
assert frame_count == len(segment_idxs)
assert frame_count == len(encode_idxs)
# Check for duplicates or skips
self.assertEqual(0, segment_idxs[0])
self.assertEqual(len(set(segment_idxs)), len(segment_idxs))
assert 0 == segment_idxs[0]
assert len(set(segment_idxs)) == len(segment_idxs)
self.assertTrue(all(valid))
assert all(valid)
self.assertEqual(expected_frames * i, encode_idxs[0])
assert expected_frames * i == encode_idxs[0]
first_frames.append(frame_idxs[0])
self.assertEqual(len(set(encode_idxs)), len(encode_idxs))
assert len(set(encode_idxs)) == len(encode_idxs)
self.assertEqual(1, len(set(first_frames)))
assert 1 == len(set(first_frames))
if TICI:
expected_frames = fps * SEGMENT_LENGTH
self.assertEqual(min(counts), expected_frames)
assert min(counts) == expected_frames
shutil.rmtree(f"{route_prefix_path}--{i}")
try:
@@ -150,7 +148,3 @@ class TestEncoder(unittest.TestCase):
managed_processes['encoderd'].stop()
managed_processes['camerad'].stop()
managed_processes['sensord'].stop()
if __name__ == "__main__":
unittest.main()
+2 -3
View File
@@ -1,4 +1,3 @@
#!/usr/bin/env python3
import numpy as np
import os
import re
@@ -19,11 +18,11 @@ from openpilot.common.timeout import Timeout
from openpilot.system.hardware.hw import Paths
from openpilot.system.loggerd.xattr_cache import getxattr
from openpilot.system.loggerd.deleter import PRESERVE_ATTR_NAME, PRESERVE_ATTR_VALUE
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
from openpilot.system.version import get_version
from openpilot.tools.lib.helpers import RE
from openpilot.tools.lib.logreader import LogReader
from cereal.visionipc import VisionIpcServer, VisionStreamType
from msgq.visionipc import VisionIpcServer, VisionStreamType
from openpilot.common.transformations.camera import DEVICE_CAMERAS
SentinelType = log.Sentinel.SentinelType
+25 -31
View File
@@ -1,8 +1,6 @@
#!/usr/bin/env python3
import os
import time
import threading
import unittest
import logging
import json
from pathlib import Path
@@ -38,8 +36,8 @@ cloudlog.addHandler(log_handler)
class TestUploader(UploaderTestCase):
def setUp(self):
super().setUp()
def setup_method(self):
super().setup_method()
log_handler.reset()
def start_thread(self):
@@ -80,13 +78,13 @@ class TestUploader(UploaderTestCase):
exp_order = self.gen_order([self.seg_num], [])
self.assertTrue(len(log_handler.upload_ignored) == 0, "Some files were ignored")
self.assertFalse(len(log_handler.upload_order) < len(exp_order), "Some files failed to upload")
self.assertFalse(len(log_handler.upload_order) > len(exp_order), "Some files were uploaded twice")
assert len(log_handler.upload_ignored) == 0, "Some files were ignored"
assert not len(log_handler.upload_order) < len(exp_order), "Some files failed to upload"
assert not len(log_handler.upload_order) > len(exp_order), "Some files were uploaded twice"
for f_path in exp_order:
self.assertEqual(os.getxattr((Path(Paths.log_root()) / f_path).with_suffix(""), UPLOAD_ATTR_NAME), UPLOAD_ATTR_VALUE, "All files not uploaded")
assert os.getxattr((Path(Paths.log_root()) / f_path).with_suffix(""), UPLOAD_ATTR_NAME) == UPLOAD_ATTR_VALUE, "All files not uploaded"
self.assertTrue(log_handler.upload_order == exp_order, "Files uploaded in wrong order")
assert log_handler.upload_order == exp_order, "Files uploaded in wrong order"
def test_upload_with_wrong_xattr(self):
self.gen_files(lock=False, xattr=b'0')
@@ -98,13 +96,13 @@ class TestUploader(UploaderTestCase):
exp_order = self.gen_order([self.seg_num], [])
self.assertTrue(len(log_handler.upload_ignored) == 0, "Some files were ignored")
self.assertFalse(len(log_handler.upload_order) < len(exp_order), "Some files failed to upload")
self.assertFalse(len(log_handler.upload_order) > len(exp_order), "Some files were uploaded twice")
assert len(log_handler.upload_ignored) == 0, "Some files were ignored"
assert not len(log_handler.upload_order) < len(exp_order), "Some files failed to upload"
assert not len(log_handler.upload_order) > len(exp_order), "Some files were uploaded twice"
for f_path in exp_order:
self.assertEqual(os.getxattr((Path(Paths.log_root()) / f_path).with_suffix(""), UPLOAD_ATTR_NAME), UPLOAD_ATTR_VALUE, "All files not uploaded")
assert os.getxattr((Path(Paths.log_root()) / f_path).with_suffix(""), UPLOAD_ATTR_NAME) == UPLOAD_ATTR_VALUE, "All files not uploaded"
self.assertTrue(log_handler.upload_order == exp_order, "Files uploaded in wrong order")
assert log_handler.upload_order == exp_order, "Files uploaded in wrong order"
def test_upload_ignored(self):
self.set_ignore()
@@ -117,13 +115,13 @@ class TestUploader(UploaderTestCase):
exp_order = self.gen_order([self.seg_num], [])
self.assertTrue(len(log_handler.upload_order) == 0, "Some files were not ignored")
self.assertFalse(len(log_handler.upload_ignored) < len(exp_order), "Some files failed to ignore")
self.assertFalse(len(log_handler.upload_ignored) > len(exp_order), "Some files were ignored twice")
assert len(log_handler.upload_order) == 0, "Some files were not ignored"
assert not len(log_handler.upload_ignored) < len(exp_order), "Some files failed to ignore"
assert not len(log_handler.upload_ignored) > len(exp_order), "Some files were ignored twice"
for f_path in exp_order:
self.assertEqual(os.getxattr((Path(Paths.log_root()) / f_path).with_suffix(""), UPLOAD_ATTR_NAME), UPLOAD_ATTR_VALUE, "All files not ignored")
assert os.getxattr((Path(Paths.log_root()) / f_path).with_suffix(""), UPLOAD_ATTR_NAME) == UPLOAD_ATTR_VALUE, "All files not ignored"
self.assertTrue(log_handler.upload_ignored == exp_order, "Files ignored in wrong order")
assert log_handler.upload_ignored == exp_order, "Files ignored in wrong order"
def test_upload_files_in_create_order(self):
seg1_nums = [0, 1, 2, 10, 20]
@@ -142,13 +140,13 @@ class TestUploader(UploaderTestCase):
time.sleep(5)
self.join_thread()
self.assertTrue(len(log_handler.upload_ignored) == 0, "Some files were ignored")
self.assertFalse(len(log_handler.upload_order) < len(exp_order), "Some files failed to upload")
self.assertFalse(len(log_handler.upload_order) > len(exp_order), "Some files were uploaded twice")
assert len(log_handler.upload_ignored) == 0, "Some files were ignored"
assert not len(log_handler.upload_order) < len(exp_order), "Some files failed to upload"
assert not len(log_handler.upload_order) > len(exp_order), "Some files were uploaded twice"
for f_path in exp_order:
self.assertEqual(os.getxattr((Path(Paths.log_root()) / f_path).with_suffix(""), UPLOAD_ATTR_NAME), UPLOAD_ATTR_VALUE, "All files not uploaded")
assert os.getxattr((Path(Paths.log_root()) / f_path).with_suffix(""), UPLOAD_ATTR_NAME) == UPLOAD_ATTR_VALUE, "All files not uploaded"
self.assertTrue(log_handler.upload_order == exp_order, "Files uploaded in wrong order")
assert log_handler.upload_order == exp_order, "Files uploaded in wrong order"
def test_no_upload_with_lock_file(self):
self.start_thread()
@@ -163,7 +161,7 @@ class TestUploader(UploaderTestCase):
for f_path in f_paths:
fn = f_path.with_suffix(f_path.suffix.replace(".bz2", ""))
uploaded = UPLOAD_ATTR_NAME in os.listxattr(fn) and os.getxattr(fn, UPLOAD_ATTR_NAME) == UPLOAD_ATTR_VALUE
self.assertFalse(uploaded, "File upload when locked")
assert not uploaded, "File upload when locked"
def test_no_upload_with_xattr(self):
self.gen_files(lock=False, xattr=UPLOAD_ATTR_VALUE)
@@ -173,7 +171,7 @@ class TestUploader(UploaderTestCase):
time.sleep(5)
self.join_thread()
self.assertEqual(len(log_handler.upload_order), 0, "File uploaded again")
assert len(log_handler.upload_order) == 0, "File uploaded again"
def test_clear_locks_on_startup(self):
f_paths = self.gen_files(lock=True, boot=False)
@@ -183,8 +181,4 @@ class TestUploader(UploaderTestCase):
for f_path in f_paths:
lock_path = f_path.with_suffix(f_path.suffix + ".lock")
self.assertFalse(lock_path.is_file(), "File lock not cleared on startup")
if __name__ == "__main__":
unittest.main()
assert not lock_path.is_file(), "File lock not cleared on startup"
View File
+90
View File
@@ -0,0 +1,90 @@
#!/usr/bin/env python3
import os
import subprocess
from pathlib import Path
# NOTE: Do NOT import anything here that needs be built (e.g. params)
from openpilot.common.basedir import BASEDIR
from openpilot.common.spinner import Spinner
from openpilot.common.text_window import TextWindow
from openpilot.system.hardware import AGNOS
from openpilot.common.swaglog import cloudlog, add_file_handler
from openpilot.system.version import get_build_metadata
MAX_CACHE_SIZE = 4e9 if "CI" in os.environ else 2e9
CACHE_DIR = Path("/data/scons_cache" if AGNOS else "/tmp/scons_cache")
TOTAL_SCONS_NODES = 2820
MAX_BUILD_PROGRESS = 100
def build(spinner: Spinner, dirty: bool = False, minimal: bool = False) -> None:
env = os.environ.copy()
env['SCONS_PROGRESS'] = "1"
nproc = os.cpu_count()
if nproc is None:
nproc = 2
extra_args = ["--minimal"] if minimal else []
# building with all cores can result in using too
# much memory, so retry with less parallelism
compile_output: list[bytes] = []
for n in (nproc, nproc/2, 1):
compile_output.clear()
scons: subprocess.Popen = subprocess.Popen(["scons", f"-j{int(n)}", "--cache-populate", *extra_args], cwd=BASEDIR, env=env, stderr=subprocess.PIPE)
assert scons.stderr is not None
# Read progress from stderr and update spinner
while scons.poll() is None:
try:
line = scons.stderr.readline()
if line is None:
continue
line = line.rstrip()
prefix = b'progress: '
if line.startswith(prefix):
i = int(line[len(prefix):])
spinner.update_progress(MAX_BUILD_PROGRESS * min(1., i / TOTAL_SCONS_NODES), 100.)
elif len(line):
compile_output.append(line)
print(line.decode('utf8', 'replace'))
except Exception:
pass
if scons.returncode == 0:
break
if scons.returncode != 0:
# Read remaining output
if scons.stderr is not None:
compile_output += scons.stderr.read().split(b'\n')
# Build failed log errors
error_s = b"\n".join(compile_output).decode('utf8', 'replace')
add_file_handler(cloudlog)
cloudlog.error("scons build failed\n" + error_s)
# Show TextWindow
spinner.close()
if not os.getenv("CI"):
with TextWindow("openpilot failed to build\n \n" + error_s) as t:
t.wait_for_exit()
exit(1)
# enforce max cache size
cache_files = [f for f in CACHE_DIR.rglob('*') if f.is_file()]
cache_files.sort(key=lambda f: f.stat().st_mtime)
cache_size = sum(f.stat().st_size for f in cache_files)
for f in cache_files:
if cache_size < MAX_CACHE_SIZE:
break
cache_size -= f.stat().st_size
f.unlink()
if __name__ == "__main__":
spinner = Spinner()
spinner.update_progress(0, 100)
build_metadata = get_build_metadata()
build(spinner, build_metadata.openpilot.is_dirty, minimal = AGNOS)
+67
View File
@@ -0,0 +1,67 @@
import errno
import fcntl
import os
import sys
import pathlib
import shutil
import signal
import subprocess
import tempfile
import threading
from openpilot.common.basedir import BASEDIR
from openpilot.common.params import Params
def unblock_stdout() -> None:
# get a non-blocking stdout
child_pid, child_pty = os.forkpty()
if child_pid != 0: # parent
# child is in its own process group, manually pass kill signals
signal.signal(signal.SIGINT, lambda signum, frame: os.kill(child_pid, signal.SIGINT))
signal.signal(signal.SIGTERM, lambda signum, frame: os.kill(child_pid, signal.SIGTERM))
fcntl.fcntl(sys.stdout, fcntl.F_SETFL, fcntl.fcntl(sys.stdout, fcntl.F_GETFL) | os.O_NONBLOCK)
while True:
try:
dat = os.read(child_pty, 4096)
except OSError as e:
if e.errno == errno.EIO:
break
continue
if not dat:
break
try:
sys.stdout.write(dat.decode('utf8'))
except (OSError, UnicodeDecodeError):
pass
# os.wait() returns a tuple with the pid and a 16 bit value
# whose low byte is the signal number and whose high byte is the exit status
exit_status = os.wait()[1] >> 8
os._exit(exit_status)
def write_onroad_params(started, params):
params.put_bool("IsOnroad", started)
params.put_bool("IsOffroad", not started)
def save_bootlog():
# copy current params
tmp = tempfile.mkdtemp()
params_dirname = pathlib.Path(Params().get_param_path()).name
params_dir = os.path.join(tmp, params_dirname)
shutil.copytree(Params().get_param_path(), params_dir, dirs_exist_ok=True)
def fn(tmpdir):
env = os.environ.copy()
env['PARAMS_COPY_PATH'] = tmpdir
subprocess.call("./bootlog", cwd=os.path.join(BASEDIR, "system/loggerd"), env=env)
shutil.rmtree(tmpdir)
t = threading.Thread(target=fn, args=(tmp, ))
t.daemon = True
t.start()
+239
View File
@@ -0,0 +1,239 @@
#!/usr/bin/env python3
import datetime
import os
import signal
import sys
import traceback
from cereal import log
import cereal.messaging as messaging
import openpilot.system.sentry as sentry
from openpilot.common.params import Params, ParamKeyType
from openpilot.common.text_window import TextWindow
from openpilot.system.hardware import HARDWARE, PC
from openpilot.system.manager.helpers import unblock_stdout, write_onroad_params, save_bootlog
from openpilot.system.manager.process import ensure_running
from openpilot.system.manager.process_config import managed_processes
from openpilot.system.athena.registration import register, UNREGISTERED_DONGLE_ID
from openpilot.common.swaglog import cloudlog, add_file_handler
from openpilot.system.version import get_build_metadata, terms_version, training_version
def manager_init() -> None:
save_bootlog()
build_metadata = get_build_metadata()
params = Params()
params.clear_all(ParamKeyType.CLEAR_ON_MANAGER_START)
params.clear_all(ParamKeyType.CLEAR_ON_ONROAD_TRANSITION)
params.clear_all(ParamKeyType.CLEAR_ON_OFFROAD_TRANSITION)
if build_metadata.release_channel:
params.clear_all(ParamKeyType.DEVELOPMENT_ONLY)
default_params: list[tuple[str, str | bytes]] = [
("CompletedTrainingVersion", "0"),
("DisengageOnAccelerator", "0"),
("GsmMetered", "1"),
("HasAcceptedTerms", "0"),
("LanguageSetting", "main_en"),
("OpenpilotEnabledToggle", "1"),
("LongitudinalPersonality", str(log.LongitudinalPersonality.standard)),
# dp
("dp_device_display_off_mode", "0"),
("dp_ui_rainbow", "0"),
("dp_ui_flight_panel", "0"),
("dp_long_de2e", "0"),
("dp_long_personality_btn", "0"),
("dp_nav_full_screen", "0"),
]
if not PC:
default_params.append(("LastUpdateTime", datetime.datetime.now(datetime.UTC).replace(tzinfo=None).isoformat().encode('utf8')))
if params.get_bool("RecordFrontLock"):
params.put_bool("RecordFront", True)
# set unset params
for k, v in default_params:
if params.get(k) is None:
params.put(k, v)
# Create folders needed for msgq
try:
os.mkdir("/dev/shm")
except FileExistsError:
pass
except PermissionError:
print("WARNING: failed to make /dev/shm")
# set version params
params.put("Version", build_metadata.openpilot.version)
params.put("TermsVersion", terms_version)
params.put("TrainingVersion", training_version)
params.put("GitCommit", build_metadata.openpilot.git_commit)
params.put("GitCommitDate", build_metadata.openpilot.git_commit_date)
params.put("GitBranch", build_metadata.channel)
params.put("GitRemote", build_metadata.openpilot.git_origin)
params.put_bool("IsTestedBranch", build_metadata.tested_channel)
params.put_bool("IsReleaseBranch", build_metadata.release_channel)
# set dongle id
reg_res = register(show_spinner=True)
if reg_res:
dongle_id = reg_res
else:
serial = params.get("HardwareSerial")
raise Exception(f"Registration failed for device {serial}")
os.environ['DONGLE_ID'] = dongle_id # Needed for swaglog
os.environ['GIT_ORIGIN'] = build_metadata.openpilot.git_normalized_origin # Needed for swaglog
os.environ['GIT_BRANCH'] = build_metadata.channel # Needed for swaglog
os.environ['GIT_COMMIT'] = build_metadata.openpilot.git_commit # Needed for swaglog
if not build_metadata.openpilot.is_dirty:
os.environ['CLEAN'] = '1'
# init logging
sentry.init(sentry.SentryProject.SELFDRIVE)
cloudlog.bind_global(dongle_id=dongle_id,
version=build_metadata.openpilot.version,
origin=build_metadata.openpilot.git_normalized_origin,
branch=build_metadata.channel,
commit=build_metadata.openpilot.git_commit,
dirty=build_metadata.openpilot.is_dirty,
device=HARDWARE.get_device_type())
# preimport all processes
for p in managed_processes.values():
p.prepare()
def manager_cleanup() -> None:
# send signals to kill all procs
for p in managed_processes.values():
p.stop(block=False)
# ensure all are killed
for p in managed_processes.values():
p.stop(block=True)
cloudlog.info("everything is dead")
def manager_thread() -> None:
cloudlog.bind(daemon="manager")
cloudlog.info("manager start")
cloudlog.info({"environ": os.environ})
params = Params()
ignore: list[str] = []
if params.get("DongleId", encoding='utf8') in (None, UNREGISTERED_DONGLE_ID):
ignore += ["manage_athenad", "uploader"]
if os.getenv("NOBOARD") is not None:
ignore.append("pandad")
ignore += [x for x in os.getenv("BLOCK", "").split(",") if len(x) > 0]
sm = messaging.SubMaster(['deviceState', 'carParams'], poll='deviceState')
pm = messaging.PubMaster(['managerState'])
write_onroad_params(False, params)
ensure_running(managed_processes.values(), False, params=params, CP=sm['carParams'], not_run=ignore)
started_prev = False
while True:
sm.update(1000)
started = sm['deviceState'].started
if started and not started_prev:
params.clear_all(ParamKeyType.CLEAR_ON_ONROAD_TRANSITION)
elif not started and started_prev:
params.clear_all(ParamKeyType.CLEAR_ON_OFFROAD_TRANSITION)
# update onroad params, which drives pandad's safety setter thread
if started != started_prev:
write_onroad_params(started, params)
started_prev = started
ensure_running(managed_processes.values(), started, params=params, CP=sm['carParams'], not_run=ignore)
running = ' '.join("{}{}\u001b[0m".format("\u001b[32m" if p.proc.is_alive() else "\u001b[31m", p.name)
for p in managed_processes.values() if p.proc)
print(running)
cloudlog.debug(running)
# send managerState
msg = messaging.new_message('managerState', valid=True)
msg.managerState.processes = [p.get_process_state_msg() for p in managed_processes.values()]
pm.send('managerState', msg)
# Exit main loop when uninstall/shutdown/reboot is needed
shutdown = False
for param in ("DoUninstall", "DoShutdown", "DoReboot"):
if params.get_bool(param):
shutdown = True
params.put("LastManagerExitReason", f"{param} {datetime.datetime.now()}")
cloudlog.warning(f"Shutting down manager - {param} set")
if shutdown:
break
def main() -> None:
manager_init()
if os.getenv("PREPAREONLY") is not None:
return
# SystemExit on sigterm
signal.signal(signal.SIGTERM, lambda signum, frame: sys.exit(1))
try:
manager_thread()
except Exception:
traceback.print_exc()
sentry.capture_exception()
finally:
manager_cleanup()
params = Params()
if params.get_bool("DoUninstall"):
cloudlog.warning("uninstalling")
HARDWARE.uninstall()
elif params.get_bool("DoReboot"):
cloudlog.warning("reboot")
HARDWARE.reboot()
elif params.get_bool("DoShutdown"):
cloudlog.warning("shutdown")
HARDWARE.shutdown()
if __name__ == "__main__":
unblock_stdout()
try:
main()
except KeyboardInterrupt:
print("got CTRL-C, exiting")
except Exception:
add_file_handler(cloudlog)
cloudlog.exception("Manager failed to start")
try:
managed_processes['ui'].stop()
except Exception:
pass
# Show last 3 lines of traceback
error = traceback.format_exc(-3)
error = "Manager failed to start\n\n" + error
with TextWindow(error) as t:
t.wait_for_exit()
raise
# manual exit because we are forked
sys.exit(0)
+293
View File
@@ -0,0 +1,293 @@
import importlib
import os
import signal
import struct
import time
import subprocess
from collections.abc import Callable, ValuesView
from abc import ABC, abstractmethod
from multiprocessing import Process
from openpilot.common.threadname import setthreadname
from cereal import car, log
import cereal.messaging as messaging
import openpilot.system.sentry as sentry
from openpilot.common.basedir import BASEDIR
from openpilot.common.params import Params
from openpilot.common.swaglog import cloudlog
WATCHDOG_FN = "/dev/shm/wd_"
ENABLE_WATCHDOG = os.getenv("NO_WATCHDOG") is None
def launcher(proc: str, name: str) -> None:
try:
# import the process
mod = importlib.import_module(proc)
# rename the process
setthreadname(proc)
# create new context since we forked
messaging.context = messaging.Context()
# add daemon name tag to logs
cloudlog.bind(daemon=name)
sentry.set_tag("daemon", name)
# exec the process
mod.main()
except KeyboardInterrupt:
cloudlog.warning(f"child {proc} got SIGINT")
except Exception:
# can't install the crash handler because sys.excepthook doesn't play nice
# with threads, so catch it here.
sentry.capture_exception()
raise
def nativelauncher(pargs: list[str], cwd: str, name: str) -> None:
os.environ['MANAGER_DAEMON'] = name
# exec the process
os.chdir(cwd)
os.execvp(pargs[0], pargs)
def join_process(process: Process, timeout: float) -> None:
# Process().join(timeout) will hang due to a python 3 bug: https://bugs.python.org/issue28382
# We have to poll the exitcode instead
t = time.monotonic()
while time.monotonic() - t < timeout and process.exitcode is None:
time.sleep(0.001)
class ManagerProcess(ABC):
daemon = False
sigkill = False
should_run: Callable[[bool, Params, car.CarParams], bool]
proc: Process | None = None
enabled = True
name = ""
last_watchdog_time = 0
watchdog_max_dt: int | None = None
watchdog_seen = False
shutting_down = False
@abstractmethod
def prepare(self) -> None:
pass
@abstractmethod
def start(self) -> None:
pass
def restart(self) -> None:
self.stop(sig=signal.SIGKILL)
self.start()
def check_watchdog(self, started: bool) -> None:
if self.watchdog_max_dt is None or self.proc is None:
return
try:
fn = WATCHDOG_FN + str(self.proc.pid)
with open(fn, "rb") as f:
# TODO: why can't pylint find struct.unpack?
self.last_watchdog_time = struct.unpack('Q', f.read())[0]
except Exception:
pass
dt = time.monotonic() - self.last_watchdog_time / 1e9
if dt > self.watchdog_max_dt:
if self.watchdog_seen and ENABLE_WATCHDOG:
cloudlog.error(f"Watchdog timeout for {self.name} (exitcode {self.proc.exitcode}) restarting ({started=})")
self.restart()
else:
self.watchdog_seen = True
def stop(self, retry: bool = True, block: bool = True, sig: signal.Signals = None) -> int | None:
if self.proc is None:
return None
if self.proc.exitcode is None:
if not self.shutting_down:
cloudlog.info(f"killing {self.name}")
if sig is None:
sig = signal.SIGKILL if self.sigkill else signal.SIGINT
self.signal(sig)
self.shutting_down = True
if not block:
return None
join_process(self.proc, 5)
# If process failed to die send SIGKILL
if self.proc.exitcode is None and retry:
cloudlog.info(f"killing {self.name} with SIGKILL")
self.signal(signal.SIGKILL)
self.proc.join()
ret = self.proc.exitcode
cloudlog.info(f"{self.name} is dead with {ret}")
if self.proc.exitcode is not None:
self.shutting_down = False
self.proc = None
return ret
def signal(self, sig: int) -> None:
if self.proc is None:
return
# Don't signal if already exited
if self.proc.exitcode is not None and self.proc.pid is not None:
return
# Can't signal if we don't have a pid
if self.proc.pid is None:
return
cloudlog.info(f"sending signal {sig} to {self.name}")
os.kill(self.proc.pid, sig)
def get_process_state_msg(self):
state = log.ManagerState.ProcessState.new_message()
state.name = self.name
if self.proc:
state.running = self.proc.is_alive()
state.shouldBeRunning = self.proc is not None and not self.shutting_down
state.pid = self.proc.pid or 0
state.exitCode = self.proc.exitcode or 0
return state
class NativeProcess(ManagerProcess):
def __init__(self, name, cwd, cmdline, should_run, enabled=True, sigkill=False, watchdog_max_dt=None):
self.name = name
self.cwd = cwd
self.cmdline = cmdline
self.should_run = should_run
self.enabled = enabled
self.sigkill = sigkill
self.watchdog_max_dt = watchdog_max_dt
self.launcher = nativelauncher
def prepare(self) -> None:
pass
def start(self) -> None:
# In case we only tried a non blocking stop we need to stop it before restarting
if self.shutting_down:
self.stop()
if self.proc is not None:
return
cwd = os.path.join(BASEDIR, self.cwd)
cloudlog.info(f"starting process {self.name}")
self.proc = Process(name=self.name, target=self.launcher, args=(self.cmdline, cwd, self.name))
self.proc.start()
self.watchdog_seen = False
self.shutting_down = False
class PythonProcess(ManagerProcess):
def __init__(self, name, module, should_run, enabled=True, sigkill=False, watchdog_max_dt=None):
self.name = name
self.module = module
self.should_run = should_run
self.enabled = enabled
self.sigkill = sigkill
self.watchdog_max_dt = watchdog_max_dt
self.launcher = launcher
def prepare(self) -> None:
if self.enabled:
cloudlog.info(f"preimporting {self.module}")
importlib.import_module(self.module)
def start(self) -> None:
# In case we only tried a non blocking stop we need to stop it before restarting
if self.shutting_down:
self.stop()
if self.proc is not None:
return
cloudlog.info(f"starting python {self.module}")
self.proc = Process(name=self.name, target=self.launcher, args=(self.module, self.name))
self.proc.start()
self.watchdog_seen = False
self.shutting_down = False
class DaemonProcess(ManagerProcess):
"""Python process that has to stay running across manager restart.
This is used for athena so you don't lose SSH access when restarting manager."""
def __init__(self, name, module, param_name, enabled=True):
self.name = name
self.module = module
self.param_name = param_name
self.enabled = enabled
self.params = None
@staticmethod
def should_run(started, params, CP):
return True
def prepare(self) -> None:
pass
def start(self) -> None:
if self.params is None:
self.params = Params()
pid = self.params.get(self.param_name, encoding='utf-8')
if pid is not None:
try:
os.kill(int(pid), 0)
with open(f'/proc/{pid}/cmdline') as f:
if self.module in f.read():
# daemon is running
return
except (OSError, FileNotFoundError):
# process is dead
pass
cloudlog.info(f"starting daemon {self.name}")
proc = subprocess.Popen(['python', '-m', self.module],
stdin=open('/dev/null'),
stdout=open('/dev/null', 'w'),
stderr=open('/dev/null', 'w'),
preexec_fn=os.setpgrp)
self.params.put(self.param_name, str(proc.pid))
def stop(self, retry=True, block=True, sig=None) -> None:
pass
def ensure_running(procs: ValuesView[ManagerProcess], started: bool, params=None, CP: car.CarParams=None,
not_run: list[str] | None=None) -> list[ManagerProcess]:
if not_run is None:
not_run = []
running = []
for p in procs:
if p.enabled and p.name not in not_run and p.should_run(started, params, CP):
running.append(p)
else:
p.stop(block=False)
p.check_watchdog(started)
for p in running:
p.start()
return running
+91
View File
@@ -0,0 +1,91 @@
import os
from cereal import car
from openpilot.common.params import Params
from openpilot.system.hardware import PC, TICI
from openpilot.system.manager.process import PythonProcess, NativeProcess, DaemonProcess
WEBCAM = os.getenv("USE_WEBCAM") is not None
def driverview(started: bool, params: Params, CP: car.CarParams) -> bool:
return started or params.get_bool("IsDriverViewEnabled")
def notcar(started: bool, params: Params, CP: car.CarParams) -> bool:
return started and CP.notCar
def iscar(started: bool, params: Params, CP: car.CarParams) -> bool:
return started and not CP.notCar
def logging(started, params, CP: car.CarParams) -> bool:
run = (not CP.notCar) or not params.get_bool("DisableLogging")
return started and run
def ublox_available() -> bool:
return os.path.exists('/dev/ttyHS0') and not os.path.exists('/persist/comma/use-quectel-gps')
def ublox(started, params, CP: car.CarParams) -> bool:
use_ublox = ublox_available()
if use_ublox != params.get_bool("UbloxAvailable"):
params.put_bool("UbloxAvailable", use_ublox)
return started and use_ublox
def qcomgps(started, params, CP: car.CarParams) -> bool:
return started and not ublox_available()
def always_run(started, params, CP: car.CarParams) -> bool:
return True
def only_onroad(started: bool, params, CP: car.CarParams) -> bool:
return started
def only_offroad(started, params, CP: car.CarParams) -> bool:
return not started
procs = [
DaemonProcess("manage_athenad", "system.athena.manage_athenad", "AthenadPid"),
NativeProcess("camerad", "system/camerad", ["./camerad"], driverview),
NativeProcess("logcatd", "system/logcatd", ["./logcatd"], only_onroad),
NativeProcess("proclogd", "system/proclogd", ["./proclogd"], only_onroad),
PythonProcess("logmessaged", "system.logmessaged", always_run),
PythonProcess("micd", "system.micd", iscar),
PythonProcess("timed", "system.timed", always_run, enabled=not PC),
PythonProcess("dmonitoringmodeld", "selfdrive.modeld.dmonitoringmodeld", driverview, enabled=(not PC or WEBCAM)),
NativeProcess("encoderd", "system/loggerd", ["./encoderd"], only_onroad),
NativeProcess("stream_encoderd", "system/loggerd", ["./encoderd", "--stream"], notcar),
NativeProcess("loggerd", "system/loggerd", ["./loggerd"], logging),
NativeProcess("modeld", "selfdrive/modeld", ["./modeld"], only_onroad),
NativeProcess("sensord", "system/sensord", ["./sensord"], only_onroad, enabled=not PC),
NativeProcess("ui", "selfdrive/ui", ["./ui"], always_run, watchdog_max_dt=(5 if not PC else None)),
PythonProcess("soundd", "selfdrive.ui.soundd", only_onroad),
NativeProcess("locationd", "selfdrive/locationd", ["./locationd"], only_onroad),
NativeProcess("pandad", "selfdrive/pandad", ["./pandad"], always_run, enabled=False),
PythonProcess("calibrationd", "selfdrive.locationd.calibrationd", only_onroad),
PythonProcess("torqued", "selfdrive.locationd.torqued", only_onroad),
PythonProcess("controlsd", "selfdrive.controls.controlsd", only_onroad),
PythonProcess("card", "selfdrive.car.card", only_onroad),
PythonProcess("deleter", "system.loggerd.deleter", always_run),
PythonProcess("dmonitoringd", "selfdrive.monitoring.dmonitoringd", driverview, enabled=(not PC or WEBCAM)),
PythonProcess("qcomgpsd", "system.qcomgpsd.qcomgpsd", qcomgps, enabled=TICI),
#PythonProcess("ugpsd", "system.ugpsd", only_onroad, enabled=TICI),
PythonProcess("navd", "selfdrive.navd.navd", only_onroad),
PythonProcess("pandad", "selfdrive.pandad.pandad", always_run),
PythonProcess("paramsd", "selfdrive.locationd.paramsd", only_onroad),
NativeProcess("ubloxd", "system/ubloxd", ["./ubloxd"], ublox, enabled=TICI),
PythonProcess("pigeond", "system.ubloxd.pigeond", ublox, enabled=TICI),
PythonProcess("plannerd", "selfdrive.controls.plannerd", only_onroad),
PythonProcess("radard", "selfdrive.controls.radard", only_onroad),
PythonProcess("hardwared", "system.hardware.hardwared", always_run),
PythonProcess("tombstoned", "system.tombstoned", always_run, enabled=not PC),
PythonProcess("updated", "system.updated.updated", only_offroad, enabled=not PC),
PythonProcess("uploader", "system.loggerd.uploader", always_run),
PythonProcess("statsd", "system.statsd", always_run),
# debug procs
NativeProcess("bridge", "cereal/messaging", ["./bridge"], notcar),
PythonProcess("webrtcd", "system.webrtc.webrtcd", notcar),
PythonProcess("webjoystick", "tools.bodyteleop.web", notcar),
]
managed_processes = {p.name: p for p in procs}
View File
+76
View File
@@ -0,0 +1,76 @@
import os
import pytest
import signal
import time
from parameterized import parameterized
from cereal import car
from openpilot.common.params import Params
import openpilot.system.manager.manager as manager
from openpilot.system.manager.process import ensure_running
from openpilot.system.manager.process_config import managed_processes
from openpilot.system.hardware import HARDWARE
os.environ['FAKEUPLOAD'] = "1"
MAX_STARTUP_TIME = 3
BLACKLIST_PROCS = ['manage_athenad', 'pandad', 'pigeond']
@pytest.mark.tici
class TestManager:
def setup_method(self):
HARDWARE.set_power_save(False)
# ensure clean CarParams
params = Params()
params.clear_all()
def teardown_method(self):
manager.manager_cleanup()
def test_manager_prepare(self):
os.environ['PREPAREONLY'] = '1'
manager.main()
def test_blacklisted_procs(self):
# TODO: ensure there are blacklisted procs until we have a dedicated test
assert len(BLACKLIST_PROCS), "No blacklisted procs to test not_run"
@parameterized.expand([(i,) for i in range(10)])
def test_startup_time(self, index):
start = time.monotonic()
os.environ['PREPAREONLY'] = '1'
manager.main()
t = time.monotonic() - start
assert t < MAX_STARTUP_TIME, f"startup took {t}s, expected <{MAX_STARTUP_TIME}s"
@pytest.mark.skip("this test is flaky the way it's currently written, should be moved to test_onroad")
def test_clean_exit(self, subtests):
"""
Ensure all processes exit cleanly when stopped.
"""
HARDWARE.set_power_save(False)
manager.manager_init()
CP = car.CarParams.new_message()
procs = ensure_running(managed_processes.values(), True, Params(), CP, not_run=BLACKLIST_PROCS)
time.sleep(10)
for p in procs:
with subtests.test(proc=p.name):
state = p.get_process_state_msg()
assert state.running, f"{p.name} not running"
exit_code = p.stop(retry=False)
assert p.name not in BLACKLIST_PROCS, f"{p.name} was started"
assert exit_code is not None, f"{p.name} failed to exit"
# TODO: interrupted blocking read exits with 1 in cereal. use a more unique return code
exit_codes = [0, 1]
if p.sigkill:
exit_codes = [-signal.SIGKILL]
assert exit_code in exit_codes, f"{p.name} died with {exit_code}"
+2 -2
View File
@@ -1,5 +1,5 @@
Import('env', 'cereal', 'messaging', 'common')
libs = [cereal, messaging, 'pthread', 'zmq', 'capnp', 'kj', 'common', 'zmq', 'json11']
Import('env', 'messaging', 'common')
libs = [messaging, 'pthread', 'common', 'zmq', 'json11']
env.Program('proclogd', ['main.cc', 'proclog.cc'], LIBS=libs)
if GetOption('extras'):
+1 -1
View File
@@ -127,7 +127,7 @@ def main() -> NoReturn:
print("qcomgpsd is running, please kill openpilot before running this script! (aborted)")
sys.exit(1)
except CalledProcessError as e:
if e.returncode != 1: # 1 == no process found (boardd not running)
if e.returncode != 1: # 1 == no process found (pandad not running)
raise e
print("power up antenna ...")
+1 -1
View File
@@ -173,7 +173,7 @@ def setup_quectel(diag: ModemDiag) -> bool:
os.remove(ASSIST_DATA_FILE)
#at_cmd("AT+QGPSXTRADATA?")
if system_time_valid():
time_str = datetime.datetime.utcnow().strftime("%Y/%m/%d,%H:%M:%S")
time_str = datetime.datetime.now(datetime.UTC).replace(tzinfo=None).strftime("%Y/%m/%d,%H:%M:%S")
at_cmd(f"AT+QGPSXTRATIME=0,\"{time_str}\",1,1,1000")
at_cmd("AT+QGPSCFG=\"outport\",\"usbnmea\"")
+11 -16
View File
@@ -1,38 +1,36 @@
#!/usr/bin/env python3
import os
import pytest
import json
import time
import datetime
import unittest
import subprocess
import cereal.messaging as messaging
from openpilot.system.qcomgpsd.qcomgpsd import at_cmd, wait_for_modem
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
GOOD_SIGNAL = bool(int(os.getenv("GOOD_SIGNAL", '0')))
@pytest.mark.tici
class TestRawgpsd(unittest.TestCase):
class TestRawgpsd:
@classmethod
def setUpClass(cls):
def setup_class(cls):
os.system("sudo systemctl start systemd-resolved")
os.system("sudo systemctl restart ModemManager lte")
wait_for_modem()
@classmethod
def tearDownClass(cls):
def teardown_class(cls):
managed_processes['qcomgpsd'].stop()
os.system("sudo systemctl restart systemd-resolved")
os.system("sudo systemctl restart ModemManager lte")
def setUp(self):
def setup_method(self):
at_cmd("AT+QGPSDEL=0")
self.sm = messaging.SubMaster(['qcomGnss', 'gpsLocation', 'gnssMeasurements'])
def tearDown(self):
def teardown_method(self):
managed_processes['qcomgpsd'].stop()
os.system("sudo systemctl restart systemd-resolved")
@@ -57,18 +55,18 @@ class TestRawgpsd(unittest.TestCase):
os.system("sudo systemctl restart ModemManager")
assert self._wait_for_output(30)
def test_startup_time(self):
def test_startup_time(self, subtests):
for internet in (True, False):
if not internet:
os.system("sudo systemctl stop systemd-resolved")
with self.subTest(internet=internet):
with subtests.test(internet=internet):
managed_processes['qcomgpsd'].start()
assert self._wait_for_output(7)
managed_processes['qcomgpsd'].stop()
def test_turns_off_gnss(self):
def test_turns_off_gnss(self, subtests):
for s in (0.1, 1, 5):
with self.subTest(runtime=s):
with subtests.test(runtime=s):
managed_processes['qcomgpsd'].start()
time.sleep(s)
managed_processes['qcomgpsd'].stop()
@@ -87,7 +85,7 @@ class TestRawgpsd(unittest.TestCase):
if should_be_loaded:
assert valid_duration == "10080" # should be max time
injected_time = datetime.datetime.strptime(injected_time_str.replace("\"", ""), "%Y/%m/%d,%H:%M:%S")
self.assertLess(abs((datetime.datetime.utcnow() - injected_time).total_seconds()), 60*60*12)
assert abs((datetime.datetime.now(datetime.UTC).replace(tzinfo=None) - injected_time).total_seconds()) < 60*60*12
else:
valid_duration, injected_time_str = out.split(",", 1)
injected_time_str = injected_time_str.replace('\"', '').replace('\'', '')
@@ -119,6 +117,3 @@ class TestRawgpsd(unittest.TestCase):
time.sleep(15)
managed_processes['qcomgpsd'].stop()
self.check_assistance(True)
if __name__ == "__main__":
unittest.main(failfast=True)
+2 -2
View File
@@ -1,4 +1,4 @@
Import('env', 'arch', 'common', 'cereal', 'messaging')
Import('env', 'arch', 'common', 'messaging')
sensors = [
'sensors/i2c_sensor.cc',
@@ -11,7 +11,7 @@ sensors = [
'sensors/lsm6ds3_temp.cc',
'sensors/mmc5603nj_magn.cc',
]
libs = [common, cereal, messaging, 'capnp', 'zmq', 'kj', 'pthread']
libs = [common, messaging, 'pthread']
if arch == "larch64":
libs.append('i2c')
env.Program('sensord', ['sensors_qcom2.cc'] + sensors, LIBS=libs)
+10 -15
View File
@@ -1,8 +1,6 @@
#!/usr/bin/env python3
import os
import pytest
import time
import unittest
import numpy as np
from collections import namedtuple, defaultdict
@@ -11,7 +9,7 @@ from cereal import log
from cereal.services import SERVICE_LIST
from openpilot.common.gpio import get_irqs_for_action
from openpilot.common.timeout import Timeout
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
BMX = {
('bmx055', 'acceleration'),
@@ -99,9 +97,9 @@ def read_sensor_events(duration_sec):
return {k: v for k, v in events.items() if len(v) > 0}
@pytest.mark.tici
class TestSensord(unittest.TestCase):
class TestSensord:
@classmethod
def setUpClass(cls):
def setup_class(cls):
# enable LSM self test
os.environ["LSM_SELF_TEST"] = "1"
@@ -119,10 +117,10 @@ class TestSensord(unittest.TestCase):
managed_processes["sensord"].stop()
@classmethod
def tearDownClass(cls):
def teardown_class(cls):
managed_processes["sensord"].stop()
def tearDown(self):
def teardown_method(self):
managed_processes["sensord"].stop()
def test_sensors_present(self):
@@ -133,9 +131,9 @@ class TestSensord(unittest.TestCase):
m = getattr(measurement, measurement.which())
seen.add((str(m.source), m.which()))
self.assertIn(seen, SENSOR_CONFIGURATIONS)
assert seen in SENSOR_CONFIGURATIONS
def test_lsm6ds3_timing(self):
def test_lsm6ds3_timing(self, subtests):
# verify measurements are sampled and published at 104Hz
sensor_t = {
@@ -152,7 +150,7 @@ class TestSensord(unittest.TestCase):
sensor_t[m.sensor].append(m.timestamp)
for s, vals in sensor_t.items():
with self.subTest(sensor=s):
with subtests.test(sensor=s):
assert len(vals) > 0
tdiffs = np.diff(vals) / 1e6 # millis
@@ -166,9 +164,9 @@ class TestSensord(unittest.TestCase):
stddev = np.std(tdiffs)
assert stddev < 2.0, f"Standard-dev to big {stddev}"
def test_sensor_frequency(self):
def test_sensor_frequency(self, subtests):
for s, msgs in self.events.items():
with self.subTest(sensor=s):
with subtests.test(sensor=s):
freq = len(msgs) / self.sample_secs
ef = SERVICE_LIST[s].frequency
assert ef*0.85 <= freq <= ef*1.15
@@ -246,6 +244,3 @@ class TestSensord(unittest.TestCase):
state_two = get_irq_count(self.sensord_irq)
assert state_one == state_two, "Interrupts received after sensord stop!"
if __name__ == "__main__":
unittest.main()
+1 -1
View File
@@ -4,7 +4,7 @@ import time
import atexit
from cereal import messaging
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
TIMEOUT = 10*60
+78
View File
@@ -0,0 +1,78 @@
"""Install exception handler for process crash."""
import sentry_sdk
from enum import Enum
from sentry_sdk.integrations.threading import ThreadingIntegration
from openpilot.common.params import Params
from openpilot.system.athena.registration import is_registered_device
from openpilot.system.hardware import HARDWARE, PC
from openpilot.common.swaglog import cloudlog
from openpilot.system.version import get_build_metadata, get_version
class SentryProject(Enum):
# python project
SELFDRIVE = "https://6f3c7076c1e14b2aa10f5dde6dda0cc4@o33823.ingest.sentry.io/77924"
# native project
SELFDRIVE_NATIVE = "https://3e4b586ed21a4479ad5d85083b639bc6@o33823.ingest.sentry.io/157615"
def report_tombstone(fn: str, message: str, contents: str) -> None:
cloudlog.error({'tombstone': message})
with sentry_sdk.configure_scope() as scope:
scope.set_extra("tombstone_fn", fn)
scope.set_extra("tombstone", contents)
sentry_sdk.capture_message(message=message)
sentry_sdk.flush()
def capture_exception(*args, **kwargs) -> None:
cloudlog.error("crash", exc_info=kwargs.get('exc_info', 1))
try:
sentry_sdk.capture_exception(*args, **kwargs)
sentry_sdk.flush() # https://github.com/getsentry/sentry-python/issues/291
except Exception:
cloudlog.exception("sentry exception")
def set_tag(key: str, value: str) -> None:
sentry_sdk.set_tag(key, value)
def init(project: SentryProject) -> bool:
build_metadata = get_build_metadata()
# forks like to mess with this, so double check
comma_remote = build_metadata.openpilot.comma_remote and "commaai" in build_metadata.openpilot.git_origin
if not comma_remote or not is_registered_device() or PC:
return False
env = "release" if build_metadata.tested_channel else "master"
dongle_id = Params().get("DongleId", encoding='utf-8')
integrations = []
if project == SentryProject.SELFDRIVE:
integrations.append(ThreadingIntegration(propagate_hub=True))
sentry_sdk.init(project.value,
default_integrations=False,
release=get_version(),
integrations=integrations,
traces_sample_rate=1.0,
max_value_length=8192,
environment=env)
build_metadata = get_build_metadata()
sentry_sdk.set_user({"id": dongle_id})
sentry_sdk.set_tag("dirty", build_metadata.openpilot.is_dirty)
sentry_sdk.set_tag("origin", build_metadata.openpilot.git_origin)
sentry_sdk.set_tag("branch", build_metadata.channel)
sentry_sdk.set_tag("commit", build_metadata.openpilot.git_commit)
sentry_sdk.set_tag("device", HARDWARE.get_device_type())
if project == SentryProject.SELFDRIVE:
sentry_sdk.Hub.current.start_session()
return True
+181
View File
@@ -0,0 +1,181 @@
#!/usr/bin/env python3
import os
import zmq
import time
from pathlib import Path
from collections import defaultdict
from datetime import datetime, UTC
from typing import NoReturn
from openpilot.common.params import Params
from cereal.messaging import SubMaster
from openpilot.system.hardware.hw import Paths
from openpilot.common.swaglog import cloudlog
from openpilot.system.hardware import HARDWARE
from openpilot.common.file_helpers import atomic_write_in_dir
from openpilot.system.version import get_build_metadata
from openpilot.system.loggerd.config import STATS_DIR_FILE_LIMIT, STATS_SOCKET, STATS_FLUSH_TIME_S
class METRIC_TYPE:
GAUGE = 'g'
SAMPLE = 'sa'
class StatLog:
def __init__(self):
self.pid = None
self.zctx = None
self.sock = None
def connect(self) -> None:
self.zctx = zmq.Context()
self.sock = self.zctx.socket(zmq.PUSH)
self.sock.setsockopt(zmq.LINGER, 10)
self.sock.connect(STATS_SOCKET)
self.pid = os.getpid()
def __del__(self):
if self.sock is not None:
self.sock.close()
if self.zctx is not None:
self.zctx.term()
def _send(self, metric: str) -> None:
if os.getpid() != self.pid:
self.connect()
try:
self.sock.send_string(metric, zmq.NOBLOCK)
except zmq.error.Again:
# drop :/
pass
def gauge(self, name: str, value: float) -> None:
self._send(f"{name}:{value}|{METRIC_TYPE.GAUGE}")
# Samples will be recorded in a buffer and at aggregation time,
# statistical properties will be logged (mean, count, percentiles, ...)
def sample(self, name: str, value: float):
self._send(f"{name}:{value}|{METRIC_TYPE.SAMPLE}")
def main() -> NoReturn:
dongle_id = Params().get("DongleId", encoding='utf-8')
def get_influxdb_line(measurement: str, value: float | dict[str, float], timestamp: datetime, tags: dict) -> str:
res = f"{measurement}"
for k, v in tags.items():
res += f",{k}={str(v)}"
res += " "
if isinstance(value, float):
value = {'value': value}
for k, v in value.items():
res += f"{k}={v},"
res += f"dongle_id=\"{dongle_id}\" {int(timestamp.timestamp() * 1e9)}\n"
return res
# open statistics socket
ctx = zmq.Context.instance()
sock = ctx.socket(zmq.PULL)
sock.bind(STATS_SOCKET)
STATS_DIR = Paths.stats_root()
# initialize stats directory
Path(STATS_DIR).mkdir(parents=True, exist_ok=True)
build_metadata = get_build_metadata()
# initialize tags
tags = {
'started': False,
'version': build_metadata.openpilot.version,
'branch': build_metadata.channel,
'dirty': build_metadata.openpilot.is_dirty,
'origin': build_metadata.openpilot.git_normalized_origin,
'deviceType': HARDWARE.get_device_type(),
}
# subscribe to deviceState for started state
sm = SubMaster(['deviceState'])
idx = 0
last_flush_time = time.monotonic()
gauges = {}
samples: dict[str, list[float]] = defaultdict(list)
try:
while True:
started_prev = sm['deviceState'].started
sm.update()
# Update metrics
while True:
try:
metric = sock.recv_string(zmq.NOBLOCK)
try:
metric_type = metric.split('|')[1]
metric_name = metric.split(':')[0]
metric_value = float(metric.split('|')[0].split(':')[1])
if metric_type == METRIC_TYPE.GAUGE:
gauges[metric_name] = metric_value
elif metric_type == METRIC_TYPE.SAMPLE:
samples[metric_name].append(metric_value)
else:
cloudlog.event("unknown metric type", metric_type=metric_type)
except Exception:
cloudlog.event("malformed metric", metric=metric)
except zmq.error.Again:
break
# flush when started state changes or after FLUSH_TIME_S
if (time.monotonic() > last_flush_time + STATS_FLUSH_TIME_S) or (sm['deviceState'].started != started_prev):
result = ""
current_time = datetime.now(UTC)
tags['started'] = sm['deviceState'].started
for key, value in gauges.items():
result += get_influxdb_line(f"gauge.{key}", value, current_time, tags)
for key, values in samples.items():
values.sort()
sample_count = len(values)
sample_sum = sum(values)
stats = {
'count': sample_count,
'min': values[0],
'max': values[-1],
'mean': sample_sum / sample_count,
}
for percentile in [0.05, 0.5, 0.95]:
value = values[int(round(percentile * (sample_count - 1)))]
stats[f"p{int(percentile * 100)}"] = value
result += get_influxdb_line(f"sample.{key}", stats, current_time, tags)
# clear intermediate data
gauges.clear()
samples.clear()
last_flush_time = time.monotonic()
# check that we aren't filling up the drive
if len(os.listdir(STATS_DIR)) < STATS_DIR_FILE_LIMIT:
if len(result) > 0:
stats_path = os.path.join(STATS_DIR, f"{current_time.timestamp():.0f}_{idx}")
with atomic_write_in_dir(stats_path) as f:
f.write(result)
idx += 1
else:
cloudlog.error("stats dir full")
finally:
sock.close()
ctx.term()
if __name__ == "__main__":
main()
else:
statlog = StatLog()
Executable → Regular
+4 -9
View File
@@ -1,17 +1,15 @@
#!/usr/bin/env python3
import glob
import os
import time
import unittest
import cereal.messaging as messaging
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
from openpilot.system.hardware.hw import Paths
from openpilot.common.swaglog import cloudlog, ipchandler
class TestLogmessaged(unittest.TestCase):
def setUp(self):
class TestLogmessaged:
def setup_method(self):
# clear the IPC buffer in case some other tests used cloudlog and filled it
ipchandler.close()
ipchandler.connect()
@@ -25,7 +23,7 @@ class TestLogmessaged(unittest.TestCase):
messaging.drain_sock(self.sock)
messaging.drain_sock(self.error_sock)
def tearDown(self):
def teardown_method(self):
del self.sock
del self.error_sock
managed_processes['logmessaged'].stop(block=True)
@@ -55,6 +53,3 @@ class TestLogmessaged(unittest.TestCase):
logsize = sum([os.path.getsize(f) for f in self._get_log_files()])
assert (n*len(msg)) < logsize < (n*(len(msg)+1024))
if __name__ == "__main__":
unittest.main()
+176
View File
@@ -0,0 +1,176 @@
#!/usr/bin/env python3
import datetime
import os
import re
import shutil
import signal
import subprocess
import time
import glob
from typing import NoReturn
import openpilot.system.sentry as sentry
from openpilot.system.hardware.hw import Paths
from openpilot.common.swaglog import cloudlog
from openpilot.system.version import get_build_metadata
MAX_SIZE = 1_000_000 * 100 # allow up to 100M
MAX_TOMBSTONE_FN_LEN = 62 # 85 - 23 ("<dongle id>/crash/")
TOMBSTONE_DIR = "/data/tombstones/"
APPORT_DIR = "/var/crash/"
def safe_fn(s):
extra = ['_']
return "".join(c for c in s if c.isalnum() or c in extra).rstrip()
def clear_apport_folder():
for f in glob.glob(APPORT_DIR + '*'):
try:
os.remove(f)
except Exception:
pass
def get_apport_stacktrace(fn):
try:
cmd = f'apport-retrace -s <(cat <(echo "Package: openpilot") "{fn}")'
return subprocess.check_output(cmd, shell=True, encoding='utf8', timeout=30, executable='/bin/bash')
except subprocess.CalledProcessError:
return "Error getting stacktrace"
except subprocess.TimeoutExpired:
return "Timeout getting stacktrace"
def get_tombstones():
"""Returns list of (filename, ctime) for all crashlogs"""
files = []
if os.path.exists(APPORT_DIR):
with os.scandir(APPORT_DIR) as d:
# Loop over first 1000 directory entries
for _, f in zip(range(1000), d, strict=False):
if f.name.startswith("tombstone"):
files.append((f.path, int(f.stat().st_ctime)))
elif f.name.endswith(".crash") and f.stat().st_mode == 0o100640:
files.append((f.path, int(f.stat().st_ctime)))
return files
def report_tombstone_apport(fn):
f_size = os.path.getsize(fn)
if f_size > MAX_SIZE:
cloudlog.error(f"Tombstone {fn} too big, {f_size}. Skipping...")
return
message = "" # One line description of the crash
contents = "" # Full file contents without coredump
path = "" # File path relative to openpilot directory
proc_maps = False
with open(fn) as f:
for line in f:
if "CoreDump" in line:
break
elif "ProcMaps" in line:
proc_maps = True
elif "ProcStatus" in line:
proc_maps = False
if not proc_maps:
contents += line
if "ExecutablePath" in line:
path = line.strip().split(': ')[-1]
path = path.replace('/data/openpilot/', '')
message += path
elif "Signal" in line:
message += " - " + line.strip()
try:
sig_num = int(line.strip().split(': ')[-1])
message += " (" + signal.Signals(sig_num).name + ")"
except ValueError:
pass
stacktrace = get_apport_stacktrace(fn)
stacktrace_s = stacktrace.split('\n')
crash_function = "No stacktrace"
if len(stacktrace_s) > 2:
found = False
# Try to find first entry in openpilot, fall back to first line
for line in stacktrace_s:
if "at selfdrive/" in line:
crash_function = line
found = True
break
if not found:
crash_function = stacktrace_s[1]
# Remove arguments that can contain pointers to make sentry one-liner unique
crash_function = " ".join(x for x in crash_function.split(' ')[1:] if not x.startswith('0x'))
crash_function = re.sub(r'\(.*?\)', '', crash_function)
contents = stacktrace + "\n\n" + contents
message = message + " - " + crash_function
sentry.report_tombstone(fn, message, contents)
# Copy crashlog to upload folder
clean_path = path.replace('/', '_')
date = datetime.datetime.now().strftime("%Y-%m-%d--%H-%M-%S")
build_metadata = get_build_metadata()
new_fn = f"{date}_{(build_metadata.openpilot.git_commit or 'nocommit')[:8]}_{safe_fn(clean_path)}"[:MAX_TOMBSTONE_FN_LEN]
crashlog_dir = os.path.join(Paths.log_root(), "crash")
os.makedirs(crashlog_dir, exist_ok=True)
# Files could be on different filesystems, copy, then delete
shutil.copy(fn, os.path.join(crashlog_dir, new_fn))
try:
os.remove(fn)
except PermissionError:
pass
def main() -> NoReturn:
should_report = sentry.init(sentry.SentryProject.SELFDRIVE_NATIVE)
# Clear apport folder on start, otherwise duplicate crashes won't register
clear_apport_folder()
initial_tombstones = set(get_tombstones())
while True:
now_tombstones = set(get_tombstones())
for fn, _ in (now_tombstones - initial_tombstones):
# clear logs if we're not interested in them
if not should_report:
try:
os.remove(fn)
except Exception:
pass
continue
try:
cloudlog.info(f"reporting new tombstone {fn}")
if fn.endswith(".crash"):
report_tombstone_apport(fn)
else:
cloudlog.error(f"unknown crash type: {fn}")
except Exception:
cloudlog.exception(f"Error reporting tombstone {fn}")
initial_tombstones = now_tombstones
time.sleep(5)
if __name__ == "__main__":
main()
+2 -2
View File
@@ -1,6 +1,6 @@
Import('env', 'common', 'cereal', 'messaging')
Import('env', 'common', 'messaging')
loc_libs = [cereal, messaging, 'zmq', common, 'capnp', 'kj', 'kaitai', 'pthread']
loc_libs = [messaging, common, 'kaitai', 'pthread']
if GetOption('kaitai'):
generated = Dir('generated').srcnode().abspath
+3 -3
View File
@@ -6,7 +6,7 @@ import serial
import struct
import requests
import urllib.parse
from datetime import datetime
from datetime import datetime, UTC
from cereal import messaging
from openpilot.common.params import Params
@@ -61,7 +61,7 @@ def get_assistnow_messages(token: bytes) -> list[bytes]:
return msgs
class TTYPigeon():
class TTYPigeon:
def __init__(self):
self.tty = serial.VTIMESerial(UBLOX_TTY, baudrate=9600, timeout=0)
@@ -196,7 +196,7 @@ def initialize_pigeon(pigeon: TTYPigeon) -> bool:
cloudlog.error(f"failed to restore almanac backup, status: {restore_status}")
# sending time to ublox
t_now = datetime.utcnow()
t_now = datetime.now(UTC).replace(tzinfo=None)
if t_now >= datetime(2021, 6, 1):
cloudlog.warning("Sending current time to ublox")
+3 -9
View File
@@ -1,21 +1,19 @@
#!/usr/bin/env python3
import pytest
import time
import unittest
import cereal.messaging as messaging
from cereal.services import SERVICE_LIST
from openpilot.common.gpio import gpio_read
from openpilot.selfdrive.test.helpers import with_processes
from openpilot.selfdrive.manager.process_config import managed_processes
from openpilot.system.manager.process_config import managed_processes
from openpilot.system.hardware.tici.pins import GPIO
# TODO: test TTFF when we have good A-GNSS
@pytest.mark.tici
class TestPigeond(unittest.TestCase):
class TestPigeond:
def tearDown(self):
def teardown_method(self):
managed_processes['pigeond'].stop()
@with_processes(['pigeond'])
@@ -54,7 +52,3 @@ class TestPigeond(unittest.TestCase):
assert gpio_read(GPIO.UBLOX_RST_N) == 0
assert gpio_read(GPIO.GNSS_PWR_EN) == 0
if __name__ == "__main__":
unittest.main()
+2 -1
View File
@@ -1,6 +1,7 @@
import pathlib
import tarfile
from typing import IO, Callable
from typing import IO
from collections.abc import Callable
def include_default(_) -> bool:
+43 -46
View File
@@ -1,7 +1,6 @@
#!/usr/bin/env python3
import pytest
import os
import pathlib
import unittest
import tempfile
import subprocess
@@ -14,9 +13,10 @@ from openpilot.system.updated.casync import tar
LOOPBACK = os.environ.get('LOOPBACK', None)
class TestCasync(unittest.TestCase):
@pytest.mark.skip("not used yet")
class TestCasync:
@classmethod
def setUpClass(cls):
def setup_class(cls):
cls.tmpdir = tempfile.TemporaryDirectory()
# Build example contents
@@ -43,7 +43,7 @@ class TestCasync(unittest.TestCase):
# Ensure we have chunk reuse
assert len(hashes) > len(set(hashes))
def setUp(self):
def setup_method(self):
# Clear target_lo
if LOOPBACK is not None:
self.target_lo = LOOPBACK
@@ -53,7 +53,7 @@ class TestCasync(unittest.TestCase):
self.target_fn = os.path.join(self.tmpdir.name, next(tempfile._get_candidate_names()))
self.seed_fn = os.path.join(self.tmpdir.name, next(tempfile._get_candidate_names()))
def tearDown(self):
def teardown_method(self):
for fn in [self.target_fn, self.seed_fn]:
try:
os.unlink(fn)
@@ -67,9 +67,9 @@ class TestCasync(unittest.TestCase):
stats = casync.extract(target, sources, self.target_fn)
with open(self.target_fn, 'rb') as target_f:
self.assertEqual(target_f.read(), self.contents)
assert target_f.read() == self.contents
self.assertEqual(stats['remote'], len(self.contents))
assert stats['remote'] == len(self.contents)
def test_seed(self):
target = casync.parse_caibx(self.manifest_fn)
@@ -83,10 +83,10 @@ class TestCasync(unittest.TestCase):
stats = casync.extract(target, sources, self.target_fn)
with open(self.target_fn, 'rb') as target_f:
self.assertEqual(target_f.read(), self.contents)
assert target_f.read() == self.contents
self.assertGreater(stats['seed'], 0)
self.assertLess(stats['remote'], len(self.contents))
assert stats['seed'] > 0
assert stats['remote'] < len(self.contents)
def test_already_done(self):
"""Test that an already flashed target doesn't download any chunks"""
@@ -101,9 +101,9 @@ class TestCasync(unittest.TestCase):
stats = casync.extract(target, sources, self.target_fn)
with open(self.target_fn, 'rb') as f:
self.assertEqual(f.read(), self.contents)
assert f.read() == self.contents
self.assertEqual(stats['target'], len(self.contents))
assert stats['target'] == len(self.contents)
def test_chunk_reuse(self):
"""Test that chunks that are reused are only downloaded once"""
@@ -119,11 +119,11 @@ class TestCasync(unittest.TestCase):
stats = casync.extract(target, sources, self.target_fn)
with open(self.target_fn, 'rb') as f:
self.assertEqual(f.read(), self.contents)
assert f.read() == self.contents
self.assertLess(stats['remote'], len(self.contents))
assert stats['remote'] < len(self.contents)
@unittest.skipUnless(LOOPBACK, "requires loopback device")
@pytest.mark.skipif(not LOOPBACK, reason="requires loopback device")
def test_lo_simple_extract(self):
target = casync.parse_caibx(self.manifest_fn)
sources = [('remote', casync.RemoteChunkReader(self.store_fn), casync.build_chunk_dict(target))]
@@ -131,11 +131,11 @@ class TestCasync(unittest.TestCase):
stats = casync.extract(target, sources, self.target_lo)
with open(self.target_lo, 'rb') as target_f:
self.assertEqual(target_f.read(len(self.contents)), self.contents)
assert target_f.read(len(self.contents)) == self.contents
self.assertEqual(stats['remote'], len(self.contents))
assert stats['remote'] == len(self.contents)
@unittest.skipUnless(LOOPBACK, "requires loopback device")
@pytest.mark.skipif(not LOOPBACK, reason="requires loopback device")
def test_lo_chunk_reuse(self):
"""Test that chunks that are reused are only downloaded once"""
target = casync.parse_caibx(self.manifest_fn)
@@ -146,12 +146,13 @@ class TestCasync(unittest.TestCase):
stats = casync.extract(target, sources, self.target_lo)
with open(self.target_lo, 'rb') as f:
self.assertEqual(f.read(len(self.contents)), self.contents)
assert f.read(len(self.contents)) == self.contents
self.assertLess(stats['remote'], len(self.contents))
assert stats['remote'] < len(self.contents)
class TestCasyncDirectory(unittest.TestCase):
@pytest.mark.skip("not used yet")
class TestCasyncDirectory:
"""Tests extracting a directory stored as a casync tar archive"""
NUM_FILES = 16
@@ -174,7 +175,7 @@ class TestCasyncDirectory(unittest.TestCase):
os.symlink(f"file_{i}.txt", os.path.join(directory, f"link_{i}.txt"))
@classmethod
def setUpClass(cls):
def setup_class(cls):
cls.tmpdir = tempfile.TemporaryDirectory()
# Create casync files
@@ -190,16 +191,16 @@ class TestCasyncDirectory(unittest.TestCase):
subprocess.check_output(["casync", "make", "--compression=xz", "--store", cls.store_fn, cls.manifest_fn, cls.orig_fn])
@classmethod
def tearDownClass(cls):
def teardown_class(cls):
cls.tmpdir.cleanup()
cls.directory_to_extract.cleanup()
def setUp(self):
def setup_method(self):
self.cache_dir = tempfile.TemporaryDirectory()
self.working_dir = tempfile.TemporaryDirectory()
self.out_dir = tempfile.TemporaryDirectory()
def tearDown(self):
def teardown_method(self):
self.cache_dir.cleanup()
self.working_dir.cleanup()
self.out_dir.cleanup()
@@ -216,32 +217,32 @@ class TestCasyncDirectory(unittest.TestCase):
stats = casync.extract_directory(target, sources, pathlib.Path(self.out_dir.name), tmp_filename)
with open(os.path.join(self.out_dir.name, "file_0.txt"), "rb") as f:
self.assertEqual(f.read(), self.contents)
assert f.read() == self.contents
with open(os.path.join(self.out_dir.name, "link_0.txt"), "rb") as f:
self.assertEqual(f.read(), self.contents)
self.assertEqual(os.readlink(os.path.join(self.out_dir.name, "link_0.txt")), "file_0.txt")
assert f.read() == self.contents
assert os.readlink(os.path.join(self.out_dir.name, "link_0.txt")) == "file_0.txt"
return stats
def test_no_cache(self):
self.setup_cache(self.cache_dir.name, [])
stats = self.run_test()
self.assertGreater(stats['remote'], 0)
self.assertEqual(stats['cache'], 0)
assert stats['remote'] > 0
assert stats['cache'] == 0
def test_full_cache(self):
self.setup_cache(self.cache_dir.name, range(self.NUM_FILES))
stats = self.run_test()
self.assertEqual(stats['remote'], 0)
self.assertGreater(stats['cache'], 0)
assert stats['remote'] == 0
assert stats['cache'] > 0
def test_one_file_cache(self):
self.setup_cache(self.cache_dir.name, range(1))
stats = self.run_test()
self.assertGreater(stats['remote'], 0)
self.assertGreater(stats['cache'], 0)
self.assertLess(stats['cache'], stats['remote'])
assert stats['remote'] > 0
assert stats['cache'] > 0
assert stats['cache'] < stats['remote']
def test_one_file_incorrect_cache(self):
self.setup_cache(self.cache_dir.name, range(self.NUM_FILES))
@@ -249,19 +250,15 @@ class TestCasyncDirectory(unittest.TestCase):
f.write(b"1234")
stats = self.run_test()
self.assertGreater(stats['remote'], 0)
self.assertGreater(stats['cache'], 0)
self.assertGreater(stats['cache'], stats['remote'])
assert stats['remote'] > 0
assert stats['cache'] > 0
assert stats['cache'] > stats['remote']
def test_one_file_missing_cache(self):
self.setup_cache(self.cache_dir.name, range(self.NUM_FILES))
os.unlink(os.path.join(self.cache_dir.name, "file_12.txt"))
stats = self.run_test()
self.assertGreater(stats['remote'], 0)
self.assertGreater(stats['cache'], 0)
self.assertGreater(stats['cache'], stats['remote'])
if __name__ == "__main__":
unittest.main()
assert stats['remote'] > 0
assert stats['cache'] > 0
assert stats['cache'] > stats['remote']
+259
View File
@@ -0,0 +1,259 @@
import os
import pathlib
import shutil
import signal
import stat
import subprocess
import tempfile
import time
import pytest
from openpilot.common.params import Params
from openpilot.system.manager.process import ManagerProcess
from openpilot.selfdrive.test.helpers import processes_context
def get_consistent_flag(path: str) -> bool:
consistent_file = pathlib.Path(os.path.join(path, ".overlay_consistent"))
return consistent_file.is_file()
def run(args, **kwargs):
return subprocess.check_output(args, **kwargs)
def update_release(directory, name, version, agnos_version, release_notes):
with open(directory / "RELEASES.md", "w") as f:
f.write(release_notes)
(directory / "common").mkdir(exist_ok=True)
with open(directory / "common" / "version.h", "w") as f:
f.write(f'#define COMMA_VERSION "{version}"')
launch_env = directory / "launch_env.sh"
with open(launch_env, "w") as f:
f.write(f'export AGNOS_VERSION="{agnos_version}"')
st = os.stat(launch_env)
os.chmod(launch_env, st.st_mode | stat.S_IEXEC)
test_symlink = directory / "test_symlink"
if not os.path.exists(str(test_symlink)):
os.symlink("common/version.h", test_symlink)
def get_version(path: str) -> str:
with open(os.path.join(path, "common", "version.h")) as f:
return f.read().split('"')[1]
@pytest.mark.slow # TODO: can we test overlayfs in GHA?
class TestBaseUpdate:
@classmethod
def setup_class(cls):
if "Base" in cls.__name__:
pytest.skip()
def setup_method(self):
self.tmpdir = tempfile.mkdtemp()
run(["sudo", "mount", "-t", "tmpfs", "tmpfs", self.tmpdir]) # overlayfs doesn't work inside of docker unless this is a tmpfs
self.mock_update_path = pathlib.Path(self.tmpdir)
self.params = Params()
self.basedir = self.mock_update_path / "openpilot"
self.basedir.mkdir()
self.staging_root = self.mock_update_path / "safe_staging"
self.staging_root.mkdir()
self.remote_dir = self.mock_update_path / "remote"
self.remote_dir.mkdir()
os.environ["UPDATER_STAGING_ROOT"] = str(self.staging_root)
os.environ["UPDATER_LOCK_FILE"] = str(self.mock_update_path / "safe_staging_overlay.lock")
self.MOCK_RELEASES = {
"release3": ("0.1.2", "1.2", "0.1.2 release notes"),
"master": ("0.1.3", "1.2", "0.1.3 release notes"),
}
@pytest.fixture(autouse=True)
def mock_basedir(self, mocker):
mocker.patch("openpilot.common.basedir.BASEDIR", self.basedir)
def set_target_branch(self, branch):
self.params.put("UpdaterTargetBranch", branch)
def setup_basedir_release(self, release):
self.params = Params()
self.set_target_branch(release)
def update_remote_release(self, release):
raise NotImplementedError("")
def setup_remote_release(self, release):
raise NotImplementedError("")
def additional_context(self):
raise NotImplementedError("")
def teardown_method(self):
try:
run(["sudo", "umount", "-l", str(self.staging_root / "merged")])
run(["sudo", "umount", "-l", self.tmpdir])
shutil.rmtree(self.tmpdir)
except Exception:
print("cleanup failed...")
def wait_for_condition(self, condition, timeout=12):
start = time.monotonic()
while True:
waited = time.monotonic() - start
if condition():
print(f"waited {waited}s for condition ")
return waited
if waited > timeout:
raise TimeoutError("timed out waiting for condition")
time.sleep(1)
def _test_finalized_update(self, branch, version, agnos_version, release_notes):
assert get_version(str(self.staging_root / "finalized")) == version
assert get_consistent_flag(str(self.staging_root / "finalized"))
assert os.access(str(self.staging_root / "finalized" / "launch_env.sh"), os.X_OK)
with open(self.staging_root / "finalized" / "test_symlink") as f:
assert version in f.read()
class ParamsBaseUpdateTest(TestBaseUpdate):
def _test_finalized_update(self, branch, version, agnos_version, release_notes):
assert self.params.get("UpdaterNewDescription", encoding="utf-8").startswith(f"{version} / {branch}")
assert self.params.get("UpdaterNewReleaseNotes", encoding="utf-8") == f"<p>{release_notes}</p>\n"
super()._test_finalized_update(branch, version, agnos_version, release_notes)
def send_check_for_updates_signal(self, updated: ManagerProcess):
updated.signal(signal.SIGUSR1.value)
def send_download_signal(self, updated: ManagerProcess):
updated.signal(signal.SIGHUP.value)
def _test_params(self, branch, fetch_available, update_available):
assert self.params.get("UpdaterTargetBranch", encoding="utf-8") == branch
assert self.params.get_bool("UpdaterFetchAvailable") == fetch_available
assert self.params.get_bool("UpdateAvailable") == update_available
def wait_for_idle(self):
self.wait_for_condition(lambda: self.params.get("UpdaterState", encoding="utf-8") == "idle")
def wait_for_failed(self):
self.wait_for_condition(lambda: self.params.get("UpdateFailedCount", encoding="utf-8") is not None and \
int(self.params.get("UpdateFailedCount", encoding="utf-8")) > 0)
def wait_for_fetch_available(self):
self.wait_for_condition(lambda: self.params.get_bool("UpdaterFetchAvailable"))
def wait_for_update_available(self):
self.wait_for_condition(lambda: self.params.get_bool("UpdateAvailable"))
def test_no_update(self):
# Start on release3, ensure we don't fetch any updates
self.setup_remote_release("release3")
self.setup_basedir_release("release3")
with self.additional_context(), processes_context(["updated"]) as [updated]:
self._test_params("release3", False, False)
self.wait_for_idle()
self._test_params("release3", False, False)
self.send_check_for_updates_signal(updated)
self.wait_for_idle()
self._test_params("release3", False, False)
def test_new_release(self):
# Start on release3, simulate a release3 commit, ensure we fetch that update properly
self.setup_remote_release("release3")
self.setup_basedir_release("release3")
with self.additional_context(), processes_context(["updated"]) as [updated]:
self._test_params("release3", False, False)
self.wait_for_idle()
self._test_params("release3", False, False)
self.MOCK_RELEASES["release3"] = ("0.1.3", "1.2", "0.1.3 release notes")
self.update_remote_release("release3")
self.send_check_for_updates_signal(updated)
self.wait_for_fetch_available()
self._test_params("release3", True, False)
self.send_download_signal(updated)
self.wait_for_update_available()
self._test_params("release3", False, True)
self._test_finalized_update("release3", *self.MOCK_RELEASES["release3"])
def test_switch_branches(self):
# Start on release3, request to switch to master manually, ensure we switched
self.setup_remote_release("release3")
self.setup_remote_release("master")
self.setup_basedir_release("release3")
with self.additional_context(), processes_context(["updated"]) as [updated]:
self._test_params("release3", False, False)
self.wait_for_idle()
self._test_params("release3", False, False)
self.set_target_branch("master")
self.send_check_for_updates_signal(updated)
self.wait_for_fetch_available()
self._test_params("master", True, False)
self.send_download_signal(updated)
self.wait_for_update_available()
self._test_params("master", False, True)
self._test_finalized_update("master", *self.MOCK_RELEASES["master"])
def test_agnos_update(self, mocker):
# Start on release3, push an update with an agnos change
self.setup_remote_release("release3")
self.setup_basedir_release("release3")
with self.additional_context(), processes_context(["updated"]) as [updated]:
mocker.patch("openpilot.system.hardware.AGNOS", "True")
mocker.patch("openpilot.system.hardware.tici.hardware.Tici.get_os_version", "1.2")
mocker.patch("openpilot.system.hardware.tici.agnos.get_target_slot_number")
mocker.patch("openpilot.system.hardware.tici.agnos.flash_agnos_update")
self._test_params("release3", False, False)
self.wait_for_idle()
self._test_params("release3", False, False)
self.MOCK_RELEASES["release3"] = ("0.1.3", "1.3", "0.1.3 release notes")
self.update_remote_release("release3")
self.send_check_for_updates_signal(updated)
self.wait_for_fetch_available()
self._test_params("release3", True, False)
self.send_download_signal(updated)
self.wait_for_update_available()
self._test_params("release3", False, True)
self._test_finalized_update("release3", *self.MOCK_RELEASES["release3"])
+22
View File
@@ -0,0 +1,22 @@
import contextlib
from openpilot.system.updated.tests.test_base import ParamsBaseUpdateTest, run, update_release
class TestUpdateDGitStrategy(ParamsBaseUpdateTest):
def update_remote_release(self, release):
update_release(self.remote_dir, release, *self.MOCK_RELEASES[release])
run(["git", "add", "."], cwd=self.remote_dir)
run(["git", "commit", "-m", f"openpilot release {release}"], cwd=self.remote_dir)
def setup_remote_release(self, release):
run(["git", "init"], cwd=self.remote_dir)
run(["git", "checkout", "-b", release], cwd=self.remote_dir)
self.update_remote_release(release)
def setup_basedir_release(self, release):
super().setup_basedir_release(release)
run(["git", "clone", "-b", release, self.remote_dir, self.basedir])
@contextlib.contextmanager
def additional_context(self):
yield
+509
View File
@@ -0,0 +1,509 @@
#!/usr/bin/env python3
import os
import re
import datetime
import subprocess
import psutil
import shutil
import signal
import fcntl
import time
import threading
from collections import defaultdict
from pathlib import Path
from openpilot.common.basedir import BASEDIR
from openpilot.common.params import Params
from openpilot.common.time import system_time_valid
from openpilot.common.markdown import parse_markdown
from openpilot.common.swaglog import cloudlog
from openpilot.selfdrive.controls.lib.alertmanager import set_offroad_alert
from openpilot.system.hardware import AGNOS, HARDWARE
from openpilot.system.version import get_build_metadata
LOCK_FILE = os.getenv("UPDATER_LOCK_FILE", "/tmp/safe_staging_overlay.lock")
STAGING_ROOT = os.getenv("UPDATER_STAGING_ROOT", "/data/safe_staging")
OVERLAY_UPPER = os.path.join(STAGING_ROOT, "upper")
OVERLAY_METADATA = os.path.join(STAGING_ROOT, "metadata")
OVERLAY_MERGED = os.path.join(STAGING_ROOT, "merged")
FINALIZED = os.path.join(STAGING_ROOT, "finalized")
OVERLAY_INIT = Path(os.path.join(BASEDIR, ".overlay_init"))
DAYS_NO_CONNECTIVITY_MAX = 14 # do not allow to engage after this many days
DAYS_NO_CONNECTIVITY_PROMPT = 10 # send an offroad prompt after this many days
class UserRequest:
NONE = 0
CHECK = 1
FETCH = 2
class WaitTimeHelper:
def __init__(self):
self.ready_event = threading.Event()
self.user_request = UserRequest.NONE
signal.signal(signal.SIGHUP, self.update_now)
signal.signal(signal.SIGUSR1, self.check_now)
def update_now(self, signum: int, frame) -> None:
cloudlog.info("caught SIGHUP, attempting to downloading update")
self.user_request = UserRequest.FETCH
self.ready_event.set()
def check_now(self, signum: int, frame) -> None:
cloudlog.info("caught SIGUSR1, checking for updates")
self.user_request = UserRequest.CHECK
self.ready_event.set()
def sleep(self, t: float) -> None:
self.ready_event.wait(timeout=t)
def write_time_to_param(params, param) -> None:
t = datetime.datetime.now(datetime.UTC).replace(tzinfo=None)
params.put(param, t.isoformat().encode('utf8'))
def read_time_from_param(params, param) -> datetime.datetime | None:
t = params.get(param, encoding='utf8')
try:
return datetime.datetime.fromisoformat(t)
except (TypeError, ValueError):
pass
return None
def run(cmd: list[str], cwd: str = None) -> str:
return subprocess.check_output(cmd, cwd=cwd, stderr=subprocess.STDOUT, encoding='utf8')
def set_consistent_flag(consistent: bool) -> None:
os.sync()
consistent_file = Path(os.path.join(FINALIZED, ".overlay_consistent"))
if consistent:
consistent_file.touch()
elif not consistent:
consistent_file.unlink(missing_ok=True)
os.sync()
def parse_release_notes(basedir: str) -> bytes:
try:
with open(os.path.join(basedir, "RELEASES.md"), "rb") as f:
r = f.read().split(b'\n\n', 1)[0] # Slice latest release notes
try:
return bytes(parse_markdown(r.decode("utf-8")), encoding="utf-8")
except Exception:
return r + b"\n"
except FileNotFoundError:
pass
except Exception:
cloudlog.exception("failed to parse release notes")
return b""
def setup_git_options(cwd: str) -> None:
# We sync FS object atimes (which NEOS doesn't use) and mtimes, but ctimes
# are outside user control. Make sure Git is set up to ignore system ctimes,
# because they change when we make hard links during finalize. Otherwise,
# there is a lot of unnecessary churn. This appears to be a common need on
# OSX as well: https://www.git-tower.com/blog/make-git-rebase-safe-on-osx/
# We are using copytree to copy the directory, which also changes
# inode numbers. Ignore those changes too.
# Set protocol to the new version (default after git 2.26) to reduce data
# usage on git fetch --dry-run from about 400KB to 18KB.
git_cfg = [
("core.trustctime", "false"),
("core.checkStat", "minimal"),
("protocol.version", "2"),
("gc.auto", "0"),
("gc.autoDetach", "false"),
]
for option, value in git_cfg:
run(["git", "config", option, value], cwd)
def dismount_overlay() -> None:
if os.path.ismount(OVERLAY_MERGED):
cloudlog.info("unmounting existing overlay")
run(["sudo", "umount", "-l", OVERLAY_MERGED])
def init_overlay() -> None:
# Re-create the overlay if BASEDIR/.git has changed since we created the overlay
if OVERLAY_INIT.is_file() and os.path.ismount(OVERLAY_MERGED):
git_dir_path = os.path.join(BASEDIR, ".git")
new_files = run(["find", git_dir_path, "-newer", str(OVERLAY_INIT)])
if not len(new_files.splitlines()):
# A valid overlay already exists
return
else:
cloudlog.info(".git directory changed, recreating overlay")
cloudlog.info("preparing new safe staging area")
params = Params()
params.put_bool("UpdateAvailable", False)
set_consistent_flag(False)
dismount_overlay()
run(["sudo", "rm", "-rf", STAGING_ROOT])
if os.path.isdir(STAGING_ROOT):
shutil.rmtree(STAGING_ROOT)
for dirname in [STAGING_ROOT, OVERLAY_UPPER, OVERLAY_METADATA, OVERLAY_MERGED]:
os.mkdir(dirname, 0o755)
if os.lstat(BASEDIR).st_dev != os.lstat(OVERLAY_MERGED).st_dev:
raise RuntimeError("base and overlay merge directories are on different filesystems; not valid for overlay FS!")
# Leave a timestamped canary in BASEDIR to check at startup. The device clock
# should be correct by the time we get here. If the init file disappears, or
# critical mtimes in BASEDIR are newer than .overlay_init, continue.sh can
# assume that BASEDIR has used for local development or otherwise modified,
# and skips the update activation attempt.
consistent_file = Path(os.path.join(BASEDIR, ".overlay_consistent"))
if consistent_file.is_file():
consistent_file.unlink()
OVERLAY_INIT.touch()
os.sync()
overlay_opts = f"lowerdir={BASEDIR},upperdir={OVERLAY_UPPER},workdir={OVERLAY_METADATA}"
mount_cmd = ["mount", "-t", "overlay", "-o", overlay_opts, "none", OVERLAY_MERGED]
run(["sudo"] + mount_cmd)
run(["sudo", "chmod", "755", os.path.join(OVERLAY_METADATA, "work")])
git_diff = run(["git", "diff"], OVERLAY_MERGED)
params.put("GitDiff", git_diff)
cloudlog.info(f"git diff output:\n{git_diff}")
def finalize_update() -> None:
"""Take the current OverlayFS merged view and finalize a copy outside of
OverlayFS, ready to be swapped-in at BASEDIR. Copy using shutil.copytree"""
# Remove the update ready flag and any old updates
cloudlog.info("creating finalized version of the overlay")
set_consistent_flag(False)
# Copy the merged overlay view and set the update ready flag
if os.path.exists(FINALIZED):
shutil.rmtree(FINALIZED)
shutil.copytree(OVERLAY_MERGED, FINALIZED, symlinks=True)
run(["git", "reset", "--hard"], FINALIZED)
run(["git", "submodule", "foreach", "--recursive", "git", "reset", "--hard"], FINALIZED)
cloudlog.info("Starting git cleanup in finalized update")
t = time.monotonic()
try:
run(["git", "gc"], FINALIZED)
run(["git", "lfs", "prune"], FINALIZED)
cloudlog.event("Done git cleanup", duration=time.monotonic() - t)
except subprocess.CalledProcessError:
cloudlog.exception(f"Failed git cleanup, took {time.monotonic() - t:.3f} s")
set_consistent_flag(True)
cloudlog.info("done finalizing overlay")
def handle_agnos_update() -> None:
from openpilot.system.hardware.tici.agnos import flash_agnos_update, get_target_slot_number
cur_version = HARDWARE.get_os_version()
updated_version = run(["bash", "-c", r"unset AGNOS_VERSION && source launch_env.sh && \
echo -n $AGNOS_VERSION"], OVERLAY_MERGED).strip()
cloudlog.info(f"AGNOS version check: {cur_version} vs {updated_version}")
if cur_version == updated_version:
return
# prevent an openpilot getting swapped in with a mismatched or partially downloaded agnos
set_consistent_flag(False)
cloudlog.info(f"Beginning background installation for AGNOS {updated_version}")
set_offroad_alert("Offroad_NeosUpdate", True)
manifest_path = os.path.join(OVERLAY_MERGED, "system/hardware/tici/agnos.json")
target_slot_number = get_target_slot_number()
flash_agnos_update(manifest_path, target_slot_number, cloudlog)
set_offroad_alert("Offroad_NeosUpdate", False)
class Updater:
def __init__(self):
self.params = Params()
self.branches = defaultdict(str)
self._has_internet: bool = False
@property
def has_internet(self) -> bool:
return self._has_internet
@property
def target_branch(self) -> str:
b: str | None = self.params.get("UpdaterTargetBranch", encoding='utf-8')
if b is None:
b = self.get_branch(BASEDIR)
return b
@property
def update_ready(self) -> bool:
consistent_file = Path(os.path.join(FINALIZED, ".overlay_consistent"))
if consistent_file.is_file():
hash_mismatch = self.get_commit_hash(BASEDIR) != self.branches[self.target_branch]
branch_mismatch = self.get_branch(BASEDIR) != self.target_branch
on_target_branch = self.get_branch(FINALIZED) == self.target_branch
return ((hash_mismatch or branch_mismatch) and on_target_branch)
return False
@property
def update_available(self) -> bool:
if os.path.isdir(OVERLAY_MERGED) and len(self.branches) > 0:
hash_mismatch = self.get_commit_hash(OVERLAY_MERGED) != self.branches[self.target_branch]
branch_mismatch = self.get_branch(OVERLAY_MERGED) != self.target_branch
return hash_mismatch or branch_mismatch
return False
def get_branch(self, path: str) -> str:
return run(["git", "rev-parse", "--abbrev-ref", "HEAD"], path).rstrip()
def get_commit_hash(self, path: str = OVERLAY_MERGED) -> str:
return run(["git", "rev-parse", "HEAD"], path).rstrip()
def set_params(self, update_success: bool, failed_count: int, exception: str | None) -> None:
self.params.put("UpdateFailedCount", str(failed_count))
self.params.put("UpdaterTargetBranch", self.target_branch)
self.params.put_bool("UpdaterFetchAvailable", self.update_available)
if len(self.branches):
self.params.put("UpdaterAvailableBranches", ','.join(self.branches.keys()))
last_update = datetime.datetime.now(datetime.UTC).replace(tzinfo=None)
if update_success:
write_time_to_param(self.params, "LastUpdateTime")
else:
t = read_time_from_param(self.params, "LastUpdateTime")
if t is not None:
last_update = t
if exception is None:
self.params.remove("LastUpdateException")
else:
self.params.put("LastUpdateException", exception)
# Write out current and new version info
def get_description(basedir: str) -> str:
if not os.path.exists(basedir):
return ""
version = ""
branch = ""
commit = ""
commit_date = ""
try:
branch = self.get_branch(basedir)
commit = self.get_commit_hash(basedir)[:7]
with open(os.path.join(basedir, "common", "version.h")) as f:
version = f.read().split('"')[1]
commit_unix_ts = run(["git", "show", "-s", "--format=%ct", "HEAD"], basedir).rstrip()
dt = datetime.datetime.fromtimestamp(int(commit_unix_ts))
commit_date = dt.strftime("%b %d")
except Exception:
cloudlog.exception("updater.get_description")
return f"{version} / {branch} / {commit} / {commit_date}"
self.params.put("UpdaterCurrentDescription", get_description(BASEDIR))
self.params.put("UpdaterCurrentReleaseNotes", parse_release_notes(BASEDIR))
self.params.put("UpdaterNewDescription", get_description(FINALIZED))
self.params.put("UpdaterNewReleaseNotes", parse_release_notes(FINALIZED))
self.params.put_bool("UpdateAvailable", self.update_ready)
# Handle user prompt
for alert in ("Offroad_UpdateFailed", "Offroad_ConnectivityNeeded", "Offroad_ConnectivityNeededPrompt"):
set_offroad_alert(alert, False)
now = datetime.datetime.now(datetime.UTC).replace(tzinfo=None)
dt = now - last_update
build_metadata = get_build_metadata()
if failed_count > 15 and exception is not None and self.has_internet:
if build_metadata.tested_channel:
extra_text = "Ensure the software is correctly installed. Uninstall and re-install if this error persists."
else:
extra_text = exception
set_offroad_alert("Offroad_UpdateFailed", True, extra_text=extra_text)
elif failed_count > 0:
if dt.days > DAYS_NO_CONNECTIVITY_MAX:
set_offroad_alert("Offroad_ConnectivityNeeded", True)
elif dt.days > DAYS_NO_CONNECTIVITY_PROMPT:
remaining = max(DAYS_NO_CONNECTIVITY_MAX - dt.days, 1)
set_offroad_alert("Offroad_ConnectivityNeededPrompt", True, extra_text=f"{remaining} day{'' if remaining == 1 else 's'}.")
def check_for_update(self) -> None:
cloudlog.info("checking for updates")
excluded_branches = ('release2', 'release2-staging')
try:
run(["git", "ls-remote", "origin", "HEAD"], OVERLAY_MERGED)
self._has_internet = True
except subprocess.CalledProcessError:
self._has_internet = False
setup_git_options(OVERLAY_MERGED)
output = run(["git", "ls-remote", "--heads"], OVERLAY_MERGED)
self.branches = defaultdict(lambda: None)
for line in output.split('\n'):
ls_remotes_re = r'(?P<commit_sha>\b[0-9a-f]{5,40}\b)(\s+)(refs\/heads\/)(?P<branch_name>.*$)'
x = re.fullmatch(ls_remotes_re, line.strip())
if x is not None and x.group('branch_name') not in excluded_branches:
self.branches[x.group('branch_name')] = x.group('commit_sha')
cur_branch = self.get_branch(OVERLAY_MERGED)
cur_commit = self.get_commit_hash(OVERLAY_MERGED)
new_branch = self.target_branch
new_commit = self.branches[new_branch]
if (cur_branch, cur_commit) != (new_branch, new_commit):
cloudlog.info(f"update available, {cur_branch} ({str(cur_commit)[:7]}) -> {new_branch} ({str(new_commit)[:7]})")
else:
cloudlog.info(f"up to date on {cur_branch} ({str(cur_commit)[:7]})")
def fetch_update(self) -> None:
cloudlog.info("attempting git fetch inside staging overlay")
self.params.put("UpdaterState", "downloading...")
# TODO: cleanly interrupt this and invalidate old update
set_consistent_flag(False)
self.params.put_bool("UpdateAvailable", False)
setup_git_options(OVERLAY_MERGED)
branch = self.target_branch
git_fetch_output = run(["git", "fetch", "origin", branch], OVERLAY_MERGED)
cloudlog.info("git fetch success: %s", git_fetch_output)
cloudlog.info("git reset in progress")
cmds = [
["git", "checkout", "--force", "--no-recurse-submodules", "-B", branch, "FETCH_HEAD"],
["git", "reset", "--hard"],
["git", "clean", "-xdff"],
["git", "submodule", "sync"],
["git", "submodule", "update", "--init", "--recursive"],
["git", "submodule", "foreach", "--recursive", "git", "reset", "--hard"],
]
r = [run(cmd, OVERLAY_MERGED) for cmd in cmds]
cloudlog.info("git reset success: %s", '\n'.join(r))
# TODO: show agnos download progress
if AGNOS:
handle_agnos_update()
# Create the finalized, ready-to-swap update
self.params.put("UpdaterState", "finalizing update...")
finalize_update()
cloudlog.info("finalize success!")
def main() -> None:
params = Params()
if params.get_bool("DisableUpdates"):
cloudlog.warning("updates are disabled by the DisableUpdates param")
exit(0)
with open(LOCK_FILE, 'w') as ov_lock_fd:
try:
fcntl.flock(ov_lock_fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
except OSError as e:
raise RuntimeError("couldn't get overlay lock; is another instance running?") from e
# Set low io priority
proc = psutil.Process()
if psutil.LINUX:
proc.ionice(psutil.IOPRIO_CLASS_BE, value=7)
# Check if we just performed an update
if Path(os.path.join(STAGING_ROOT, "old_openpilot")).is_dir():
cloudlog.event("update installed")
if not params.get("InstallDate"):
t = datetime.datetime.now(datetime.UTC).replace(tzinfo=None).isoformat()
params.put("InstallDate", t.encode('utf8'))
updater = Updater()
update_failed_count = 0 # TODO: Load from param?
wait_helper = WaitTimeHelper()
# invalidate old finalized update
set_consistent_flag(False)
# set initial state
params.put("UpdaterState", "idle")
# Run the update loop
first_run = True
while True:
wait_helper.ready_event.clear()
# Attempt an update
exception = None
try:
# TODO: reuse overlay from previous updated instance if it looks clean
init_overlay()
# ensure we have some params written soon after startup
updater.set_params(False, update_failed_count, exception)
if not system_time_valid() or first_run:
first_run = False
wait_helper.sleep(60)
continue
update_failed_count += 1
# check for update
params.put("UpdaterState", "checking...")
updater.check_for_update()
# download update
last_fetch = read_time_from_param(params, "UpdaterLastFetchTime")
timed_out = last_fetch is None or (datetime.datetime.now(datetime.UTC).replace(tzinfo=None) - last_fetch > datetime.timedelta(days=3))
user_requested_fetch = wait_helper.user_request == UserRequest.FETCH
if params.get_bool("NetworkMetered") and not timed_out and not user_requested_fetch:
cloudlog.info("skipping fetch, connection metered")
elif wait_helper.user_request == UserRequest.CHECK:
cloudlog.info("skipping fetch, only checking")
else:
updater.fetch_update()
write_time_to_param(params, "UpdaterLastFetchTime")
update_failed_count = 0
except subprocess.CalledProcessError as e:
cloudlog.event(
"update process failed",
cmd=e.cmd,
output=e.output,
returncode=e.returncode
)
exception = f"command failed: {e.cmd}\n{e.output}"
OVERLAY_INIT.unlink(missing_ok=True)
except Exception as e:
cloudlog.exception("uncaught updated exception, shouldn't happen")
exception = str(e)
OVERLAY_INIT.unlink(missing_ok=True)
try:
params.put("UpdaterState", "idle")
update_successful = (update_failed_count == 0)
updater.set_params(update_successful, update_failed_count, exception)
except Exception:
cloudlog.exception("uncaught updated exception while setting params, shouldn't happen")
# infrequent attempts if we successfully updated recently
wait_helper.user_request = UserRequest.NONE
wait_helper.sleep(5*60 if update_failed_count > 0 else 1.5*60*60)
if __name__ == "__main__":
main()
+2 -3
View File
@@ -1,14 +1,13 @@
#!/usr/bin/env python3
from dataclasses import dataclass
from functools import cache
import json
import os
import pathlib
import subprocess
from openpilot.common.basedir import BASEDIR
from openpilot.common.swaglog import cloudlog
from openpilot.common.utils import cache
from openpilot.common.git import get_commit, get_origin, get_branch, get_short_branch, get_commit_date
RELEASE_BRANCHES = ['release3-staging', 'release3', 'nightly']
@@ -27,7 +26,7 @@ def get_version(path: str = BASEDIR) -> str:
def get_release_notes(path: str = BASEDIR) -> str:
with open(os.path.join(path, "RELEASES.md"), "r") as f:
with open(os.path.join(path, "RELEASES.md")) as f:
return f.read().split('\n\n', 1)[0]
+35 -42
View File
@@ -1,17 +1,14 @@
#!/usr/bin/env python3
import asyncio
import unittest
from unittest.mock import Mock, MagicMock, patch
import json
# for aiortc and its dependencies
import warnings
warnings.filterwarnings("ignore", category=DeprecationWarning)
warnings.filterwarnings("ignore", category=RuntimeWarning) # TODO: remove this when google-crc32c publish a python3.12 wheel
from aiortc import RTCDataChannel
from aiortc.mediastreams import VIDEO_CLOCK_RATE, VIDEO_TIME_BASE
import capnp
import pyaudio
from cereal import messaging, log
from openpilot.system.webrtc.webrtcd import CerealOutgoingMessageProxy, CerealIncomingMessageProxy
@@ -20,15 +17,15 @@ from openpilot.system.webrtc.device.audio import AudioInputStreamTrack
from openpilot.common.realtime import DT_DMON
class TestStreamSession(unittest.TestCase):
def setUp(self):
class TestStreamSession:
def setup_method(self):
self.loop = asyncio.new_event_loop()
def tearDown(self):
def teardown_method(self):
self.loop.stop()
self.loop.close()
def test_outgoing_proxy(self):
def test_outgoing_proxy(self, mocker):
test_msg = log.Event.new_message()
test_msg.logMonoTime = 123
test_msg.valid = True
@@ -36,27 +33,27 @@ class TestStreamSession(unittest.TestCase):
expected_dict = {"type": "customReservedRawData0", "logMonoTime": 123, "valid": True, "data": "test"}
expected_json = json.dumps(expected_dict).encode()
channel = Mock(spec=RTCDataChannel)
channel = mocker.Mock(spec=RTCDataChannel)
mocked_submaster = messaging.SubMaster(["customReservedRawData0"])
def mocked_update(t):
mocked_submaster.update_msgs(0, [test_msg])
with patch.object(messaging.SubMaster, "update", side_effect=mocked_update):
proxy = CerealOutgoingMessageProxy(mocked_submaster)
proxy.add_channel(channel)
mocker.patch.object(messaging.SubMaster, "update", side_effect=mocked_update)
proxy = CerealOutgoingMessageProxy(mocked_submaster)
proxy.add_channel(channel)
proxy.update()
proxy.update()
channel.send.assert_called_once_with(expected_json)
channel.send.assert_called_once_with(expected_json)
def test_incoming_proxy(self):
def test_incoming_proxy(self, mocker):
tested_msgs = [
{"type": "customReservedRawData0", "data": "test"}, # primitive
{"type": "can", "data": [{"address": 0, "busTime": 0, "dat": "", "src": 0}]}, # list
{"type": "testJoystick", "data": {"axes": [0, 0], "buttons": [False]}}, # dict
]
mocked_pubmaster = MagicMock(spec=messaging.PubMaster)
mocked_pubmaster = mocker.MagicMock(spec=messaging.PubMaster)
proxy = CerealIncomingMessageProxy(mocked_pubmaster)
@@ -65,44 +62,40 @@ class TestStreamSession(unittest.TestCase):
mocked_pubmaster.send.assert_called_once()
mt, md = mocked_pubmaster.send.call_args.args
self.assertEqual(mt, msg["type"])
self.assertIsInstance(md, capnp._DynamicStructBuilder)
self.assertTrue(hasattr(md, msg["type"]))
assert mt == msg["type"]
assert isinstance(md, capnp._DynamicStructBuilder)
assert hasattr(md, msg["type"])
mocked_pubmaster.reset_mock()
def test_livestream_track(self):
def test_livestream_track(self, mocker):
fake_msg = messaging.new_message("livestreamDriverEncodeData")
config = {"receive.return_value": fake_msg.to_bytes()}
with patch("cereal.messaging.SubSocket", spec=True, **config):
track = LiveStreamVideoStreamTrack("driver")
mocker.patch("msgq.SubSocket", spec=True, **config)
track = LiveStreamVideoStreamTrack("driver")
self.assertTrue(track.id.startswith("driver"))
self.assertEqual(track.codec_preference(), "H264")
assert track.id.startswith("driver")
assert track.codec_preference() == "H264"
for i in range(5):
packet = self.loop.run_until_complete(track.recv())
self.assertEqual(packet.time_base, VIDEO_TIME_BASE)
self.assertEqual(packet.pts, int(i * DT_DMON * VIDEO_CLOCK_RATE))
self.assertEqual(packet.size, 0)
for i in range(5):
packet = self.loop.run_until_complete(track.recv())
assert packet.time_base == VIDEO_TIME_BASE
assert packet.pts == int(i * DT_DMON * VIDEO_CLOCK_RATE)
assert packet.size == 0
def test_input_audio_track(self):
def test_input_audio_track(self, mocker):
packet_time, rate = 0.02, 16000
sample_count = int(packet_time * rate)
mocked_stream = MagicMock(spec=pyaudio.Stream)
mocked_stream = mocker.MagicMock(spec=pyaudio.Stream)
mocked_stream.read.return_value = b"\x00" * 2 * sample_count
config = {"open.side_effect": lambda *args, **kwargs: mocked_stream}
with patch("pyaudio.PyAudio", spec=True, **config):
track = AudioInputStreamTrack(audio_format=pyaudio.paInt16, packet_time=packet_time, rate=rate)
mocker.patch("pyaudio.PyAudio", spec=True, **config)
track = AudioInputStreamTrack(audio_format=pyaudio.paInt16, packet_time=packet_time, rate=rate)
for i in range(5):
frame = self.loop.run_until_complete(track.recv())
self.assertEqual(frame.rate, rate)
self.assertEqual(frame.samples, sample_count)
self.assertEqual(frame.pts, i * sample_count)
if __name__ == "__main__":
unittest.main()
for i in range(5):
frame = self.loop.run_until_complete(track.recv())
assert frame.rate == rate
assert frame.samples == sample_count
assert frame.pts == i * sample_count
+12 -15
View File
@@ -1,11 +1,10 @@
#!/usr/bin/env python
import pytest
import asyncio
import json
import unittest
from unittest.mock import MagicMock, AsyncMock
# for aiortc and its dependencies
import warnings
warnings.filterwarnings("ignore", category=DeprecationWarning)
warnings.filterwarnings("ignore", category=RuntimeWarning) # TODO: remove this when google-crc32c publish a python3.12 wheel
from openpilot.system.webrtc.webrtcd import get_stream
@@ -20,19 +19,20 @@ from parameterized import parameterized_class
(["testJoystick"], []),
([], []),
])
class TestWebrtcdProc(unittest.IsolatedAsyncioTestCase):
@pytest.mark.asyncio
class TestWebrtcdProc:
async def assertCompletesWithTimeout(self, awaitable, timeout=1):
try:
async with asyncio.timeout(timeout):
await awaitable
except TimeoutError:
self.fail("Timeout while waiting for awaitable to complete")
pytest.fail("Timeout while waiting for awaitable to complete")
async def test_webrtcd(self):
mock_request = MagicMock()
async def test_webrtcd(self, mocker):
mock_request = mocker.MagicMock()
async def connect(offer):
body = {'sdp': offer.sdp, 'cameras': offer.video, 'bridge_services_in': self.in_services, 'bridge_services_out': self.out_services}
mock_request.json.side_effect = AsyncMock(return_value=body)
mock_request.json.side_effect = mocker.AsyncMock(return_value=body)
response = await get_stream(mock_request)
response_json = json.loads(response.text)
return aiortc.RTCSessionDescription(**response_json)
@@ -48,9 +48,9 @@ class TestWebrtcdProc(unittest.IsolatedAsyncioTestCase):
await self.assertCompletesWithTimeout(stream.start())
await self.assertCompletesWithTimeout(stream.wait_for_connection())
self.assertTrue(stream.has_incoming_video_track("road"))
self.assertTrue(stream.has_incoming_audio_track())
self.assertEqual(stream.has_messaging_channel(), len(self.in_services) > 0 or len(self.out_services) > 0)
assert stream.has_incoming_video_track("road")
assert stream.has_incoming_audio_track()
assert stream.has_messaging_channel() == (len(self.in_services) > 0 or len(self.out_services) > 0)
video_track, audio_track = stream.get_incoming_video_track("road"), stream.get_incoming_audio_track()
await self.assertCompletesWithTimeout(video_track.recv())
@@ -59,10 +59,7 @@ class TestWebrtcdProc(unittest.IsolatedAsyncioTestCase):
await self.assertCompletesWithTimeout(stream.stop())
# cleanup, very implementation specific, test may break if it changes
self.assertTrue(mock_request.app["streams"].__setitem__.called, "Implementation changed, please update this test")
assert mock_request.app["streams"].__setitem__.called, "Implementation changed, please update this test"
_, session = mock_request.app["streams"].__setitem__.call_args.args
await self.assertCompletesWithTimeout(session.post_run_cleanup())
if __name__ == "__main__":
unittest.main()
+8 -7
View File
@@ -11,6 +11,7 @@ from typing import Any, TYPE_CHECKING
# aiortc and its dependencies have lots of internal warnings :(
import warnings
warnings.filterwarnings("ignore", category=DeprecationWarning)
warnings.filterwarnings("ignore", category=RuntimeWarning) # TODO: remove this when google-crc32c publish a python3.12 wheel
import capnp
from aiohttp import web
@@ -24,7 +25,7 @@ from cereal import messaging, log
class CerealOutgoingMessageProxy:
def __init__(self, sm: messaging.SubMaster):
self.sm = sm
self.channels: list['RTCDataChannel'] = []
self.channels: list[RTCDataChannel] = []
def add_channel(self, channel: 'RTCDataChannel'):
self.channels.append(channel)
@@ -97,8 +98,8 @@ class CerealProxyRunner:
except InvalidStateError:
self.logger.warning("Cereal outgoing proxy invalid state (connection closed)")
break
except Exception as ex:
self.logger.error("Cereal outgoing proxy failure: %s", ex)
except Exception:
self.logger.exception("Cereal outgoing proxy failure")
await asyncio.sleep(0.01)
@@ -175,8 +176,8 @@ class StreamSession:
assert self.incoming_bridge is not None
try:
self.incoming_bridge.send(message)
except Exception as ex:
self.logger.error("Cereal incoming proxy failure: %s", ex)
except Exception:
self.logger.exception("Cereal incoming proxy failure")
async def run(self):
try:
@@ -200,8 +201,8 @@ class StreamSession:
await self.post_run_cleanup()
self.logger.info("Stream session (%s) ended", self.identifier)
except Exception as ex:
self.logger.error("Stream session failure: %s", ex)
except Exception:
self.logger.exception("Stream session failure")
async def post_run_cleanup(self):
await self.stream.stop()