pan131419/openpilot
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Release 260111

Browse Source
This commit is contained in:
Comma Device
2026-01-11 18:23:29 +08:00
commit 3721ecbf8a
2601 changed files with 855070 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

327
opendbc_repo/opendbc/car/tests/test_fw_fingerprint.py Normal file
View File

@@ -0,0 +1,327 @@
import pytest
import random
import time
from collections import defaultdict
from opendbc.car.can_definitions import CanData
from opendbc.car.car_helpers import interfaces
from opendbc.car.structs import CarParams
from opendbc.car.fingerprints import FW_VERSIONS
from opendbc.car.fw_versions import FW_QUERY_CONFIGS, FUZZY_EXCLUDE_ECUS, VERSIONS, build_fw_dict, \
match_fw_to_car, get_brand_ecu_matches, get_fw_versions, get_present_ecus
from opendbc.car.vin import get_vin
CarFw = CarParams.CarFw
Ecu = CarParams.Ecu
ECU_NAME = {v: k for k, v in Ecu.schema.enumerants.items()}
class TestFwFingerprint:
def assertFingerprints(self, candidates, expected):
candidates = list(candidates)
assert len(candidates) == 1, f"got more than one candidate: {candidates}"
assert candidates[0] == expected
@pytest.mark.parametrize("brand, car_model, ecus, test_non_essential",
[(b, c, e[c], n) for b, e in VERSIONS.items() for c in e for n in (True, False)])
def test_exact_match(self, brand, car_model, ecus, test_non_essential):
config = FW_QUERY_CONFIGS[brand]
CP = CarParams()
for _ in range(20):
fw = []
for ecu, fw_versions in ecus.items():
# Assume non-essential ECUs apply to all cars, so we catch cases where Car A with
# missing ECUs won't match to Car B where only Car B has labeled non-essential ECUs
if ecu[0] in config.non_essential_ecus and test_non_essential:
continue
ecu_name, addr, sub_addr = ecu
fw.append(CarFw(ecu=ecu_name, fwVersion=random.choice(fw_versions), brand=brand,
address=addr, subAddress=0 if sub_addr is None else sub_addr))
CP.carFw = fw
_, matches = match_fw_to_car(CP.carFw, CP.carVin, allow_fuzzy=False)
if not test_non_essential:
self.assertFingerprints(matches, car_model)
else:
# if we're removing ECUs we expect some match loss, but it shouldn't mismatch
if len(matches) != 0:
self.assertFingerprints(matches, car_model)
@pytest.mark.parametrize("brand, car_model, ecus", [(b, c, e[c]) for b, e in VERSIONS.items() for c in e])
def test_custom_fuzzy_match(self, brand, car_model, ecus):
# Assert brand-specific fuzzy fingerprinting function doesn't disagree with standard fuzzy function
config = FW_QUERY_CONFIGS[brand]
if config.match_fw_to_car_fuzzy is None:
pytest.skip("Brand does not implement custom fuzzy fingerprinting function")
CP = CarParams()
for _ in range(5):
fw = []
for ecu, fw_versions in ecus.items():
ecu_name, addr, sub_addr = ecu
fw.append(CarFw(ecu=ecu_name, fwVersion=random.choice(fw_versions), brand=brand,
address=addr, subAddress=0 if sub_addr is None else sub_addr))
CP.carFw = fw
_, matches = match_fw_to_car(CP.carFw, CP.carVin, allow_exact=False, log=False)
brand_matches = config.match_fw_to_car_fuzzy(build_fw_dict(CP.carFw), CP.carVin, VERSIONS[brand])
# If both have matches, they must agree
if len(matches) == 1 and len(brand_matches) == 1:
assert matches == brand_matches
@pytest.mark.parametrize("brand, car_model, ecus", [(b, c, e[c]) for b, e in VERSIONS.items() for c in e])
def test_fuzzy_match_ecu_count(self, brand, car_model, ecus):
# Asserts that fuzzy matching does not count matching FW, but ECU address keys
valid_ecus = [e for e in ecus if e[0] not in FUZZY_EXCLUDE_ECUS]
if not len(valid_ecus):
pytest.skip("Car model has no compatible ECUs for fuzzy matching")
fw = []
for ecu in valid_ecus:
ecu_name, addr, sub_addr = ecu
for _ in range(5):
# Add multiple FW versions to simulate ECU returning to multiple queries in a brand
fw.append(CarFw(ecu=ecu_name, fwVersion=random.choice(ecus[ecu]), brand=brand,
address=addr, subAddress=0 if sub_addr is None else sub_addr))
CP = CarParams(carFw=fw)
_, matches = match_fw_to_car(CP.carFw, CP.carVin, allow_exact=False, log=False)
# Assert no match if there are not enough unique ECUs
unique_ecus = {(f.address, f.subAddress) for f in fw}
if len(unique_ecus) < 2:
assert len(matches) == 0, car_model
# There won't always be a match due to shared FW, but if there is it should be correct
elif len(matches):
self.assertFingerprints(matches, car_model)
def test_fw_version_lists(self, subtests):
for car_model, ecus in FW_VERSIONS.items():
with subtests.test(car_model=car_model.value):
for ecu, ecu_fw in ecus.items():
with subtests.test(ecu):
duplicates = {fw for fw in ecu_fw if ecu_fw.count(fw) > 1}
assert not len(duplicates), f'{car_model}: Duplicate FW versions: Ecu.{ecu[0]}, {duplicates}'
assert len(ecu_fw) > 0, f'{car_model}: No FW versions: Ecu.{ecu[0]}'
def test_all_addrs_map_to_one_ecu(self):
for brand, cars in VERSIONS.items():
addr_to_ecu = defaultdict(set)
for ecus in cars.values():
for ecu_type, addr, sub_addr in ecus.keys():
addr_to_ecu[(addr, sub_addr)].add(ecu_type)
ecus_for_addr = addr_to_ecu[(addr, sub_addr)]
ecu_strings = ", ".join([f'Ecu.{ecu}' for ecu in ecus_for_addr])
assert len(ecus_for_addr) <= 1, f"{brand} has multiple ECUs that map to one address: {ecu_strings} -> ({hex(addr)}, {sub_addr})"
def test_data_collection_ecus(self, subtests):
# Asserts no extra ECUs are in the fingerprinting database
for brand, config in FW_QUERY_CONFIGS.items():
for car_model, ecus in VERSIONS[brand].items():
bad_ecus = set(ecus).intersection(config.extra_ecus)
with subtests.test(car_model=car_model.value):
assert not len(bad_ecus), f'{car_model}: Fingerprints contain ECUs added for data collection: {bad_ecus}'
def test_blacklisted_ecus(self, subtests):
blacklisted_addrs = (0x7c4, 0x7d0) # includes A/C ecu and an unknown ecu
for car_model, ecus in FW_VERSIONS.items():
with subtests.test(car_model=car_model.value):
CP = interfaces[car_model].get_non_essential_params(car_model)
if CP.brand == 'subaru':
for ecu in ecus.keys():
assert ecu[1] not in blacklisted_addrs, f'{car_model}: Blacklisted ecu: (Ecu.{ecu[0]}, {hex(ecu[1])})'
elif CP.brand == "chrysler":
# Some HD trucks have a combined TCM and ECM
if CP.carFingerprint.startswith("RAM_HD"):
for ecu in ecus.keys():
assert ecu[0] != Ecu.transmission, f"{car_model}: Blacklisted ecu: (Ecu.{ecu[0]}, {hex(ecu[1])})"
def test_missing_versions_and_configs(self, subtests):
brand_versions = set(VERSIONS.keys())
brand_configs = set(FW_QUERY_CONFIGS.keys())
if len(brand_configs - brand_versions):
with subtests.test():
pytest.fail(f"Brands do not implement FW_VERSIONS: {brand_configs - brand_versions}")
if len(brand_versions - brand_configs):
with subtests.test():
pytest.fail(f"Brands do not implement FW_QUERY_CONFIG: {brand_versions - brand_configs}")
# Ensure each brand has at least 1 ECU to query, and extra ECU retrieval
for brand, config in FW_QUERY_CONFIGS.items():
assert len(config.get_all_ecus({}, include_extra_ecus=False)) == 0
assert config.get_all_ecus({}) == set(config.extra_ecus)
if len(VERSIONS[brand]) > 0:
assert len(config.get_all_ecus(VERSIONS[brand])) > 0
def test_fw_request_ecu_whitelist(self, subtests):
for brand, config in FW_QUERY_CONFIGS.items():
with subtests.test(brand=brand):
whitelisted_ecus = {ecu for r in config.requests for ecu in r.whitelist_ecus}
brand_ecus = {fw[0] for car_fw in VERSIONS[brand].values() for fw in car_fw}
brand_ecus |= {ecu[0] for ecu in config.extra_ecus}
# each ecu in brand's fw versions + extra ecus needs to be whitelisted at least once
ecus_not_whitelisted = brand_ecus - whitelisted_ecus
ecu_strings = ", ".join([f'Ecu.{ecu}' for ecu in ecus_not_whitelisted])
assert not (len(whitelisted_ecus) and len(ecus_not_whitelisted)), \
f'{brand.title()}: ECUs not in any FW query whitelists: {ecu_strings}'
def test_request_ecus_in_versions(self):
# All ECUs in requests should be in the brand's FW versions
for brand, config in FW_QUERY_CONFIGS.items():
request_ecus = {ecu for r in config.requests for ecu in r.whitelist_ecus} - {ecu[0] for ecu in config.extra_ecus}
print(brand, request_ecus)
version_ecus = config.get_all_ecus(VERSIONS[brand], include_extra_ecus=False)
for request_ecu in request_ecus:
assert request_ecu in {e for e, _, _ in version_ecus}, f"Ecu.{ECU_NAME[request_ecu]} not in {brand} FW versions"
def test_brand_ecu_matches(self):
brand_matches = get_brand_ecu_matches(set())
assert len(brand_matches) > 0
assert all(len(e) and not any(e) for e in brand_matches.values())
# we ignore bus
brand_matches = get_brand_ecu_matches({(0x758, 0xf, 99)})
assert True in brand_matches['toyota']
assert not any(any(e) for b, e in brand_matches.items() if b != 'toyota')
class TestFwFingerprintTiming:
N: int = 5
TOL: float = 0.05
# for patched functions
current_obd_multiplexing: bool
total_time: float
@staticmethod
def fake_can_send(msgs):
pass
@staticmethod
def fake_can_recv(wait_for_one: bool = False) -> list[list[CanData]]:
return ([[CanData(random.randint(0x600, 0x800), b'\x00' * 8, 0)]]
if random.uniform(0, 1) > 0.5 else [])
def fake_set_obd_multiplexing(self, obd_multiplexing):
"""The 10Hz blocking params loop adds on average 50ms to the query time for each OBD multiplexing change"""
if obd_multiplexing != self.current_obd_multiplexing:
self.current_obd_multiplexing = obd_multiplexing
self.total_time += 0.1 / 2
def fake_get_data(self, timeout):
self.total_time += timeout
return {}
def _benchmark_brand(self, brand, num_pandas, mocker):
self.total_time = 0
mocker.patch("opendbc.car.isotp_parallel_query.IsoTpParallelQuery.get_data", self.fake_get_data)
for _ in range(self.N):
# Treat each brand as the most likely (aka, the first) brand with OBD multiplexing initially on
self.current_obd_multiplexing = True
t = time.perf_counter()
get_fw_versions(self.fake_can_recv, self.fake_can_send, self.fake_set_obd_multiplexing, brand, num_pandas=num_pandas)
self.total_time += time.perf_counter() - t
return self.total_time / self.N
def _assert_timing(self, avg_time, ref_time):
assert avg_time < ref_time + self.TOL
assert avg_time > ref_time - self.TOL, "Performance seems to have improved, update test refs."
def test_startup_timing(self, subtests, mocker):
# Tests worse-case VIN query time and typical present ECU query time
vin_ref_times = {'worst': 1.6, 'best': 0.8} # best assumes we go through all queries to get a match
present_ecu_ref_time = 0.45
def fake_get_ecu_addrs(*_, timeout):
self.total_time += timeout
return set()
self.total_time = 0.0
mocker.patch("opendbc.car.fw_versions.get_ecu_addrs", fake_get_ecu_addrs)
for _ in range(self.N):
self.current_obd_multiplexing = True
get_present_ecus(self.fake_can_recv, self.fake_can_send, self.fake_set_obd_multiplexing, num_pandas=2)
self._assert_timing(self.total_time / self.N, present_ecu_ref_time)
print(f'get_present_ecus, query time={self.total_time / self.N} seconds')
for name, args in (('worst', {}), ('best', {'retry': 1})):
with subtests.test(name=name):
self.total_time = 0.0
mocker.patch("opendbc.car.isotp_parallel_query.IsoTpParallelQuery.get_data", self.fake_get_data)
for _ in range(self.N):
get_vin(self.fake_can_recv, self.fake_can_send, (0, 1), **args)
self._assert_timing(self.total_time / self.N, vin_ref_times[name])
print(f'get_vin {name} case, query time={self.total_time / self.N} seconds')
def test_fw_query_timing(self, subtests, mocker):
total_ref_time = {1: 7.3, 2: 7.9}
brand_ref_times = {
1: {
'gm': 1.0,
'body': 0.1,
'chrysler': 0.3,
'ford': 1.5,
'honda': 0.45,
'hyundai': 0.65,
'mazda': 0.1,
'nissan': 0.8,
'subaru': 0.65,
'tesla': 0.1,
'toyota': 0.7,
'volkswagen': 0.65,
'rivian': 0.3,
},
2: {
'ford': 1.6,
'hyundai': 1.15,
}
}
total_times = {1: 0.0, 2: 0.0}
for num_pandas in (1, 2):
for brand, config in FW_QUERY_CONFIGS.items():
with subtests.test(brand=brand, num_pandas=num_pandas):
avg_time = self._benchmark_brand(brand, num_pandas, mocker)
total_times[num_pandas] += avg_time
avg_time = round(avg_time, 2)
ref_time = brand_ref_times[num_pandas].get(brand)
if ref_time is None:
# ref time should be same as 1 panda if no aux queries
ref_time = brand_ref_times[num_pandas - 1][brand]
self._assert_timing(avg_time, ref_time)
print(f'{brand=}, {num_pandas=}, {len(config.requests)=}, avg FW query time={avg_time} seconds')
for num_pandas in (1, 2):
with subtests.test(brand='all_brands', num_pandas=num_pandas):
total_time = round(total_times[num_pandas], 2)
self._assert_timing(total_time, total_ref_time[num_pandas])
print(f'all brands, total FW query time={total_time} seconds')
def test_get_fw_versions(self, subtests, mocker):
# some coverage on IsoTpParallelQuery and panda UDS library
# TODO: replace this with full fingerprint simulation testing
# https://github.com/commaai/panda/pull/1329
def fake_carlog_exception(*args, **kwargs):
raise
t = 0
def fake_monotonic():
nonlocal t
t += 0.0001
return t
mocker.patch("opendbc.car.carlog.carlog.exception", fake_carlog_exception)
mocker.patch("time.monotonic", fake_monotonic)
for brand in FW_QUERY_CONFIGS.keys():
with subtests.test(brand=brand):
get_fw_versions(self.fake_can_recv, self.fake_can_send, lambda obd: None, brand)