modelrevert: boot fixes + strip remaining variable-rate cruft

controlsd:
- stop writing the removed 'no_lat_lane_change' persistent param
  (key isn't in the whitelist anymore). Write to
  frogpilot_variables.no_lat_lane_change instead so the carcontroller
  read still works and the lane-change-disengage feature is preserved.
- initialize self.driving_gear = False in __init__; it's set at the
  end of step() in update_frogpilot_variables but clearpilot_state_control
  reads it on the first iteration before that's run.

modeld:
- remove the remaining CLEARPILOT variable-rate / standby code path
  that was baked into the initial-import commit (the one our first
  commit captured). Constant 20fps, no ModelStandby/ModelStandbyTs
  writes, no params_memory reads for 'no_lat_lane_change' (which
  isn't registered anymore).
- drop now-unused locals (model_standby, last_standby_ts_write,
  params_memory).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

common/realtime.py

@@ -12,7 +12,7 @@ from openpilot.system.hardware import PC
 # time step for each process
 DT_CTRL = 0.01  # controlsd
 DT_MDL = 0.05  # model
-DT_TRML = 0.5  # thermald and manager
+DT_TRML = 0.25  # thermald and manager — 4 Hz
 DT_DMON = 0.05  # driver monitoring
 
 

selfdrive/car/hyundai/carstate.py

@@ -13,6 +13,7 @@ from openpilot.selfdrive.car.hyundai.hyundaicanfd import CanBus
 from openpilot.selfdrive.car.hyundai.values import HyundaiFlags, CAR, DBC, CAN_GEARS, CAMERA_SCC_CAR, \
                                                    CANFD_CAR, Buttons, CarControllerParams
 from openpilot.selfdrive.car.interfaces import CarStateBase
+from openpilot.selfdrive.clearpilot.telemetry import tlog
 
 PREV_BUTTON_SAMPLES = 8
 CLUSTER_SAMPLE_RATE = 20  # frames
@@ -47,6 +48,10 @@ class CarState(CarStateBase):
     self.is_metric = False
     self.buttons_counter = 0
 
+    # CLEARPILOT: cache to avoid per-cycle atomic writes to /dev/shm (eats CPU via fsync/flock)
+    self._prev_car_speed_limit = None
+    self._prev_car_is_metric = None
+
     self.cruise_info = {}
 
     # On some cars, CLU15->CF_Clu_VehicleSpeed can oscillate faster than the dash updates. Sample at 5 Hz
@@ -209,10 +214,15 @@ class CarState(CarStateBase):
       self.lkas_previously_enabled = self.lkas_enabled
       self.lkas_enabled = cp.vl["BCM_PO_11"]["LFA_Pressed"]
 
-    # self.params_memory.put_int("CarSpeedLimitLiteral", self.calculate_speed_limit(cp, cp_cam))
-    self.params_memory.put_float("CarSpeedLimit", self.calculate_speed_limit(cp, cp_cam) * speed_conv)
-    self.params_memory.put_float("CarCruiseDisplayActual", cp_cruise.vl["SCC11"]["VSetDis"])
-    
+    # CLEARPILOT: gate on change — see same fix in update_canfd
+    car_speed_limit = self.calculate_speed_limit(cp, cp_cam) * speed_conv
+    if car_speed_limit != self._prev_car_speed_limit:
+      self.params_memory.put_float("CarSpeedLimit", car_speed_limit)
+      self._prev_car_speed_limit = car_speed_limit
+    if self.is_metric != self._prev_car_is_metric:
+      self.params_memory.put("CarIsMetric", "1" if self.is_metric else "0")
+      self._prev_car_is_metric = self.is_metric
+
 
     return ret
 
@@ -415,63 +425,23 @@ class CarState(CarStateBase):
     #   nonAdaptive = false,
     #   speedCluster = 0 )
 
-    # print("Set limit")
-    # print(self.calculate_speed_limit(cp, cp_cam))
-    # self.params_memory.put_float("CarSpeedLimitLiteral", self.calculate_speed_limit(cp, cp_cam))
-    self.params_memory.put_float("CarSpeedLimit", self.calculate_speed_limit(cp, cp_cam) * speed_factor)
+    # CLEARPILOT: gate on change — these writes run 100Hz, each is an atomic fsync/flock transaction
+    car_speed_limit = self.calculate_speed_limit(cp, cp_cam) * speed_factor
+    if car_speed_limit != self._prev_car_speed_limit:
+      self.params_memory.put_float("CarSpeedLimit", car_speed_limit)
+      self._prev_car_speed_limit = car_speed_limit
+    if self.is_metric != self._prev_car_is_metric:
+      self.params_memory.put("CarIsMetric", "1" if self.is_metric else "0")
+      self._prev_car_is_metric = self.is_metric
 
-    # CLEARPILOT: CAN-FD telemetry — preserved but disabled. Re-enable by uncommenting (also restore the import).
-    # from openpilot.selfdrive.clearpilot.telemetry import tlog
-    #
+    # CLEARPILOT: telemetry logging — disabled, re-enable when needed
     # speed_limit_bus = cp if self.CP.flags & HyundaiFlags.CANFD_HDA2 else cp_cam
     # scc = cp_cam.vl["SCC_CONTROL"] if self.CP.flags & HyundaiFlags.CANFD_CAMERA_SCC else cp.vl["SCC_CONTROL"]
     # cluster = speed_limit_bus.vl["CLUSTER_SPEED_LIMIT"]
-    #
-    # tlog("car", {
-    #   "vEgo": round(ret.vEgo, 3),
-    #   "vEgoRaw": round(ret.vEgoRaw, 3),
-    #   "aEgo": round(ret.aEgo, 3),
-    #   "steeringAngleDeg": round(ret.steeringAngleDeg, 1),
-    #   "gear": str(ret.gearShifter),
-    #   "brakePressed": ret.brakePressed,
-    #   "gasPressed": ret.gasPressed,
-    #   "standstill": ret.standstill,
-    #   "leftBlinker": ret.leftBlinker,
-    #   "rightBlinker": ret.rightBlinker,
-    # })
-    #
-    # tlog("cruise", {
-    #   "enabled": ret.cruiseState.enabled,
-    #   "available": ret.cruiseState.available,
-    #   "speed": round(ret.cruiseState.speed, 3),
-    #   "standstill": ret.cruiseState.standstill,
-    #   "accFaulted": ret.accFaulted,
-    #   "ACCMode": scc.get("ACCMode", 0),
-    #   "VSetDis": scc.get("VSetDis", 0),
-    #   "aReqRaw": round(scc.get("aReqRaw", 0), 3),
-    #   "aReqValue": round(scc.get("aReqValue", 0), 3),
-    #   "DISTANCE_SETTING": scc.get("DISTANCE_SETTING", 0),
-    #   "ACC_ObjDist": round(scc.get("ACC_ObjDist", 0), 1),
-    # })
-    #
-    # tlog("speed_limit", {
-    #   "SPEED_LIMIT_1": cluster.get("SPEED_LIMIT_1", 0),
-    #   "SPEED_LIMIT_2": cluster.get("SPEED_LIMIT_2", 0),
-    #   "SPEED_LIMIT_3": cluster.get("SPEED_LIMIT_3", 0),
-    #   "SCHOOL_ZONE": cluster.get("SCHOOL_ZONE", 0),
-    #   "CHIME_1": cluster.get("CHIME_1", 0),
-    #   "CHIME_2": cluster.get("CHIME_2", 0),
-    #   "SPEED_CHANGE_BLINKING": cluster.get("SPEED_CHANGE_BLINKING", 0),
-    #   "calculated": self.calculate_speed_limit(cp, cp_cam),
-    #   "is_metric": self.is_metric,
-    # })
-    #
-    # tlog("buttons", {
-    #   "cruise_button": self.cruise_buttons[-1],
-    #   "main_button": self.main_buttons[-1],
-    #   "lkas_enabled": self.lkas_enabled,
-    #   "main_enabled": self.main_enabled,
-    # })
+    # tlog("car", { ... })
+    # tlog("cruise", { ... })
+    # tlog("speed_limit", { ... })
+    # tlog("buttons", { ... })
 
     return ret
 

selfdrive/car/interfaces.py

@@ -484,10 +484,19 @@ class CarInterfaceBase(ABC):
           self.silent_steer_warning = True
           events.add(EventName.steerTempUnavailableSilent)
         else:
+          # CLEARPILOT: log once per instance of this warning
+          if not getattr(self, '_steer_fault_logged', False):
+            import sys
+            print(f"CLP steerTempUnavailable: steerFaultTemporary={cs_out.steerFaultTemporary} "
+                  f"steeringPressed={cs_out.steeringPressed} standstill={cs_out.standstill} "
+                  f"steering_unpressed={self.steering_unpressed} steeringAngleDeg={cs_out.steeringAngleDeg:.1f} "
+                  f"steeringTorque={cs_out.steeringTorque:.1f} vEgo={cs_out.vEgo:.2f}", file=sys.stderr)
+            self._steer_fault_logged = True
           events.add(EventName.steerTempUnavailable)
     else:
       self.no_steer_warning = False
       self.silent_steer_warning = False
+      self._steer_fault_logged = False
     if cs_out.steerFaultPermanent:
       events.add(EventName.steerUnavailable)
 

selfdrive/controls/controlsd.py

@@ -37,10 +37,9 @@ from openpilot.system.version import get_short_branch
 from openpilot.selfdrive.frogpilot.controls.lib.frogpilot_functions import CRUISING_SPEED, PROBABILITY, MovingAverageCalculator
 
 from openpilot.selfdrive.frogpilot.controls.lib.model_manager import RADARLESS_MODELS
-from openpilot.selfdrive.frogpilot.controls.lib.speed_limit_controller import SpeedLimitController
-
-# CLEARPILOT: UI plumbing for ScreenDisplayMode and the speed/cruise-warning overlay
+from openpilot.selfdrive.clearpilot.telemetry import tlog
 from openpilot.selfdrive.clearpilot.speed_logic import SpeedState
+from openpilot.selfdrive.frogpilot.controls.lib.speed_limit_controller import SpeedLimitController
 
 SOFT_DISABLE_TIME = 3  # seconds
 LDW_MIN_SPEED = 31 * CV.MPH_TO_MS
@@ -50,7 +49,7 @@ CAMERA_OFFSET = 0.04
 REPLAY = "REPLAY" in os.environ
 SIMULATION = "SIMULATION" in os.environ
 TESTING_CLOSET = "TESTING_CLOSET" in os.environ
-IGNORE_PROCESSES = {"loggerd", "encoderd", "statsd"}
+IGNORE_PROCESSES = {"loggerd", "encoderd", "statsd", "telemetryd", "dashcamd"}
 
 ThermalStatus = log.DeviceState.ThermalStatus
 State = log.ControlsState.OpenpilotState
@@ -80,14 +79,13 @@ class Controls:
     self.params_storage = Params("/persist/params")
 
     self.params_memory.put_bool("CPTLkasButtonAction", False)
-    # CLEARPILOT: ScreenDisplayMode is an int (5-state machine: 0..4); UI reads it via getInt
     self.params_memory.put_int("ScreenDisplayMode", 0)
 
-    # CLEARPILOT: speed/cruise-warning overlay state, ticked at ~2Hz from clearpilot_state_control
+    # CLEARPILOT: speed-limit/cruise-warning state machine + park→drive auto-reset
     self.speed_state = SpeedState()
     self.speed_state_frame = 0
-    # CLEARPILOT: edge tracking for park->drive auto-wake of screen
     self.was_driving_gear = False
+    self.driving_gear = False
 
     self.radarless_model = self.params.get("Model", encoding='utf-8') in RADARLESS_MODELS
 
@@ -451,6 +449,13 @@ class Controls:
     # All events here should at least have NO_ENTRY and SOFT_DISABLE.
     num_events = len(self.events)
 
+    # CLEARPILOT: compute model standby suppression early — used by multiple checks below
+    try:
+      standby_ts = float(self.params_memory.get("ModelStandbyTs") or "0")
+    except (ValueError, TypeError):
+      standby_ts = 0
+    model_suppress = (time.monotonic() - standby_ts) < 2.0
+
     not_running = {p.name for p in self.sm['managerState'].processes if not p.running and p.shouldBeRunning}
     if self.sm.recv_frame['managerState'] and (not_running - IGNORE_PROCESSES):
       self.events.add(EventName.processNotRunning)
@@ -464,9 +469,11 @@ class Controls:
         elif not self.sm.all_freq_ok(self.camera_packets):
           self.events.add(EventName.cameraFrameRate)
     if not REPLAY and self.rk.lagging:
+      import sys
+      print(f"CLP controlsdLagging: remaining={self.rk.remaining:.4f} standstill={CS.standstill} vEgo={CS.vEgo:.2f}", file=sys.stderr)
       self.events.add(EventName.controlsdLagging)
     if not self.radarless_model:
-      if len(self.sm['radarState'].radarErrors) or (not self.rk.lagging and not self.sm.all_checks(['radarState'])):
+      if not model_suppress and (len(self.sm['radarState'].radarErrors) or (not self.rk.lagging and not self.sm.all_checks(['radarState']))):
         self.events.add(EventName.radarFault)
     if not self.sm.valid['pandaStates']:
       self.events.add(EventName.usbError)
@@ -478,7 +485,7 @@ class Controls:
     # generic catch-all. ideally, a more specific event should be added above instead
     has_disable_events = self.events.contains(ET.NO_ENTRY) and (self.events.contains(ET.SOFT_DISABLE) or self.events.contains(ET.IMMEDIATE_DISABLE))
     no_system_errors = (not has_disable_events) or (len(self.events) == num_events)
-    if (not self.sm.all_checks() or self.card.can_rcv_timeout) and no_system_errors:
+    if (not self.sm.all_checks() or self.card.can_rcv_timeout) and no_system_errors and not model_suppress:
       if not self.sm.all_alive():
         self.events.add(EventName.commIssue)
       elif not self.sm.all_freq_ok():
@@ -499,13 +506,13 @@ class Controls:
       self.logged_comm_issue = None
 
     if not (self.CP.notCar and self.joystick_mode):
-      if not self.sm['liveLocationKalman'].posenetOK:
+      if not self.sm['liveLocationKalman'].posenetOK and not model_suppress:
         self.events.add(EventName.posenetInvalid)
       if not self.sm['liveLocationKalman'].deviceStable:
         self.events.add(EventName.deviceFalling)
-      if not self.sm['liveLocationKalman'].inputsOK:
+      if not self.sm['liveLocationKalman'].inputsOK and not model_suppress:
         self.events.add(EventName.locationdTemporaryError)
-      if not self.sm['liveParameters'].valid and not TESTING_CLOSET and (not SIMULATION or REPLAY):
+      if not self.sm['liveParameters'].valid and not TESTING_CLOSET and (not SIMULATION or REPLAY) and not model_suppress:
         self.events.add(EventName.paramsdTemporaryError)
 
     # conservative HW alert. if the data or frequency are off, locationd will throw an error
@@ -551,7 +558,7 @@ class Controls:
         self.distance_traveled = 0
       self.distance_traveled += CS.vEgo * DT_CTRL
 
-      if self.sm['modelV2'].frameDropPerc > 20:
+      if self.sm['modelV2'].frameDropPerc > 20 and not model_suppress:
         self.events.add(EventName.modeldLagging)
 
 
@@ -638,6 +645,14 @@ class Controls:
     # Check if openpilot is engaged and actuators are enabled
     self.enabled = self.state in ENABLED_STATES
     self.active = self.state in ACTIVE_STATES
+
+    # CLEARPILOT: engagement telemetry disabled — was running at 100Hz, causing CPU load
+    # tlog("engage", {
+    #   "state": self.state.name if hasattr(self.state, 'name') else str(self.state),
+    #   "enabled": self.enabled, "active": self.active,
+    #   "cruise_enabled": CS.cruiseState.enabled, "cruise_available": CS.cruiseState.available,
+    #   "brakePressed": CS.brakePressed,
+    # })
     if self.active:
       self.current_alert_types.append(ET.WARNING)
 
@@ -686,11 +701,8 @@ class Controls:
 
     if model_v2.meta.laneChangeState == LaneChangeState.laneChangeStarting and clearpilot_disable_lat_on_lane_change:
       CC.latActive = False
-      self.params_memory.put_bool("no_lat_lane_change", True)
-      # CLEARPILOT: hyundai carcontroller reads this off frogpilot_variables (not Params) — keep both in sync
       self.frogpilot_variables.no_lat_lane_change = True
     else:
-      self.params_memory.put_bool("no_lat_lane_change", False)
       self.frogpilot_variables.no_lat_lane_change = False
 
     if CS.leftBlinker or CS.rightBlinker:
@@ -1246,47 +1258,40 @@ class Controls:
     self.frogpilot_variables.use_ev_tables = self.params.get_bool("EVTable")
 
   def update_clearpilot_events(self, CS):
-    if (len(CS.buttonEvents) > 0): 
+    if (len(CS.buttonEvents) > 0):
       print (CS.buttonEvents)
-
-    # Uncomment to alert when lkas button pressed
-    # if any(be.pressed and be.type == FrogPilotButtonType.lkas for be in CS.buttonEvents):
-      # self.events.add(EventName.clpDebug)
+    if any(be.pressed and be.type == FrogPilotButtonType.lkas for be in CS.buttonEvents):
+      self.events.add(EventName.clpDebug)
 
   def clearpilot_state_control(self, CC, CS):
-    # CLEARPILOT: pure UI plumbing — does not modify CC/actuators. Maintains
-    # ScreenDisplayMode (5-state machine driven by the LFA/debug button + gear
-    # edges) and ticks the speed/cruise-warning overlay at ~2Hz.
-    driving_gear = CS.gearShifter not in (GearShifter.neutral, GearShifter.park,
-                                          GearShifter.reverse, GearShifter.unknown)
-
-    # Auto-wake screen when shifting into drive from screen-off
-    if driving_gear and not self.was_driving_gear:
+    # CLEARPILOT: auto-reset display when shifting into drive from screen-off
+    if self.driving_gear and not self.was_driving_gear:
       if self.params_memory.get_int("ScreenDisplayMode") == 3:
         self.params_memory.put_int("ScreenDisplayMode", 0)
-    self.was_driving_gear = driving_gear
+    self.was_driving_gear = self.driving_gear
 
-    # LFA/debug button cycles ScreenDisplayMode. Onroad and offroad use
-    # different transition tables.
     if any(be.pressed and be.type == FrogPilotButtonType.lkas for be in CS.buttonEvents):
       current = self.params_memory.get_int("ScreenDisplayMode")
-      if driving_gear:
+
+      if self.driving_gear:
         # Onroad: 0→4, 1→2, 2→3, 3→4, 4→2 (never back to auto via button)
         transitions = {0: 4, 1: 2, 2: 3, 3: 4, 4: 2}
         new_mode = transitions.get(current, 0)
       else:
-        # Not in drive: anything except 3 → 3 (screen off), state 3 → 0 (auto)
+        # Not in drive: any except 3 → 3, state 3 → 0
         new_mode = 0 if current == 3 else 3
+
       self.params_memory.put_int("ScreenDisplayMode", new_mode)
 
-    # Speed/cruise-warning overlay tick (~2Hz at 100Hz loop)
+    # ClearPilot speed processing (~2 Hz at 100 Hz loop) — drives speed-limit sign
+    # and cruise over/under warning sign via memory params read by the UI.
     self.speed_state_frame += 1
     if self.speed_state_frame % 50 == 0:
       gps = self.sm['gpsLocation']
       has_gps = self.sm.valid['gpsLocation'] and gps.hasFix
       speed_ms = gps.speed if has_gps else 0.0
-      speed_limit_ms = self.params_memory.get_float("CarSpeedLimit") or 0.0
-      is_metric = self.is_metric
+      speed_limit_ms = self.params_memory.get_float("CarSpeedLimit")
+      is_metric = (self.params_memory.get("CarIsMetric", encoding="utf-8") or "0") == "1"
       cruise_speed_ms = CS.cruiseState.speed
       cruise_active = CS.cruiseState.enabled
       cruise_standstill = CS.cruiseState.standstill

selfdrive/controls/lib/events.py

@@ -778,8 +778,8 @@ EVENTS: dict[int, dict[str, Alert | AlertCallbackType]] = {
     ET.SOFT_DISABLE: soft_disable_alert("Sensor Data Invalid"),
   },
 
+  # CLEARPILOT: alert suppressed — event still fires for screen toggle and future actions
   EventName.clpDebug: {
-    ET.PERMANENT: clp_debug_notice,
   },
 
   EventName.noGps: {

selfdrive/frogpilot/controls/frogpilot_planner.py

@@ -58,6 +58,9 @@ class FrogPilotPlanner:
     self.params = Params()
     self.params_memory = Params("/dev/shm/params")
 
+    # CLEARPILOT: track valid transitions so we only log when it flips, not every cycle
+    self._dbg_prev_valid = True
+
     self.cem = ConditionalExperimentalMode()
     self.lead_one = Lead()
     self.mtsc = MapTurnSpeedController()
@@ -242,7 +245,18 @@ class FrogPilotPlanner:
 
   def publish(self, sm, pm):
     frogpilot_plan_send = messaging.new_message('frogpilotPlan')
-    frogpilot_plan_send.valid = sm.all_checks(service_list=['carState', 'controlsState'])
+    valid = sm.all_checks(service_list=['carState', 'controlsState'])
+    # CLEARPILOT: log on transition into invalid — stderr goes to our plannerd.log
+    if valid != self._dbg_prev_valid and not valid:
+      import sys
+      print(
+        "CLP frogpilotPlan valid=False: "
+        f"carState(a={sm.alive['carState']},v={sm.valid['carState']},f={sm.freq_ok['carState']}) "
+        f"controlsState(a={sm.alive['controlsState']},v={sm.valid['controlsState']},f={sm.freq_ok['controlsState']})",
+        file=sys.stderr, flush=True
+      )
+    self._dbg_prev_valid = valid
+    frogpilot_plan_send.valid = valid
     frogpilotPlan = frogpilot_plan_send.frogpilotPlan
 
     frogpilotPlan.accelerationJerk = A_CHANGE_COST * (float(self.jerk) if self.lead_one.status else 1)

selfdrive/modeld/modeld.py

@@ -134,11 +134,15 @@ def main(demo=False):
   setproctitle(PROCESS_NAME)
   config_realtime_process(7, 54)
 
+  import time as _time
   cloudlog.warning("setting up CL context")
+  _t0 = _time.monotonic()
   cl_context = CLContext()
-  cloudlog.warning("CL context ready; loading model")
+  _t1 = _time.monotonic()
+  cloudlog.warning("CL context ready in %.3fs; loading model", _t1 - _t0)
   model = ModelState(cl_context)
-  cloudlog.warning("models loaded, modeld starting")
+  _t2 = _time.monotonic()
+  cloudlog.warning("model loaded in %.3fs (total init %.3fs), modeld starting", _t2 - _t1, _t2 - _t0)
 
   # visionipc clients
   while True:
@@ -233,6 +237,12 @@ def main(demo=False):
       meta_extra = meta_main
 
     sm.update(0)
+
+    # CLEARPILOT: constant 20fps. Variable-rate + standby logic removed — the
+    # variable-rate path caused freq_ok cascades in downstream consumers
+    # (calibrationd/locationd/paramsd). Running at the camera's native rate is
+    # simpler and keeps the full-stack localization chain happy.
+
     desire = DH.desire
     is_rhd = sm["driverMonitoringState"].isRHD
     frame_id = sm["roadCameraState"].frameId

selfdrive/monitoring/dmonitoringd.py

@@ -21,6 +21,7 @@ def dmonitoringd_thread():
 
   v_cruise_last = 0
   driver_engaged = False
+  dbg_prev_valid = True  # CLEARPILOT: track valid transitions
 
   # 10Hz <- dmonitoringmodeld
   while True:
@@ -43,7 +44,18 @@ def dmonitoringd_thread():
     # Get data from dmonitoringmodeld
     events = Events()
 
-    if sm.all_checks() and len(sm['liveCalibration'].rpyCalib):
+    # CLEARPILOT: narrow update_states gate. The original sm.all_checks() also
+    # required modelV2 fresh (stops at standstill in two-state modeld) and
+    # liveCalibration.valid (calibrationd cascades its own freq_ok to valid, which
+    # flaps). Both made DM freeze pose → face_detected stuck False → awareness
+    # decayed to 0 within 6s of engagement. Narrow the gate to the subs
+    # update_states actually reads, and only to alive+valid (skip freq_ok and
+    # skip liveCalibration.valid). rpyCalib presence is sufficient to know
+    # calibration has produced output.
+    if (sm.alive['driverStateV2'] and sm.valid['driverStateV2'] and
+        sm.alive['carState'] and sm.valid['carState'] and
+        sm.alive['controlsState'] and sm.valid['controlsState'] and
+        sm.alive['liveCalibration'] and len(sm['liveCalibration'].rpyCalib) > 0):
       driver_status.update_states(sm['driverStateV2'], sm['liveCalibration'].rpyCalib, sm['carState'].vEgo, sm['controlsState'].enabled)
 
     # Block engaging after max number of distrations
@@ -54,8 +66,16 @@ def dmonitoringd_thread():
     # Update events from driver state
     driver_status.update_events(events, driver_engaged, sm['controlsState'].enabled, sm['carState'].standstill)
 
+    # CLEARPILOT: log on transition to invalid so we can see which sub caused the cascade
+    dm_valid = sm.all_checks()
+    if dm_valid != dbg_prev_valid and not dm_valid:
+      import sys
+      bad = [s for s in sm.alive if not (sm.alive[s] and sm.valid[s] and sm.freq_ok.get(s, True))]
+      details = [f"{s}(a={sm.alive[s]},v={sm.valid[s]},f={sm.freq_ok[s]})" for s in bad]
+      print(f"CLP driverMonitoringState valid=False: {' '.join(details)}", file=sys.stderr, flush=True)
+    dbg_prev_valid = dm_valid
     # build driverMonitoringState packet
-    dat = messaging.new_message('driverMonitoringState', valid=sm.all_checks())
+    dat = messaging.new_message('driverMonitoringState', valid=dm_valid)
     dat.driverMonitoringState = {
       "events": events.to_msg(),
       "faceDetected": driver_status.face_detected,