Harden BT Locate handoff matching and start flow

utils/bt_locate.py

@@ -7,12 +7,13 @@ distance estimation, and proximity alerts for search and rescue operations.
 
 from __future__ import annotations
 
-import logging
-import queue
-import threading
-from dataclasses import dataclass
-from datetime import datetime
-from enum import Enum
+import logging
+import queue
+import threading
+import time
+from dataclasses import dataclass, field
+from datetime import datetime
+from enum import Enum
 
 from utils.bluetooth.models import BTDeviceAggregate
 from utils.bluetooth.scanner import BluetoothScanner, get_bluetooth_scanner
@@ -20,12 +21,17 @@ from utils.gps import get_current_position
 
 logger = logging.getLogger('intercept.bt_locate')
 
-# Maximum trail points to retain
-MAX_TRAIL_POINTS = 500
-
-# EMA smoothing factor for RSSI
+# Maximum trail points to retain
+MAX_TRAIL_POINTS = 500
+
+# EMA smoothing factor for RSSI
 EMA_ALPHA = 0.3
 
+# Polling/restart tuning for scanner resilience without high CPU churn.
+POLL_INTERVAL_SECONDS = 1.5
+SCAN_RESTART_BACKOFF_SECONDS = 8.0
+NO_MATCH_LOG_EVERY_POLLS = 10
+
 
 def _normalize_mac(address: str | None) -> str | None:
     """Normalize MAC string to colon-separated uppercase form when possible."""
@@ -47,6 +53,22 @@ def _normalize_mac(address: str | None) -> str | None:
 
     # Return cleaned original when not a strict MAC (caller may still use exact matching)
     return text
+
+
+def _address_looks_like_rpa(address: str | None) -> bool:
+    """
+    Return True when an address looks like a Resolvable Private Address.
+
+    RPA check: most-significant two bits of the first octet are `01`.
+    """
+    normalized = _normalize_mac(address)
+    if not normalized:
+        return False
+    try:
+        first_octet = int(normalized.split(':', 1)[0], 16)
+    except (ValueError, TypeError):
+        return False
+    return (first_octet >> 6) == 1
 
 
 class Environment(Enum):
@@ -116,8 +138,27 @@ class LocateTarget:
     known_name: str | None = None
     known_manufacturer: str | None = None
     last_known_rssi: int | None = None
+    _cached_irk_hex: str | None = field(default=None, init=False, repr=False)
+    _cached_irk_bytes: bytes | None = field(default=None, init=False, repr=False)
 
-    def matches(self, device: BTDeviceAggregate) -> bool:
+    def _get_irk_bytes(self) -> bytes | None:
+        """Parse/cache target IRK bytes once for repeated match checks."""
+        if not self.irk_hex:
+            return None
+        if self._cached_irk_hex == self.irk_hex:
+            return self._cached_irk_bytes
+        self._cached_irk_hex = self.irk_hex
+        self._cached_irk_bytes = None
+        try:
+            parsed = bytes.fromhex(self.irk_hex)
+        except (ValueError, TypeError):
+            return None
+        if len(parsed) != 16:
+            return None
+        self._cached_irk_bytes = parsed
+        return parsed
+
+    def matches(self, device: BTDeviceAggregate, irk_bytes: bytes | None = None) -> bool:
         """Check if a device matches this target."""
         # Match by stable device key (survives MAC randomization for many devices)
         if self.device_key and getattr(device, 'device_key', None) == self.device_key:
@@ -141,21 +182,23 @@ class LocateTarget:
             if dev_addr and target_addr and dev_addr == target_addr:
                 return True
 
-        # Match by payload fingerprint (guard against low-stability generic fingerprints)
+        # Match by payload fingerprint.
+        # For explicit hand-off sessions, allow exact fingerprint matches even if
+        # stability is still warming up.
         if self.fingerprint_id:
             dev_fp = getattr(device, 'payload_fingerprint_id', None)
             dev_fp_stability = getattr(device, 'payload_fingerprint_stability', 0.0) or 0.0
-            if dev_fp and dev_fp == self.fingerprint_id and dev_fp_stability >= 0.35:
-                return True
+            if dev_fp and dev_fp == self.fingerprint_id:
+                if dev_fp_stability >= 0.35:
+                    return True
+                if any([self.device_id, self.device_key, self.mac_address, self.known_name]):
+                    return True
 
         # Match by RPA resolution
-        if self.irk_hex:
-            try:
-                irk = bytes.fromhex(self.irk_hex)
-                if len(irk) == 16 and device.address and resolve_rpa(irk, device.address):
-                    return True
-            except (ValueError, TypeError):
-                pass
+        if self.irk_hex and device.address and _address_looks_like_rpa(device.address):
+            irk = irk_bytes or self._get_irk_bytes()
+            if irk and resolve_rpa(irk, device.address):
+                return True
 
         # Match by name pattern
         if self.name_pattern and device.name and self.name_pattern.lower() in device.name.lower():
@@ -257,7 +300,7 @@ class LocateSession:
         self.environment = environment
         self.fallback_lat = fallback_lat
         self.fallback_lon = fallback_lon
-        self._lock = threading.Lock()
+        self._lock = threading.Lock()
 
         # Distance estimator
         n = custom_exponent if environment == Environment.CUSTOM and custom_exponent else environment.value
@@ -281,7 +324,9 @@ class LocateSession:
         # Debug counters
         self.callback_call_count = 0
         self.poll_count = 0
-        self._last_seen_device: str | None = None
+        self._last_seen_device: str | None = None
+        self._last_scan_restart_attempt = 0.0
+        self._target_irk = target._get_irk_bytes()
 
         # Scanner reference
         self._scanner: BluetoothScanner | None = None
@@ -304,6 +349,7 @@ class LocateSession:
         if not self._scanner.is_scanning:
             logger.info("BT scanner not running, starting scan for locate session")
             self._scanner_started_by_us = True
+            self._last_scan_restart_attempt = time.monotonic()
             if not self._scanner.start_scan(mode='auto'):
                 # Surface startup failure to caller and avoid leaving stale callbacks.
                 status = self._scanner.get_status()
@@ -342,37 +388,40 @@ class LocateSession:
 
     def _poll_loop(self) -> None:
         """Poll scanner aggregator for target device updates."""
-        while not self._stop_event.is_set():
-            self._stop_event.wait(timeout=1.5)
-            if self._stop_event.is_set():
-                break
-            try:
-                self._check_aggregator()
-            except Exception as e:
-                logger.error(f"Locate poll error: {e}")
+        while not self._stop_event.is_set():
+            self._stop_event.wait(timeout=POLL_INTERVAL_SECONDS)
+            if self._stop_event.is_set():
+                break
+            try:
+                self._check_aggregator()
+            except Exception as e:
+                logger.error(f"Locate poll error: {e}")
 
     def _check_aggregator(self) -> None:
         """Check the scanner's aggregator for the target device."""
         if not self._scanner:
             return
 
-        self.poll_count += 1
-
-        # Restart scan if it expired (bleak 10s timeout)
-        if not self._scanner.is_scanning:
-            logger.info("Scanner stopped, restarting for locate session")
-            self._scanner.start_scan(mode='auto')
-
-        # Check devices seen within a recent window.  Using a short window
+        self.poll_count += 1
+
+        # Restart scan if it expired (bleak 10s timeout)
+        if not self._scanner.is_scanning:
+            now = time.monotonic()
+            if (now - self._last_scan_restart_attempt) >= SCAN_RESTART_BACKOFF_SECONDS:
+                self._last_scan_restart_attempt = now
+                logger.info("Scanner stopped, restarting for locate session")
+                self._scanner.start_scan(mode='auto')
+
+        # Check devices seen within a recent window.  Using a short window
         # (rather than the aggregator's full 120s) so that once a device
         # goes silent its stale RSSI stops producing detections.  The window
         # must survive bleak's 10s scan cycle + restart gap (~3s).
         devices = self._scanner.get_devices(max_age_seconds=15)
-        found_target = False
-        for device in devices:
-            if not self.target.matches(device):
-                continue
-            found_target = True
+        found_target = False
+        for device in devices:
+            if not self.target.matches(device, irk_bytes=self._target_irk):
+                continue
+            found_target = True
             rssi = device.rssi_current
             if rssi is None:
                 continue
@@ -380,10 +429,14 @@ class LocateSession:
             break  # One match per poll cycle is sufficient
 
         # Log periodically for debugging
-        if self.poll_count % 20 == 0 or (self.poll_count <= 5) or not found_target:
-            logger.info(
-                f"Poll #{self.poll_count}: {len(devices)} devices, "
-                f"target_found={found_target}, "
+        if (
+            self.poll_count <= 5
+            or self.poll_count % 20 == 0
+            or (not found_target and self.poll_count % NO_MATCH_LOG_EVERY_POLLS == 0)
+        ):
+            logger.info(
+                f"Poll #{self.poll_count}: {len(devices)} devices, "
+                f"target_found={found_target}, "
                 f"detections={self.detection_count}, "
                 f"scanning={self._scanner.is_scanning}"
             )
@@ -396,8 +449,8 @@ class LocateSession:
         self.callback_call_count += 1
         self._last_seen_device = f"{device.device_id}|{device.name}"
 
-        if not self.target.matches(device):
-            return
+        if not self.target.matches(device, irk_bytes=self._target_irk):
+            return
 
         rssi = device.rssi_current
         if rssi is None:
@@ -425,13 +478,9 @@ class LocateSession:
         band = DistanceEstimator.proximity_band(distance)
 
         # Check RPA resolution
-        rpa_resolved = False
-        if self.target.irk_hex and device.address:
-            try:
-                irk = bytes.fromhex(self.target.irk_hex)
-                rpa_resolved = resolve_rpa(irk, device.address)
-            except (ValueError, TypeError):
-                pass
+        rpa_resolved = False
+        if self._target_irk and device.address and _address_looks_like_rpa(device.address):
+            rpa_resolved = resolve_rpa(self._target_irk, device.address)
 
         # GPS tag — prefer live GPS, fall back to user-set coordinates
         gps_pos = get_current_position()
@@ -493,15 +542,15 @@ class LocateSession:
         with self._lock:
             return [p.to_dict() for p in self.trail if p.lat is not None]
 
-    def get_status(self) -> dict:
-        """Get session status."""
-        gps_pos = get_current_position()
+    def get_status(self, include_debug: bool = False) -> dict:
+        """Get session status."""
+        gps_pos = get_current_position()
 
         # Collect scanner/aggregator data OUTSIDE self._lock to avoid ABBA
         # deadlock: get_status would hold self._lock then wait on
         # aggregator._lock, while _poll_loop holds aggregator._lock then
         # waits on self._lock in _record_detection.
-        debug_devices = self._debug_device_sample()
+        debug_devices = self._debug_device_sample() if include_debug else []
         scanner_running = self._scanner.is_scanning if self._scanner else False
         scanner_device_count = self._scanner.device_count if self._scanner else 0
         callback_registered = (
@@ -537,8 +586,8 @@ class LocateSession:
                 'latest_rssi_ema': round(self.trail[-1].rssi_ema, 1) if self.trail else None,
                 'latest_distance': round(self.trail[-1].estimated_distance, 2) if self.trail else None,
                 'latest_band': self.trail[-1].proximity_band if self.trail else None,
-                'debug_devices': debug_devices,
-            }
+                'debug_devices': debug_devices,
+            }
 
     def set_environment(self, environment: Environment, custom_exponent: float | None = None) -> None:
         """Update the environment and recalculate distance estimator."""
@@ -553,16 +602,16 @@ class LocateSession:
             return []
         try:
             devices = self._scanner.get_devices(max_age_seconds=30)
-            return [
-                {
-                    'id': d.device_id,
-                    'addr': d.address,
-                    'name': d.name,
-                    'rssi': d.rssi_current,
-                    'match': self.target.matches(d),
-                }
-                for d in devices[:8]
-            ]
+            return [
+                {
+                    'id': d.device_id,
+                    'addr': d.address,
+                    'name': d.name,
+                    'rssi': d.rssi_current,
+                    'match': self.target.matches(d, irk_bytes=self._target_irk),
+                }
+                for d in devices[:8]
+            ]
         except Exception:
             return []