fix(reliability): rfplan review fixes — cli commit, assoc poll, ensure_attack wiring, idle reload

payload/user/remote_access/pager-webui/mk8_rfplan.py

@@ -1,9 +1,16 @@
 """Mark VIII RF role manager: radio1/phy1 is shared between an uplink STA
 (``wlan1up``) and attack work, so the roles are made mutually exclusive.
 Uplink pauses channel hopping; attack/idle resumes it."""
+import time
+
 ROLE_KEY = 'mk8.rfplan.role'
 IFACE = 'wlan1up'
 
+# wifi reload returns while wpa_supplicant is still scanning/authenticating;
+# poll instead of checking once or every real uplink would false-fail.
+ASSOC_ATTEMPTS = 5
+ASSOC_WAIT_SECONDS = 2
+
 
 def current_role():
     from server import _uci_values
@@ -22,7 +29,9 @@ def associated():
     for line in (out or '').splitlines():
         line = line.strip()
         if line.startswith('Connected to '):
-            return line.split()[2]
+            parts = line.split()
+            if len(parts) >= 3:
+                return parts[2]
     return None
 
 
@@ -31,6 +40,24 @@ def hop_paused():
     return _read_hop() == '0'
 
 
+def _ensure_cli_network():
+    """Make network 'cli' usable for the STA. Returns True when a network
+    change was staged and still needs ``uci commit network``. Stock firmware
+    ships 'cli' present but disabled; create a minimal DHCP interface when it
+    is missing entirely so netifd can bring wlan1up up either way."""
+    from server import device_run
+    rc, _, _ = device_run(['uci', '-q', 'get', 'network.cli'])
+    if rc != 0:
+        device_run(['uci', 'set', 'network.cli=interface'])
+        device_run(['uci', 'set', 'network.cli.proto=dhcp'])
+        return True
+    rc, out, _ = device_run(['uci', '-q', 'get', 'network.cli.disabled'])
+    if rc == 0 and out.strip() == '1':
+        device_run(['uci', 'set', 'network.cli.disabled=0'])
+        return True
+    return False
+
+
 def set_role(role, ssid=None, psk=None):
     from server import device_run, _pause_hop, _resume_hop
     if role not in ('uplink', 'attack', 'idle'):
@@ -52,17 +79,19 @@ def set_role(role, ssid=None, psk=None):
             cmds.append(['uci', 'set', 'wireless.wlan1up.key=%s' % psk])
         for c in cmds:
             device_run(c)
-        # The STA rides network 'cli'; stock firmware ships it disabled but
-        # present. If it is missing entirely, create a minimal DHCP interface
-        # so netifd can bring wlan1up up.
-        rc, _, _ = device_run(['uci', 'show', 'network.cli'])
-        if rc != 0:
-            device_run(['uci', 'set', 'network.cli=interface'])
-            device_run(['uci', 'set', 'network.cli.proto=dhcp'])
+        if _ensure_cli_network():
+            # netifd consumes committed config only; staging without commit
+            # would leave the STA with no L3 attachment.
+            device_run(['uci', 'commit', 'network'])
         device_run(['uci', 'commit', 'wireless'])
         _pause_hop()
         device_run(['wifi', 'reload'], timeout=60)
-        assoc = associated()
+        assoc = None
+        for _ in range(ASSOC_ATTEMPTS):
+            time.sleep(ASSOC_WAIT_SECONDS)
+            assoc = associated()
+            if assoc:
+                break
         if not assoc:
             disable_uplink()
             _resume_hop()
@@ -71,6 +100,20 @@ def set_role(role, ssid=None, psk=None):
     # attack/idle: tear down the STA so radio1 is free again.
     disable_uplink()
     _resume_hop()
+    if role == 'attack':
+        # The deploy path performs its own wifi reload right after; teardown
+        # converges there without a second reload churn on this phy.
+        try:
+            import mk8_events
+            mk8_events.log_event(
+                'rfplan', msg='rfplan role attack applied; STA teardown '
+                              'applies at next wifi reload')
+        except Exception:
+            pass
+    else:
+        # idle has no guaranteed follow-up reload anywhere else, so converge
+        # now while the gated watchdog is still armed.
+        device_run(['wifi', 'reload'], timeout=60)
     return {'ok': True, 'role': role}
 
 
@@ -81,7 +124,8 @@ def disable_uplink():
 
 
 def ensure_attack():
-    """Exclusivity hook for attack enable paths: switch uplink off first."""
+    """Exclusivity hook for radio1 attack-AP enable paths: switch the
+    uplink off first so one phy never carries STA + AP at once."""
     if current_role() == 'uplink':
         return set_role('attack')
     return None

payload/user/remote_access/pager-webui/server.py

@@ -3535,6 +3535,13 @@ def h_pineap_wifi_set_ap(ctx):
     open_band = channel_band(openap.get('channel')) if openap.get('channel') is not None else None
     use_radio1 = wpa_band in (BAND_5G, BAND_6G) or open_band in (BAND_5G, BAND_6G)
     if use_radio1:
+        # Exclusivity: an uplink STA on radio1 must go down before any
+        # radio1 AP change so one phy never carries STA + AP at once.
+        import mk8_rfplan
+        try:
+            mk8_rfplan.ensure_attack()
+        except Exception:
+            pass
         wpa_active = (wpa_band in (BAND_5G, BAND_6G) and bool(wpa.get('enabled', True))
                       and wpa.get('channel') is not None)
         open_active = (open_band in (BAND_5G, BAND_6G) and bool(openap.get('enabled', True))
@@ -4353,7 +4360,13 @@ def _enterprise_boot_recover():
 
 def h_attacks_deploy(ctx):
     import mk8_gate
+    import mk8_rfplan
     mk8_gate.enter('attack_deploy')
+    # Exclusivity: tear an active uplink down before deploying attack APs.
+    try:
+        mk8_rfplan.ensure_attack()
+    except Exception:
+        pass
     body = ctx.body or {}
     kind = (body.get('kind') or '').strip().lower()
     if kind not in ('wpa', 'open', 'enterprise'):