Use subsystem recovery for Sargo modem faults

Select RELATED before HAL startup and create oFono’s private state directory. A controlled modem restart recovered network registration without rebooting the phone; two persistent boots and native UI checks passed. The IMS firmware watchdog root cause remains unresolved.

tundra/runtime/modem-recovery.py

@@ -0,0 +1,35 @@
+#!/usr/bin/env python3
+# SPDX-License-Identifier: MPL-2.0
+"""Keep a Sargo modem fault from immediately restarting the whole phone."""
+from pathlib import Path
+import time
+
+
+def configure(devices=Path('/sys/bus/msm_subsys/devices')):
+    matches = [p.parent for p in devices.glob('*/name')
+               if p.read_text().strip() == 'modem']
+    if len(matches) != 1:
+        raise RuntimeError('Expected one Sargo modem subsystem')
+    level = matches[0]/'restart_level'
+    previous = level.read_text().strip()
+    if previous not in ('SYSTEM', 'RELATED'):
+        raise RuntimeError('Unsupported modem restart policy: '+previous)
+    # This kernel supports coupled subsystem recovery. Firmware failures and
+    # unsuccessful recovery remain visible in the kernel log; do not suppress
+    # its watchdog or crash accounting.
+    if previous != 'RELATED':
+        level.write_text('RELATED\n')
+    if level.read_text().strip() != 'RELATED':
+        raise RuntimeError('Modem restart policy was not applied')
+    print('Sargo modem recovery: '+previous+' -> RELATED', flush=True)
+
+
+if __name__ == '__main__':
+    for attempt in range(30):
+        try:
+            configure()
+            break
+        except (FileNotFoundError, RuntimeError):
+            if attempt == 29:
+                raise
+            time.sleep(1)

tundra/runtime/prepare-hardware.py

@@ -2,10 +2,16 @@
 # SPDX-License-Identifier: MPL-2.0
 """Configure isolated oFono/PulseAudio diagnostics; no call or SMS on startup."""
 from pathlib import Path
+import shutil
 
 
 def prepare(root, write):
     write('etc/tundra/hardware-enabled', 'sargo\n')
+    shutil.copy2(Path(__file__).with_name('modem-recovery.py'),
+                 root/'etc/tundra/modem-recovery.py')
+    write('etc/systemd/system/tundra-hal.service.d/modem-recovery.conf', '''[Service]
+ExecStartPre=/usr/bin/python3 /etc/tundra/modem-recovery.py
+''')
     with (root/'etc/group').open('a') as groups:
         groups.write('audio:x:1005:droidian\ncamera:x:1006:droidian\nradio:x:1001:\nmedia:x:1013:droidian\n')
     write('etc/ofono/binder.conf', '[Settings]\nExpectSlots = slot1\n\n[slot1]\npath = /ril_0\nslot = 0\n')
@@ -26,6 +32,8 @@ After=tundra-hal.service dbus.service
 [Service]
 Type=dbus
 BusName=org.ofono
+StateDirectory=ofono
+StateDirectoryMode=0700
 ExecStartPre=/usr/bin/binder-wait android.hardware.radio@1.0::IRadio/slot1
 ExecStart=/usr/sbin/ofonod --nodetach
 Restart=on-failure

tundra/tests/test_modem_recovery.py

@@ -0,0 +1,42 @@
+# SPDX-License-Identifier: MPL-2.0
+"""Check modem selection without modifying a host's subsystem policy."""
+import importlib.util
+from pathlib import Path
+import tempfile
+import unittest
+
+spec = importlib.util.spec_from_file_location(
+    'modem_recovery', Path(__file__).resolve().parents[1]/'runtime/modem-recovery.py')
+recovery = importlib.util.module_from_spec(spec)
+spec.loader.exec_module(recovery)
+
+
+class ModemRecoveryTest(unittest.TestCase):
+    def test_only_modem_policy_changes_and_repeated_setup_is_safe(self):
+        with tempfile.TemporaryDirectory() as folder:
+            root = Path(folder)
+            for identifier, name in [('subsys1', 'adsp'), ('subsys7', 'modem')]:
+                device = root/identifier
+                device.mkdir()
+                (device/'name').write_text(name+'\n')
+                (device/'restart_level').write_text('SYSTEM\n')
+            recovery.configure(root)
+            recovery.configure(root)
+            self.assertEqual((root/'subsys1/restart_level').read_text(), 'SYSTEM\n')
+            self.assertEqual((root/'subsys7/restart_level').read_text(), 'RELATED\n')
+
+    def test_missing_modem_is_reported(self):
+        with tempfile.TemporaryDirectory() as folder:
+            with self.assertRaisesRegex(RuntimeError, 'Expected one'):
+                recovery.configure(Path(folder))
+
+    def test_unknown_policy_is_preserved(self):
+        with tempfile.TemporaryDirectory() as folder:
+            root = Path(folder)
+            device = root/'subsys0'
+            device.mkdir()
+            (device/'name').write_text('modem\n')
+            (device/'restart_level').write_text('UNKNOWN\n')
+            with self.assertRaisesRegex(RuntimeError, 'Unsupported'):
+                recovery.configure(root)
+            self.assertEqual((device/'restart_level').read_text(), 'UNKNOWN\n')