Stop checkra1n when palera1n is abandoned

src/orchard_cli/apple/boot.py

@@ -1,4 +1,6 @@
 """Shared checkm8/Pongo boot sequence; adapters supply device identity constraints."""
+import os
+import signal
 import subprocess
 import shutil
 import time
@@ -45,6 +47,23 @@ def wait_mode(device, mode, *, chip, board, product, seconds=30):
         time.sleep(0.25)
 
 
+def run_process_group(argv, *, timeout, **kwargs):
+    """subprocess.run(check=True) that also stops the command's children.
+
+    palera1n posix_spawns checkra1n and only waits for it. Killing palera1n
+    alone leaves checkra1n running, and it exploits the next DFU device it
+    sees, so give up on the whole process group instead.
+    """
+    with subprocess.Popen(argv, start_new_session=True, **kwargs) as process:
+        try:
+            returncode = process.wait(timeout=timeout)
+        except BaseException:
+            os.killpg(process.pid, signal.SIGKILL)
+            raise
+    if returncode:
+        raise subprocess.CalledProcessError(returncode, argv)
+
+
 def boot(device, assets, manifest, progress, *, profile, chip, board, product):
     if device['mode'] == 'linux':
         progress.phase('Using running recovery; no reboot or upload (timing starts at resume)')
@@ -61,8 +80,8 @@ def boot(device, assets, manifest, progress, *, profile, chip, board, product):
         with tempfile.NamedTemporaryFile(prefix='palera1n-', suffix='.log', dir=logs,
                                          delete=False) as log:
             try:
-                subprocess.run([executable, '-p', '-k', str(assets['pongo'])],
-                               stdout=log, stderr=subprocess.STDOUT, check=True, timeout=120)
+                run_process_group([executable, '-p', '-k', str(assets['pongo'])],
+                                  stdout=log, stderr=subprocess.STDOUT, timeout=120)
             except (OSError, subprocess.SubprocessError) as error:
                 raise RuntimeError('Pongo bootstrap failed or outcome unknown; inspect USB before any retry. '
                                    f'Log: {log.name}') from error

tests/test_runtime.py

@@ -8,6 +8,7 @@
 import subprocess
 import sys
 import tempfile
+import time
 import unittest
 from unittest.mock import Mock, patch
 import urllib.request
@@ -46,7 +47,7 @@ def test_adapter_boot_sequence_timeouts_and_no_retry_are_preserved(self):
              patch.object(boot.shutil, 'which', return_value='/tools/palera1n'), \
              patch.object(adapter_module.usb, 'selected') as selected, \
              patch.object(boot, 'wait_mode') as wait, \
-             patch.object(boot.subprocess, 'run') as bootstrap, \
+             patch.object(boot, 'run_process_group') as bootstrap, \
              patch.object(boot.pongo, 'upload') as upload:
             adapter.boot(device, assets, record, Mock())
             self.assertEqual(bootstrap.call_count, 1)
@@ -62,6 +63,33 @@ def test_adapter_boot_sequence_timeouts_and_no_retry_are_preserved(self):
             self.assertEqual(bootstrap.call_count, 2) # one success, one failed call; no retry.
             upload.assert_not_called()
 
+    def test_bootstrap_timeout_stops_spawned_children(self):
+        # palera1n posix_spawns checkra1n; a surviving checkra1n would exploit
+        # the next DFU device behind a later command's back.
+        from orchard_cli.apple import boot
+        pidfile = self.root / 'child.pid'
+        script = f'sleep 300 & echo $! > {pidfile}; wait'
+        with self.assertRaises(subprocess.TimeoutExpired):
+            boot.run_process_group(['sh', '-c', script], timeout=1)
+        child = int(pidfile.read_text())
+        for _ in range(50):
+            try:
+                state = Path(f'/proc/{child}/stat').read_text().rsplit(')', 1)[1].split()[0]
+            except (FileNotFoundError, ProcessLookupError):  # gone mid-read
+                break
+            if state == 'Z':
+                break
+            time.sleep(0.05)
+        else:
+            os.kill(child, 9)
+            self.fail('bootstrap child outlived the timeout')
+
+    def test_bootstrap_failure_is_reported(self):
+        from orchard_cli.apple import boot
+        with self.assertRaises(subprocess.CalledProcessError):
+            boot.run_process_group(['sh', '-c', 'exit 3'], timeout=5)
+        boot.run_process_group(['true'], timeout=5)
+
     def test_manifest_becomes_existing_workflow_inputs(self):
         record, assets = releases.resolve(FIXTURE / 'index.toml', 'j71b')
         self.assertEqual(record['release'], 'example-1')