mirror of
https://github.com/home-assistant/supervisor.git
synced 2026-08-21 13:49:01 +01:00
dc6a77507b
* fix(docker): register hw listener and match by-id paths for options-based devices Two bugs caused a crash loop when a USB device re-enumerates to a different minor number (e.g. ttyACM0→ttyACM1) after a HAOS reboot: 1. _hw_listener was only registered when addon.static_devices was non-empty. Addons that expose a device via the options schema (e.g. Z-Wave JS `device:` option) never had the listener registered, so add_devices_allowed was never called when the device reappeared at a new minor. 2. _hardware_events matched only device.path and device.sysfs against static_devices. When static_devices (or the new options path) contains a by-id symlink, the match always failed because by-id paths live in device.links. Fix: extend the listener registration condition to also cover addon.devices (options-based), and expand the path-matching set to include device.links so by-id paths resolve correctly. For options-based devices, compare the incoming Device against addon.devices (which re-evaluates options.json against the live hardware list, picking up the new minor number automatically). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * fix(docker): use option_device_paths for cheap by-id hw event matching Refactor _hardware_events to avoid per-event full options validation (including pwnd hashing). Introduce AppOptions.extract_device_paths and AppModel.option_device_paths to extract raw device paths from options without resolving against live hardware. Use set-intersection against {device.path, device.sysfs, *device.links} so by-id symlinks match correctly after re-enumeration for both static and options-based devices. Update test to use real schema/options setup and simulate a minor-number change (ttyACM0→ttyACM1) with a stable by-id symlink. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com> * test: improve hw listener test coverage and add policy check Address PR review feedback: - Add hardware policy check in _hardware_events to prevent bypassing access restrictions on hotplug events (follows same pattern as startup cgroup setup) - Fix test_app_options_device_hw_listener to properly simulate USB re-enumeration with different minor numbers (166:0 → 166:1) - Add test_app_options_device_policy_check to verify policy enforcement for options-based devices - Update TEST_HW_DEVICE with realistic major/minor attributes (166:0 for tty) Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: mock HostFeature.OS_AGENT in hardware event tests The _hardware_events method has a @Job decorator with conditions=[JobCondition.OS_AGENT], which checks if HostFeature.OS_AGENT is in sys_host.features. Without mocking this, the job conditions fail and the hardware event handler is never invoked, causing add_devices_allowed to not be called and tests to fail. Add patch.object(type(coresys.host), "features", ...) to all four hardware event tests to ensure the OS_AGENT job condition is met. Fixes test failures: - test_app_new_device (all 6 parametrized cases) - test_app_new_device_no_haos - test_app_options_device_hw_listener - test_app_options_device_policy_check Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: fix TEST_DEV_PATH to match TEST_HW_DEVICE.path TEST_DEV_PATH was set to /dev/ttyAMA0 but TEST_HW_DEVICE.path is /dev/ttyACM0. This mismatch would cause the dev_path=TEST_DEV_PATH parametrized test cases to fail because the hardware event handler checks if the device path intersects with the app's allowed devices, and "/dev/ttyAMA0" != "/dev/ttyACM0". Update TEST_DEV_PATH from /dev/ttyAMA0 to /dev/ttyACM0 to match TEST_HW_DEVICE. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: mock option_device_paths in test_app_options_device_hw_listener The test sets up schema and options but option_device_paths property may not be working as expected in the test environment. Add an explicit mock for option_device_paths to ensure it returns the by-id path, guaranteeing that: 1. The hardware listener is registered (checks option_device_paths at registration) 2. The device path matching works correctly in _hardware_events This ensures the test properly validates that hardware events are processed for options-based devices after re-enumeration. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com> * test: make policy check test actually exercise the policy guard test_app_options_device_policy_check set the device option via persist["options"], but option_device_paths reads the merged options and does not pick that override up during the test, so it returned an empty set. The hardware event therefore failed the path-match guard and returned before reaching the allowed_for_access check. The assert_not_called() assertion then passed regardless of the policy outcome -- it would still pass if the policy guard were removed entirely. Mock option_device_paths to return the configured by-id path (mirroring test_app_options_device_hw_listener) so the event device matches and execution actually reaches the policy guard the test is meant to verify. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * test: add unit test for AppOptions.extract_device_paths The integration tests exercise extract_device_paths only through a mocked option_device_paths property, so the schema-walking logic introduced for the hardware-event matching had no direct coverage. Add a unit test that drives every schema shape the recursion handles -- flat, optional, filtered, list, nested dict and list of dicts -- and asserts that non-device options, unset keys and empty values are skipped, without requiring the devices to exist in hardware. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com> Co-authored-by: Stefan Agner <stefan@agner.ch>