BMC/OpenBmc/bmcweb.git, branch master

redfish: Refactor FabricAdapter lambda

2026-10-09T10:43:51+00:00

getFabricAdapterLocation() registered its D-Bus LocationCode property response handler as an inline lambda, requiring an ast-grep-ignore suppression for the long-lambda rule. This commit extracts the body into a named function (afterGetFabricAdapterLocation) wired via std::bind_front into the safe dbus::utility::getProperty API. This is a pure refactor: logic is unchanged. The ast-grep-ignore comment is removed since the code is no longer flagged by the rule. Tested: - Booted an AST2600 SoC platform and exercised the endpoint that actually invokes the refactored code live: GET /redfish/v1/Systems/system/FabricAdapters// (single-adapter endpoint, not the collection) returned 200 OK with Location/PartLocation/ServiceLabel populated from the adapter's LocationCode D-Bus property, matching the value set on the test object. No errors logged for fabric_adapters.hpp for the full test session. - Redfish Service Validator: 5923 Pass, 353 Warn, 0 Fail. Change-Id: I1f6594f690b8131ac17d6d38502e5674f15e7cb6 Signed-off-by: Yuvakumar Selvamani

Refactor chassis_utils getValidChassisPath lambda

2026-10-08T16:34:36+00:00

The getValidChassisPath function contained an inline lambda that triggered the long-lambda ast-grep rule, requiring an inline suppression (// ast-grep-ignore: long-lambda). Since callback is a forwarding- reference template parameter, std::bind_front cannot be used directly here. This commit extracts the lambda body into a separate templated afterGetValidChassisPath function, leaving a short delegating lambda in getValidChassisPath that simply forwards to it. This follows the alternative pattern documented for forwarding-reference callbacks, improves readability, and allows the removal of the static analysis suppression. Tested on AST2600 SoC (live D-Bus tested). - `HEAD /redfish/v1/Chassis//ThermalSubsystem` (exercises getValidChassisPath()) Expected: HTTP 404 Base.1.19.ResourceNotFound Actual: HTTP 404 Base.1.19.ResourceNotFound - `HEAD /redfish/v1/Chassis//ThermalSubsystem` (a Chassis object registered live on D-Bus for this test) Expected: HTTP 200 (chassis resolves) Actual: HTTP 200 with the correct Link header - confirms the templated afterGetValidChassisPath() match-found branch - RSV: 5845 Pass / 353 Warn / 0 Fail Change-Id: I76ed9cc42fa4e49e49e6dc7c3cfc85b75f0d2741 Signed-off-by: Yuvakumar Selvamani

OWNERS: add myself as a reviewer

2026-10-08T16:09:27+00:00

Add myself to the bmcweb OWNERS file as a reviewer. Contributions: https://github.com/openbmc/bmcweb/commits?author=vinothkumarshanmugavel Change-Id: I50d5e704bbe3a3ea2d55f058c52dae8798e443a1 Signed-off-by: Vinothkumar Shanmugavel

redfish/processor: add tests for helpers

2026-10-08T12:03:27+00:00

Add unit coverage for processor helper paths that were previously untested. These tests focus on the logic that maps D-Bus data into the Redfish processor response and handles common error cases. These helpers directly populate processor response fields for throttle information and manufacturer-based architecture mapping. Adding unit tests here improves coverage for those paths and their error handling. Tested: - Built and verified, all tests passed. Change-Id: I78a588d43fe4459c1f6197dcc58646632490c041 Signed-off-by: Joel Pullokaran Jesin

redfish: Refactor EventService GET handler lambda

2026-10-08T01:10:18+00:00

requestRoutesEventService() registered its GET route as a long inline lambda building the static EventService resource JSON, requiring an ast-grep-ignore suppression for the long-lambda rule. This commit extracts the body into a named free function (handleEventServiceGet) registered via std::bind_front(handleEventServiceGet, std::ref(app)), matching the style already used for the POST subscriptions route in this same file (handleEventServiceSubscriptionsPost). This is a pure refactor: logic is unchanged. The ast-grep-ignore comment is removed since the code is no longer flagged by the rule. Tested: - Booted an AST2600 SoC platform with the new bmcweb and exercised the endpoint live: GET /redfish/v1/EventService/ returned 200 OK with the expected EventService JSON body (Id, Name, ServerSentEventUri, DeliveryRetryAttempts/IntervalSeconds, EventFormatTypes, RegistryPrefixes, ResourceTypes, SSEFilterPropertiesSupported, etc.) identical to the pre-refactor lambda's output. bmcweb.service was active and journalctl -u bmcweb showed only the normal startup line, no errors. - Redfish Service Validator: 5829 Pass, 353 Warn, 0 Fail. Change-Id: Iea6bb1c6366a9ead26cb5edb921f7eb5deec976c Signed-off-by: Yuvakumar Selvamani

Add OpenSSL 4.0 support

2026-10-07T15:44:16+00:00

With OpenSSL 4.0, ASN1_STRING and several other types are now opaque [0] and cannot be reached like before, and X509_NAME is now constified [1], causing the build to fail. Given that Yocto has now migrated to OpenSSL 4.0, the code must be compatible with both 3.x and 4.0. Modify the accesses for ASN1_STRING to use the appropriate functions like ASN1_STRING_get0_data() and construct the type with all options directly, instead of assigning them afterwards. X509_NAME is now its own RAII object as it must be duped due to the constification. Methods like X509_NAME_add_entry_by_txt however require a non-constified pointer. The truncation to 255 bytes for the X509 common name is now gone since get_index_by_NID and get0_data made the fixed-size buffer unnecessary. Setting subject entries such as country, organization or the subject name now returns a boolean and certificate generation aborts on failure instead of silently continuing with an incomplete subject. The wrap file for OpenSSL can't be updated due to a missing dependency (wrapdb) that doesn't provide OpenSSL 4-compatible patches. Tested: Code compiles with both OpenSSL 3.6.3, as well as 4.0.2, and the unit tests pass. On an AST2600 platform with OpenSSL 3.6.3, as well as 4.0.2, the cert generation works like before, and mTLS authentication as described in [2] how to set it up works too. Adding the certificate to the trusted store works like before. References: [0] https://github.com/openssl/openssl/blob/5ce57ab932b70e030249190a4b5a02232c080eea/CHANGES.md?plain=1#L1523 [1] https://github.com/openssl/openssl/commit/b0f2107b44049805b39434d104c8bb0daf8ae079 [2] https://github.com/openbmc/docs/blob/93b1e07c59cd8bc11e7afaaba4172b21a0a5b552/security/TLS-configuration.md Change-Id: Iefe7d87ddda3dea1ca46929110dea06cff9079ad Signed-off-by: Tan Siewert

Redfish: Return correct EventDestination errors

2026-10-07T05:20:40+00:00

Return a Redfish ResourceNotFound message when an EventDestination subscription cannot be found. Use EventDestination as the resource type in both the regular subscription lookup and the SNMP trap client lookup. This makes the error payload accurately describe the Redfish resource requested. GET /redfish/v1/EventService/Subscriptions/{SubscriptionId}/ now returns HTTP 404 with `Base.1.19.ResourceNotFound` and message arguments containing `EventDestination` and the requested ID. Tested: GET /redfish/v1/EventService/Subscriptions/{SubscriptionId}/ with a missing subscription ID returns HTTP 404 and `Base.1.19.ResourceNotFound`. DELETE /redfish/v1/EventService/Subscriptions/{SubscriptionId}/ now contains `EventDestination`. Redfish Service Validator passed. Change-Id: I17f19fef887b4cea736257a9bc1e1d7424eb24f3 Signed-off-by: Joel Pullokaran Jesin

sensor: add ReadingTime to sensor responses

2026-10-07T04:51:22+00:00

Add ReadingTime so a client can tell how old a reading is instead of assuming it is current, which is what a consumer closing a control loop on power and thermal readings needs in order to notice a device that has stopped responding. The value comes from the UpdatedTime property defined in: https://gerrit.openbmc.org/c/openbmc/phosphor-dbus-interfaces/+/84171 A service that has not obtained a reading leaves the property zero, which is skipped rather than published as a bogus date. ReadingTime is already in the installed Sensor schema and is not part of any sensor excerpt, so only Redfish.md needed updating. Tested: ``` curl -k -H "X-Auth-Token: $token" \ https:///redfish/v1/Chassis/Nvidia_*/Sensors/temperature_GPU_*_TEMP_0 { "@odata.id": "/redfish/v1/Chassis/Nvidia_*/Sensors/temperature_...", "@odata.type": "#Sensor.v1_11_1.Sensor", "Id": "temperature_...", "Name": "", "PhysicalContext": "Accelerator", "Reading": 26.19921875, "ReadingRangeMax": 127.0, "ReadingRangeMin": -128.0, "ReadingTime": "2026-08-17T15:54:30.392391+00:00", "ReadingType": "Temperature", "ReadingUnits": "Cel", "Status": { "Health": "OK", "State": "Enabled" } } Redfish Service Validator: Summary - PASS: 9453, WARN: 361, FAIL: 0, NOT TESTED: 12407 Validating /redfish/v1/Chassis/Nvidia_*/Sensors/temperature_... - Pass: 14, Warn: 0, Fail: 0, Skip: 54 ``` 1. ReadingTime carried the time the reading was acquired from the sensor, matching the clock to microsecond precision 2. All 58 sensor resources validated with no new failures 3. A service that does not publish UpdatedTime omits ReadingTime rather than reporting the Epoch Change-Id: Ib97dff91b19a7da4b3b99384dc9940333a92c3e1 Signed-off-by: Ender Hsieh

account-service: return proper error on deletion

2026-10-06T21:22:45+00:00

phosphor-user-manager may return an error on deletion if userdel fails for arbitrary reasons (e.g. the user is still logged in and running processes). Previously, any D-Bus error was unconditionally mapped to 404 Not Found, which was misleading for users that existed but could not be removed. Inspect the D-Bus error name and return the most appropriate Redfish response: - ResourceNotFound for unknown objects - ResourceCannotBeDeleted for NotAllowed (e.g. removing the root user) - InternalError for anything else. Tested: Verified correct error responses when deleting a non-existent user, when attempting to delete root (prohibited by phosphor-user-manager), and on unknown errors. Ran all openbmc-test-automation AccountService tests with no unexpected regressions or, more precisely, it fixed non-included tests like the root user deletion test. Change-Id: I559a4081d992b424319ae243c176baae91b1c322 Signed-off-by: Tan Siewert

Test translateMemoryTypeToRedfish conversions

2026-10-06T16:43:17+00:00

Add test/redfish-core/lib/memory_test.cpp and register it in test/meson.build. Cover translation of supported D-Bus DIMM device types to MemoryDeviceType enum values. Verify each supported device type maps to its corresponding Redfish enum. Also verify unknown D-Bus values return MemoryDeviceType::Invalid. Tested: - meson test -C builddir memory_test --print-errorlogs - memory_test passed Change-Id: Iff4f05a3005e9f202e3c9c27c1b308b6490c3afa Signed-off-by: Vinothkumar Shanmugavel