Togis Thomas

This commit **adjusts SAI port serdes hardware tests** to **skip POST3 readback verification** specifically for **Tomahawk6 (TH6)** ASICs. This **bug fix** prevents false test failures caused by the TH6 SDK returning a sentinel value (32767) instead of actual POST3 serdes parameters during readback. The change ensures that `verifyTxSettting` in `SaiPortUtils.cpp` correctly passes on TH6/tahansb platforms by only skipping the readback check, while the POST3 write/configuration remains fully active. This prevents erroneous test failures due to an SDK limitation, allowing the tests to accurately reflect the system's state.

This commit introduces a **robustness improvement** within the **QSFP service's CMIS module** by adding a crucial guardrail. It modifies the `fillVdmPerfMonitorCoherentVdm` function in `CmisModule.cpp` to ensure its logic is exclusively applied to **tunable optics** or coherent modules. This **fix** prevents the function from being incorrectly invoked or processing data for non-coherent transceivers. The change enhances the **accuracy and reliability** of VDM performance monitoring by correctly scoping data collection to applicable hardware.

This commit **fixes a bug** in the **CMIS module's** `getChannelNumFromFrequency` function, which led to **incorrect optical module programming**. Previously, floating-point precision errors in the conversion from MHz to THz could cause an off-by-one channel assignment for frequencies on a 6.25 GHz grid. The fix replaces the imprecise floating-point arithmetic with **exact integer arithmetic** using `int64_t`, eliminating `IEEE 754` truncation issues. This ensures **accurate frequency-to-channel conversion** for optical transceivers, preventing incorrect wavelength configuration and guaranteeing optics are programmed to their intended wavelengths.

This commit **enhances the `wedge_qsfp_util` CLI tool** by implementing display support for **coherent 800G ZR VDM (Vendor Diagnostic Monitoring) statistics**. Specifically, the **`vdm-info` command** now presents detailed **Coherent VDM Parameters**, **FEC Performance Monitoring (Page 34h)**, and **Link Performance Monitoring (Page 35h)**, covering metrics such as modulator bias, CD, DGD, OSNR, Tx/Rx power, and Q-factor. This **new feature** also integrates VDM config type entries (128-149) for direct I2C `vdm-info` display. The change provides **critical visibility into 800G ZR coherent optics performance**, significantly aiding in monitoring and troubleshooting these advanced transceivers.

This commit introduces a **new capability** to **enhance performance monitoring** for **coherent 800G ZR modules** within the **`qsfp_service` module**. It implements the extraction and aggregation of **Link Performance Monitoring (PM) statistics** by combining historical average/min/max values from **C-CMIS Page 35h** with real-time current values from various **VDM pages**. This work involves adding new CMIS page definitions, VDM configuration types, and the `fillVdmPerfMonitorLinkPm()` function to process and integrate this comprehensive data. The change significantly improves the **monitoring granularity and completeness** for **800G ZR coherent optics**, providing critical insights into optical link health.

This commit introduces a **new capability** to extract **FEC Performance Monitoring (PM) statistics** from C-CMIS Page 34h, specifically for **800G ZR coherent optical modules**. It populates per-lane counters such as received bits, corrected bits, received frames, and uncorrectable error frames, which are crucial for diagnosing link health and are integrated into the **VDM statistics collection**. The **CmisModule** is updated with new enums (`PAGE34`, `PAGE_UPPER34H`), a dedicated cache, and a new `fillVdmPerfMonitorFecPm()` function to parse this data, with **test coverage** added to verify correct decoding.

This commit introduces a **new capability** to the **CMIS module** by implementing the extraction and population of **coherent VDM parameters** specifically for **800G ZR transceivers**. It adds support for reading Modulator Bias (XI/XQ/YI/YQ/XPhase/YPhase), Chromatic Dispersion (CD), and State of Polarization Dependent Loss (SOPMD) values from pages 20h-23h, adhering to the OIF C-CMIS-01.3 standard. This enhancement provides critical **performance monitoring data** for advanced optical modules, enabling more granular diagnostics and operational insights by storing these new statistics under `coherentVdmStats` in `VdmPerfMonitorPortSideStats`.

This commit **adds comprehensive unit tests** for **Vendor Diagnostic Monitoring (VDM)** functionality on **800G ZR optical modules**. A new test case, `cmis800GZrVdmTest`, is introduced in `CmisTest.cpp` to verify VDM group support, correct parsing of coherent VDM config descriptors, and accurate data value retrieval from specific pages for all 8 coherent VDM types. To enable these tests, `FakeTransceiverImpl.cpp` is updated with **real register data** from an actual 800G ZR module for VDM configuration and data pages (20h-27h). This work significantly enhances the **CMIS implementation's test coverage** within the `qsfp_service` and ensures the reliability of VDM data reporting for these high-speed transceivers.

This commit **introduces support for VDM Group 4** within the **CMIS module** by enabling the reading and processing of VDM configuration data from page 0x23h and VDM data from page 0x27h. It updates **Thrift definitions** in `transceiver.thrift` and `cmis.thrift`, extends `CmisPages` and `CmisField` enums, and modifies internal data structures and accessors in `CmisModule.h` and `CmisModule.cpp` to accommodate these new pages. Additionally, the VDM page reading logic is **refactored** from hard-coded conditional checks to a more extensible loop-based iteration driven by `vdmSupportedGroupsMax_` and constexpr arrays. This **new capability** enhances the `qsfp_service`'s ability to correctly interpret CMIS VDM data for transceivers utilizing Group 4, such as 800G ZR, while improving the maintainability and scalability of the VDM page parsing mechanism.

This commit **introduces new Thrift struct definitions** to enable comprehensive **monitoring of 800G ZR coherent optics**. It adds `CoherentVdmStats`, `FecPm`, and `LinkPm` structs within the `fboss/qsfp_service/if/transceiver.thrift` module. This **new capability** extends the existing `VdmPerfMonitorPortSideStats` with VDM-unique coherent parameters, such as modulator bias and CD/SOPMD low granularity. The new `FecPm` and `LinkPm` structs integrate FEC and Link Performance Monitoring data, adhering to the OIF C-CMIS-01.3 specification for advanced transceiver diagnostics.

This commit **enhances I2C transaction logging** by incorporating the `page` parameter into `TransceiverAccessParameter` structs. This **feature enhancement** addresses a previous limitation where I2C transaction logs for **CMIS firmware upgrades** lacked crucial page information, impeding debugging efforts. The `page` parameter is now propagated through `i2cWriteAndContinue()` and consistently applied to `readTransceiver` and `writeTransceiver` calls across the **`CdbCommandBlock`** and **`FirmwareUpgrader`** modules. This change ensures that all CDB register, page select, and EPL payload writes are logged with their correct page context, significantly **improving the debuggability** of complex I2C operations.

This commit introduces a **new capability** to the **`wedge_qsfp_util` CLI**, enabling it to retrieve and display the **firmware build number for CMIS modules**. Previously, the utility only showed major and minor firmware versions, but this enhancement now fetches the vendor-specific build number using the `CMD 0100h: Get Firmware Info` CDB command. The change primarily affects the `printCmisDetailService` function within `fboss/util/wedge_qsfp_util.cpp`. This provides more comprehensive firmware details to users of the utility, improving the diagnostic information available for **CMIS transceivers**.

This commit introduces a **new capability** to retrieve the **firmware build number** for **CMIS modules** within the `qsfp_service`. It implements a new **CDB command** (`CMD 0100h: Get Firmware Info`) via `createCdbCmdGetFirmwareInfo` to fetch this additional firmware version detail, which vendors use alongside major and minor numbers. The **CMIS module** now includes logic to `fetchFwBuildNumberFromCdb`, cache it, and integrate it into existing status retrieval and update processes like `getFwStatus` and `updateQsfpData`. This enhancement provides more granular **firmware identification** for transceivers, improving diagnostics and inventory management.

This commit introduces **support for retrieving and representing the firmware build number** for **CMIS modules**. It **enhances the `FirmwareStatus` Thrift structure** in `transceiver.thrift` by adding an optional `buildNumber` field, allowing the system to capture this vendor-specific detail alongside major and minor versions. Consequently, the generated FSDB model in `fboss/fsdb/if/oss/fsdb_model_thriftpath.h` is updated to reflect this new field. This **new capability** enables **more precise firmware version tracking and diagnostics** for transceivers, improving overall device management.

This commit **updates the hardware test suite** within the **QSFP service** to correctly verify transceiver settings for **ZR modules**. It modifies the `verifyOpticsSettings` function in `fboss/qsfp_service/test/hw_test/HwTransceiverUtils.cpp` to assert that `txSquelch` and `rxSquelch` disable flags are `true` for tunable optics, aligning with their meta-spec requirements. This **maintenance change** ensures that **ZR module testing** accurately reflects their expected behavior, improving the **reliability of hardware validation** for these specialized transceivers.

This commit provides a **bug fix** for a **hitless warm boot** unit test issue affecting **tunable (ZR) optics** within the **CMIS module**. It modifies the `AppSel` code matching condition to check `progStartTimer` for `inittimer`, preventing unnecessary application programming during warm boot when `dpinitdone` is false. Additionally, the `readyState` state machine is adjusted for tunable optics to align with IM-DD optics, specifically to avoid invoking `datapathprogam`. This prevents `dp_deinit` during warm boot, which would otherwise lead to **traffic impact** for these critical optical modules. The changes ensure a more robust and truly hitless warm boot experience by preventing unintended de-initialization.

This commit introduces a **test fix** within the **`AgentEnsembleOpticsTest`** suite to prevent erroneous failures. Specifically, it **excludes 800G ZR optics** from `RX_LOL` (Receiver Loss of Lock) signal checks in the `verifyTxRxLatches` test. This **maintenance change** is necessary because `RX_LOL` is not reliably raised for 800G ZR modules due to disabled squelching, which would otherwise cause false positives in the **link test framework**. The modification ensures the **link tests** accurately reflect the operational status of **800G ZR optics** without reporting expected non-events as errors.

This commit implements a crucial **safety validation** within the **`qsfp_service`** to prevent **tunable optics (ZR) modules** from entering high power mode without proper optical channel configuration. The **`TransceiverManager`** now explicitly checks for this configuration and passes the status to the module's readiness checks. Specifically, `CmisModule::ensureTransceiverReadyLocked()` will now throw a `FbossError` if a **ZR module** lacks the required configuration, acting as a **bug fix** for an unsafe operational scenario. This enhancement ensures **system stability** and prevents potential hardware damage or unexpected behavior when initializing **tunable optics**.

This commit **enhances the QSFP service's CMIS module handling** by **removing the hardcoded `kUsecDatapathStateUpdateTimeMaxFboss` constant**. It introduces a **new capability** to **dynamically determine the maximum datapath delay** for **AEC and ZR optics modules** by retrieving this information directly from module specifications. Specifically, a new helper function `getDatapathMaxDelayFromModuleSpec` is implemented and integrated into `getExpectedDatapathDelayUsec`, which in turn refines the retry calculation within `dataPathProgram`. This **feature enhancement** ensures that datapath state machine polling for **tunable optics and AEC modules** is more accurately aligned with their advertised capabilities, improving the robustness of module management.

This commit introduces an **optimization** to the **CmisModule**'s transceiver programming logic, specifically within the `programTransceiver` event and `resetDataPath` function. It addresses a scenario where the state machine might get stuck in `DP_ACTIVATION` by preventing redundant laser frequency and AppSel re-programming. By adding checks in `customizeTransceiverLocked` and `resetDataPathForTunableOptics`, this change ensures that these critical programming steps for **tunable optics** occur only once, improving the efficiency and robustness of transceiver initialization and datapath resets.