Protick Bhowmick

This commit **enhances** the **Credo SDK versioning mechanism** by introducing explicit support for **Credo SAI SDK version 1.2.8**. It defines the `CREDO_SDK_VERSION_1_2_8` macro with its corresponding integer value (10208) within `fboss/lib/phy/CredoSdkVersion.h`. The conditional compilation logic in this header is also updated to correctly identify and set the active `CREDO_SDK_VERSION` when `CREDO_SDK_1_2_8` is defined. This **maintenance update** ensures that the FBOSS project can properly integrate and build against the specified Credo SDK, enabling the use of features or fixes present in that specific SDK release.

This commit **updates the server port** for the **Link Explorer application** from `443` to `8080` within `fboss/util/link_explorer/server.py`. This **configuration change** is essential for enabling **corp-to-prod access** to the Link Explorer web UI via **Silfra**, Meta's internal web proxy. By switching to port `8080`, which is within Silfra's default allowed range, the web UI becomes **directly accessible** from browsers without requiring additional VIP or traffic infrastructure setup. This significantly improves the **accessibility and usability** of the Link Explorer tool for internal users.

This commit introduces **Link Explorer**, a new internal web service designed to provide **live visualization of FBOSS switch topology** and **link diagnostics**. Implemented as a **standalone Flask web server** with a single-page GUI, it directly queries `FbossCtrl` and `QsfpService` via Thrift to fetch **LLDP topology**, **port statistics**, and **transceiver diagnostics**. This **new capability** resides in `fboss/util/link_explorer` and offers a comprehensive view of switch health and connectivity, providing a dedicated tool for network engineers with **zero risk to existing production FBOSS services**.

This commit **refactors** the way Credo SDK versions are handled within FBOSS by introducing `CredoSdkVersion.h`. This new header defines numeric macros for SDK versions, replacing scattered `ifdef` checks with ordered comparisons like `#if CREDO_SDK_VERSION >= CREDO_SDK_VERSION_0_9_0`. This **maintenance** improvement affects the **Credo SDK integration** across `SaiSwitch`, `SaiElbert8DDPhyPlatform`, and `SaiPhyManager`, modifying preprocessor directives in functions like `gracefulExitLocked` and `listObjectsLocked`. The change makes SDK version checks more robust and **future-proof**, centralizing version logic and simplifying conditional compilation for different SDK capabilities.

This commit introduces a **new capability** by adding an optional `phyConfigContent` field to the **`QsfpServiceConfig` Thrift struct**, which will store raw PHY/retimer configuration content as a single-line string. This enhancement is crucial for **platforms with external PHYs** (e.g., Ladakh) that require specific configuration. The **FSDB model** is also updated to include this new configuration path, ensuring observability. At runtime, the stored content will be written to a temporary file and passed to **SAI initialization** via the `phy_config_file` flag, enabling dynamic and flexible **PHY management**.

This commit introduces a **new capability** to extract and distribute **PHY/retimer configuration** from the QSFP service config, primarily enabling the **Ladakh platform**. The `qsfp_service` now reads `phyConfig` during `loadConfig`, persists it to `/dev/shm/fboss/qsfp_service/phy_hw_config` via `writePhyConfigToFile`, and `WedgeManager` enforces its presence for platforms requiring it during `initExternalPhyMap`. Concurrently, the `SaiPhyPlatform` in `fboss/agent` is updated to default to this new shared memory path for reading the configuration. This ensures that platforms like Ladakh can reliably access their necessary hardware settings, preventing hard failures if the configuration is absent.

This commit introduces **hardware reset support** for **MDIO bus controllers** (`BspPhyIO`) and **PHY retimers** (`BspPhyContainer`) by leveraging sysfs reset paths. A new utility, `BspResetUtils`, centralizes the reset logic, providing `holdResetViaSysfs()` and `releaseResetViaSysfs()` functions to perform a 100μs pulse reset. This **new capability** ensures proper initialization and recovery for these components, aligning with existing transceiver reset patterns and improving hardware management within the BSP. The changes affect the `BspPhyIO` and `BspPhyContainer` classes, adding methods like `init()` and `initPhy()` to manage the reset sequence before initialization. This provides a standardized and robust mechanism for hardware resets across relevant BSP components.

This commit introduces **PIM-level orchestration** for physical layer (PHY) component resets within the **Board Support Package (BSP) PIM Container**. It adds new methods, `initAllPhyIOControllers()` and `initAllPhys()`, to `BspPimContainer` which enable the **resetting of all MDIO bus controllers and PHY retimers** across a Pluggable Interface Module. This **new feature** centralizes the management of PHY initialization, providing a unified interface for ensuring proper hardware state. It builds upon existing lower-level reset functionalities, streamlining the overall PHY subsystem bring-up and recovery processes.

This commit **integrates coldboot PHY reset functionality** into the **`BspSaiPhyManager`**'s `initExternalPhyMap()` method. It introduces a new sequence that **resets MDIO bus controllers and PHY retimers** for each PIM during **coldboot initialization**, ensuring a clean hardware state. This **enhancement** prevents potential issues by performing necessary hardware resets only when not performing a warmboot, thus preserving existing PHY state in that scenario.

This commit introduces a **new configuration option** by adding the `FLAGS_enable_xphy_link_training` gflag, which provides **dynamic control over link training for XPHY ports**. This allows for the **conditional enabling or disabling of link training** within the **SAI port management** subsystem. Specifically, the `attributesFromSwPort` function in `fboss/agent/hw/sai/switch/phy/SaiPortManager.cpp` now utilizes this flag to determine the link training behavior. This enhancement provides crucial flexibility, enabling **link training on platforms like Ladakh solely via a configuration change** without requiring code modifications.

This commit **enables** the `verifyPhyPortConnector` test for the **LADAKH800BCLS platform** within the **QSFP service hardware tests**. Previously, this critical **hardware verification test** was being skipped for `LADAKH800BCLS` because the platform was not correctly identified as a SAI platform in `fboss/qsfp_service/test/hw_test/HwQsfpEnsemble.cpp`. By adding `PLATFORM_LADAKH800BCLS` to the `isSaiPlatform` check, this **maintenance** change ensures comprehensive **test coverage** and improved reliability for the platform's physical port connectors.

This commit **implements** the `getConfigOnePort` function within the **`SaiPhyRetimer`** component, introducing a **new capability** to retrieve the programmed configuration of an external PHY port. This function supports reading configuration both from the **SAI cache** (when `readFromHw=false`) and **directly from hardware** (when `readFromHw=true`). The addition of this direct hardware read functionality is essential for future hardware tests, enabling **verification of correct XPHY port configurations**. The header file `fboss/lib/phy/SaiPhyRetimer.h` is updated to reflect this new implementation.

This commit **introduces a new `readFromHw` boolean parameter** to the `getConfigOnePort` interface within the **`ExternalPhy` subsystem**. This **API modification** allows callers to explicitly control whether physical layer configuration data is fetched directly from hardware or from a cached SAI store. While initially a no-op for legacy PHY implementations, this parameter is a **new capability** that prepares the **`SaiPhyRetimer` class** for future direct hardware interaction, particularly for platforms like Ladakh. The `PhyManager` functions `getHwPhyPortConfig` and `getHwPhyPortConfigLocked` were also updated to propagate this new parameter, ensuring the parameter can be passed through the call stack.

This commit **improves the accuracy of QSFP service hardware tests** by modifying the `verifyPhyPortConfig` helper in `fboss/qsfp_service/test/hw_test/HwPortUtils.cpp`. It now explicitly calls `getHwPhyPortConfig` with `readFromHw=true`, ensuring that physical port configurations are read directly from the **physical hardware** rather than relying on cached SAI store values. This **test enhancement** is critical for **reliable verification** on platforms like `ladakh` and future `xphy NPIs` that utilize the generic `saiphyretimer` path, preventing tests from passing incorrectly due to stale data.

This commit **fixes a failing hardware test** by extending the **XPHY firmware testing framework** to support the **PLATFORM_LADAKH800BCLS** platform. It adds a specific case to the `CheckDefaultXphyFirmwareVersion` test within `HwXphyFirmwareTest.cpp`, which was previously failing due to a lack of Ladakh-specific handling. This **maintenance update** ensures the test correctly validates XPHY firmware by accounting for Ladakh's unique reporting, where only the `SAI_SWITCH_ATTR_FIRMWARE_MAJOR_VERSION` is supported. The change prevents future test failures and guarantees accurate firmware version checks for the Ladakh platform.

This commit **enhances the `CheckPortsProgrammed` test** within the `HwStateMachineTest` suite to include verification for **backplane XPHY ports**. Previously, the test only validated XPHY ports associated with transceivers, leaving direct switch-to-switch connections unchecked. This **test coverage improvement** ensures that all XPHY ports, including those without transceivers, are correctly programmed during the **port configuration flow**, thereby preventing undetected issues in critical backplane connections.

This commit **introduces a suite of helper functions** to `fboss/lib/phy/SaiPhyRetimer.cpp`, significantly enhancing the **SAI PHY retimer configuration capabilities**. These new utilities, including `getFecMode`, `toLaneVector`, and `getSerdesAttrFromHw`, provide essential logic for **converting FEC modes**, **manipulating lane vectors**, and **extracting serdes attributes** directly from hardware. This **new feature** lays crucial groundwork for the upcoming `getConfigOnePort` functionality, improving code organization and reusability within the **FBOSS PHY library** for managing retimer settings. The added functions streamline the process of building Tx settings and retrieving various port and serdes attributes.

This commit **removes management port information** from the **platform-mapping CSVs** as a **temporary workaround** to prevent a **critical agent crash**. The agent currently fails when attempting to process these specific details within the platform mappings. This change ensures agent stability by omitting the problematic data, with the intention to reintroduce it once the underlying agent bug is resolved.

This commit **optimizes platform initialization and resource management** by **removing the default invocation of ACL APIs for SAI-based PHY platforms**, specifically targeting **ASIC_TYPE_AGERA3 (Ladakh)**. It modifies `SaiPhyPlatform::getSupportedApiList` to explicitly exclude the ACL API from the supported list and updates the `SaiAclTableManager` constructor to set ACL entry priorities to zero for `ASIC_TYPE_AGERA3`. This **maintenance** change prevents unnecessary ACL API calls after switch creation, as these APIs are typically only required when MACSEC support is enabled, which is not the default for `AGERA3`. The change ensures that the system only attempts to use ACL functionality when it is genuinely needed, improving efficiency.

This commit introduces a **bug fix** to prevent **undefined behavior** within the `getCabledPortTranceivers` function in `fboss/qsfp_service/test/hw_test/HwPortUtils.cpp`. Previously, the function would dereference an optional `TransceiverID` without checking its presence, leading to issues when processing **xphy-backplane ports** that inherently lack transceivers. This oversight caused **hardware tests** to incorrectly report non-existent transceivers, resulting in false failures. The fix now properly checks for the existence of a transceiver ID, ensuring accurate reporting and correct test outcomes for all port types.