Shiva Menta

This commit performs a **maintenance chore** by **standardizing logging configurations** for several FBOSS platform services. It updates the `systemd` service unit files for `data_corral_service`, `fan_service`, `platform_manager`, and `sensor_service` to redirect their standard output and error streams. This ensures that all logs from these critical **platform services** are consistently written to dedicated files within the `/var/facebook/logs/fboss/` directory, significantly improving **log management and debugging** on FBOSS devices.

This commit provides a **maintenance fix** for the **`qsfp_service`** by updating its **systemd unit file** (`qsfp_service.service`) within the **CentOS 09.0 image templates**. The change **redirects standard output and error** from the `qsfp_service` to a specific log file, moving away from `journald`. This improves the **diagnosability and log management** for the QSFP service, making its operational output more accessible for troubleshooting on affected systems.

This commit **resolves inconsistent network connectivity** for devices built with the `centos-09.0` image, specifically addressing issues with IPv6 Stateless Address Autoconfiguration (SLAAC). It introduces an **explicit NetworkManager connection profile** for the `eth0` interface, configuring it for reliable automatic IPv4 and IPv6 address assignment via `eth0.nmconnection`. Furthermore, it enables necessary IPv6 features like Router Advertisement acceptance and autoconfiguration through a new `sysctl` configuration, and updates the image build process to ensure correct file permissions. This **bug fix** significantly improves the **network stability and consistent interaction** with affected devices.

This commit performs **maintenance updates** to the **FBOSS image build process** and several **FBOSS service configurations**. It standardizes the default log directory for `qsfp_service` and `wedge_agent` to match production environments, with `config.sh` now creating these paths and configuring SELinux permissions. Additionally, the `platform_manager` service has its Thrift SSL policy disabled, aligning with other platform stack services for the time being. These changes ensure consistent logging practices and inter-service communication settings across the FBOSS ecosystem, requiring the `policycoreutils-python-utils` package for SELinux management.

This commit **fixes a critical logging bug** in the **`distro_cli` image builder** that caused misleading output regarding the final location of generated image artifacts. Previously, logs incorrectly indicated artifacts remained in a temporary directory, but they were actually moved; the `_stage_artifact` function in `image_builder.py` now explicitly logs the correct destination path. Concurrently, this change **refactors** the **`distro_cli` logging infrastructure** by standardizing the root logger name to `'distro_cli'` and ensuring module-specific logger initialization. This ensures users receive **accurate and consistent log messages**, particularly concerning the storage paths of built images.

This commit **fixes a configuration bug** within the `kernel_only.json` distribution manifest, which is primarily used for **CI testing**. It **corrects an incorrect relative filepath** for kernel building scripts, which was previously misinterpreted because the manifest itself resides in `fboss-image/manifests` while the scripts are in `fboss-image`. This **bug fix** ensures that the **kernel build process** for the `fboss-image` project now correctly locates its dependencies, preventing build failures in the automated testing environment.

This commit implements a **bug fix** for the **artifact download functionality** within the `distro_cli` module. Previously, local files referenced in build manifests via `"download": "file:xxxx"` were inadvertently deleted from their original paths during processing. The fix modifies the `download_artifact` function in `fboss-image/distro_cli/lib/download.py` to first **copy** these local source files to a temporary directory. This prevents the accidental removal of original artifacts, ensuring the integrity of local build inputs and enhancing the reliability of manifest-based artifact handling.

This commit **introduces a new system for defining image builds** by creating the `fboss-image/manifests` directory and adding the initial `kernel_only.json` file. This **new capability** provides a structured way to specify components and the build process for various images, starting with a **kernel-only distribution**. It directly impacts the **Conveyor** image generation pipeline, enabling the creation of **hardware-specific images** for internal use and external vendors. This foundational change facilitates more tailored and efficient image deployments.

This commit performs **maintenance updates** to the **platform stack's systemd service configurations**, specifically for the **`fan_service`**, **`data_corral_service`**, and **`sensor_service`**. It **resolves service startup issues** by changing their `Type` from `notify` to `simple`, preventing them from getting stuck in an "activating" state due to unsupported `sdNotify` functionality. Additionally, the `ExecStart` commands are modified to temporarily disable `thrift_ssl_policy`, mitigating potential SSL-related startup problems until full support is implemented. This ensures these critical **platform services** can initialize correctly.

This commit introduces a **new capability** by **modifying the SSH daemon configuration** for the **CentOS 09.0 distro image**. It adds `PermitRootLogin yes` to the `99-fboss.conf` file within `sshd_config.d`, explicitly allowing root users to SSH into devices. This change ensures initial access to the device, overriding default CentOS settings due to the configuration file's numerical precedence. This is considered a **temporary measure** and will be re-evaluated once a more robust access policy, such as adding a secondary user, is established.

This commit **integrates essential debugging utilities** into the **CentOS-09.0 Fboss distro image**. It specifically adds `tcpdump`, `usbutils` (for `lsusb` access), and `ndisc6` to the image configuration via `fboss-image/image_builder/templates/centos-09.0/config.xml`. This **new feature** significantly **enhances diagnostic capabilities** for the Fboss platform, providing critical tools for network analysis, USB device enumeration, and IPv6 SLAAC troubleshooting, especially valuable during **early system bring-up** and **USB boot scenarios**.

This commit provides a **bug fix** for the **Port Manager mode** within the **Transceiver State Machine**. It addresses an issue where `needMarkLastDownTime` was incorrectly reset during remediation, leading to `lastDownTime` being re-stamped on every failed cycle. The fix modifies the `operator()` in `fboss/qsfp_service/TransceiverStateMachine.h` to preserve the correct `false` value for `needMarkLastDownTime` during the `TRANSCEIVER_PROGRAMMED` to `DISCOVERED` remediation path. This ensures that `lastDownTime` accurately reflects the last actual down event, aligning **Port Manager mode** behavior with traditional mode. New tests in `fboss/qsfp_service/test/PortStateMachineTest.cpp` validate this corrected behavior.

This commit **fixes a bug** in the **QSFP Service's `PortManager`** where the `lastDownTime` attribute was not being correctly utilized, leading to premature remediation in opticsTx.* related link tests. It introduces new helper functions like `buildTcvrActivePortSnapshot` and `updateTcvrLastDownTime` within `PortManager.cpp` and `TransceiverManager.cpp` to accurately track and update transceiver `lastDownTime` based on port active states. This **remediation** ensures the system properly accounts for port downtime before initiating recovery actions, preventing unnecessary interventions. Additionally, **test coverage** is enhanced in `PortStateMachineTest.cpp` to explicitly validate the correct marking and usage of `lastDownTime`.

This commit **refactors** and **enhances** the **QSFP service's Port Manager** by resolving seven TODO comments in `QsfpServiceHandler.cpp`. It **implements new PRBS (Pseudo-Random Binary Sequence) methods** (`getSupportedPrbsPolynomials`, `getInterfacePrbsState`, `getAllInterfacePrbsStates`) within `PortManager`, which delegate to `TransceiverManager` for transceiver components, and adds conditional routing in `QsfpServiceHandler` when `FLAGS_port_manager_mode` is enabled. This **prepares the system for future XPHY support** by correctly handling component-specific PRBS calls. Furthermore, it clarifies the behavior of four non-PRBS methods (`getPortMediaInterface`, `getPortsRequiringOpticsFwUpgrade`, `getQsfpServiceRunState`, `setInterfaceTxRx`), explaining that their data remains controlled by `TransceiverManager` even in Port Manager mode due to their transceiver-specific nature.

This commit introduces a **new capability** to integrate **XPHY `getPortInfo` failures** into the common **PIM error handling pipeline**, enabling automated **self-healing actions**. It modifies the `PhyManager` to record these specific failures within the `MultiPimPlatformSystemContainer`, which then exposes them through `getPimStates` after adding a new `XPHY_GET_PORT_INFO_FAILED` enum to `pim_state.thrift`. This enhancement to the **PIM error reporting subsystem** allows **Netstate** to detect and trigger recovery for critical hardware port information retrieval issues, significantly improving the **system's resilience** and automated recovery for **XPHY management**.

This commit performs a **refactoring** to **break the dependency** of the **PIM container statistics initialization** on `StatsPublisher` within the `fboss/lib/fpga` subsystem. Specifically, `MultiPimPlatformPimContainer` now implements a new internal function, `initPerPimFb303Stats`, to initialize per-PIM fb303 statistics directly. The `setPimContainer` method in `MultiPimPlatformSystemContainer` is updated to utilize this new standalone function, eliminating the need for `StatsPublisher`. This change enhances modularity and updates build dependencies to support the new, self-contained statistics handling.

This commit **fixes** an issue in **link tests** where `QsfpService::PortStateMachineState` was incorrectly checked for all ports defined by the agent, leading to failures on devices with backplane ports not managed by `QsfpService`. It introduces a new `qsfpServiceManagedPorts_` list within the `AgentEnsembleLinkTest` and `LinkTest` frameworks, populated with ports that have transceivers or XPHYs, mirroring the logic used by `PortManager`. Consequently, the `waitForPortStateMachineState` method now correctly restricts its checks to only these **QSFP service managed ports**, preventing erroneous test failures and improving the reliability of the **FBOSS agent's link testing infrastructure**. This is a **bug fix** that refines the scope of port state machine checks in the testing framework.