Wei Dai

This commit **enhances the robustness and reliability of multinode tests** within the `fboss/agent/test/agent_multinode_tests` suite. It **improves the `verifyDsfCluster` utility** by increasing its retry count from 10 to 100, providing more time for **DSF cluster verification** to converge. Concurrently, the **`verifyHyperPortMemberCableLengthForSwitch` function** is made more accurate and informative by adding a check that returns `false` if a hyper port member's `operState` is not `UP`, preventing erroneous cable length checks on inactive ports. Furthermore, **successful cable length measurements are now logged**, significantly improving observability and debugging capabilities for **hyper port configurations**. These **test enhancements** will lead to more stable and informative test results for **DSF and hyper port functionalities**.

This commit **enhances the multi-NPU test framework** by converting the `reachDiscard` test in `AgentFabricSwitchTests.cpp` to operate on a **per-NPU basis**. It replaces the `for (auto switchId : getFabricSwitchIdsWithPorts())` loop with `getCurrentSwitchIdForTesting()`, making the test more granular and aligned with the multi-NPU testing paradigm. Additionally, this change **fixes a pre-existing bug** in hardware statistics collection where `getHwSwitchStats()` was incorrectly called with `switchId` instead of `switchIndex`. This bug previously caused stats fetching to fail for secondary NPUs on platforms where `switchId` and `switchIndex` differ, ensuring **reliable statistics reporting** for multi-NPU setups by correctly using `getSwitchIndexFromSwitchId()`.

This commit **fixes a bug** in the **FBOSS agent multi-node tests** by modifying the `verifyHyperPortMemberCableLengthForSwitch` function. Previously, this test incorrectly attempted to verify cable length for hyper port members that were in a 'DOWN' operational state, leading to test failures and 60-second timeouts because cable length data is only collected for active ports. The change in `AgentMultiNodeDsfUtils.cpp` now **skips cable length verification for down hyper port members**, aligning the test's behavior with the agent's data collection logic. This **improves test reliability** and stability by preventing spurious failures caused by transient port states.

This commit **fixes a bug** in the **SAI port management** logic by **correcting the application of global flow control mode** for **HyperPort Members**. Specifically, the `attributesFromSwPort` function in `fboss/agent/hw/sai/switch/npu/SaiPortManager.cpp` is updated to **skip setting `SAI_PORT_ATTR_GLOBAL_FLOW_CONTROL_MODE` on HyperPort Members**, as their flow control should be inherited from the parent **HyperPort**. This change aligns with the SAI attributes proposal, ensuring **compliant NPU configuration** and preventing potential misconfigurations for **HyperPort** setups. Additionally, it introduces a check for unsupported pause configurations, enhancing the overall robustness of the **port configuration process**.

This commit **fixes** the handling of **Priority Flow Control (PFC) attributes** for **hyper port member ports** within the FBOSS agent's SAI layer. It **prevents PFC attributes from being set directly on member ports** during port creation (`attributesFromSwPort`) and guards against runtime PFC programming (`programPfc`) for them. This **compliance update** ensures that PFC configuration for hyper port members is correctly determined by their parent hyper port, aligning with the hyperport SAI attributes proposal and maintaining **configuration consistency** across the `SaiPortManager`.

This commit **fixes** an issue in the **SAI port management** by **preventing the incorrect application of PRBS (Pseudo-Random Binary Sequence) attributes to aggregate HyperPorts**. Previously, PRBS attributes were configured for all port types supporting the `SAI_PRBS` feature, but per the SAI attributes proposal, these physical layer diagnostics are only relevant for individual member ports or normal ports. The change in `fboss/agent/hw/sai/switch/npu/SaiPortManager.cpp` within the `attributesFromSwPort` function ensures that `SAI_PORT_ATTR_PRBS_POLYNOMIAL` and `SAI_PORT_ATTR_PRBS_CONFIG` are now correctly skipped for HyperPorts, **improving compliance** with the SAI specification and **ensuring proper diagnostic behavior** for physical layer ports.

This commit **refines the configuration logic for Telemetry And Monitoring (TAM) objects** within the **SaiPortManager**, specifically addressing **HyperPort** functionality. It introduces port-type guards in the `setTamObject()` and `resetTamObject()` functions in `SaiPortManager.cpp` to prevent TAM object programming on **HyperPort Member** ports. This change is a **compliance update** aligning with the SAI attributes proposal, which specifies that `SAI_PORT_ATTR_TAM_OBJECT` should only be applied to logical aggregate HyperPorts. By ensuring TAM for mirror-on-drop operates exclusively at the aggregate port level, this **improves the correctness and efficiency of TAM operations** by avoiding unnecessary or incorrect configurations on individual member lanes.

This commit **enhances the multinode testing framework** by introducing a new test case, `AgentMultiNodeHyperPortTest`, to **verify cable length reporting for hyper port members on EDSWs**. It adds robust logic to query aggregate port information and confirm that `cableLengthMeters` is correctly populated (>= 0) for each hyper port member, utilizing retry mechanisms to account for periodic stats collection. This **new test capability** also resolves a previous TODO, integrating cable length verification into the existing `verifyPortsForEdsws()` function as part of the broader **DSF cluster validation**, ensuring critical physical link properties are accurately reported.

This commit introduces a **stability improvement** to the **FBOSS agent's Link Aggregation Manager** by adding a crucial null check. Specifically, it **fixes a potential segfault** within the `updateHyperPortState` function in `LinkAggregationManager.cpp`. Previously, the result of `getNodeIf()` was immediately dereferenced without validation, which could lead to a crash if a hyper port was concurrently removed due to a **race condition**. The change now gracefully handles this scenario by checking for `nullptr`, logging a warning, and returning an empty state, thereby preventing agent crashes and enhancing overall system robustness.

This commit **enhances network diagnostics** by **enabling FEC error detection (FDR)** for **HYPER_PORT_MEMBER** ports within the **FBOSS SAI port management** layer. Previously, FDR was only active for `INTERFACE_PORT` types, but `HYPER_PORT_MEMBER` ports are the actual physical interfaces carrying traffic where FEC errors occur. By modifying the `attributesFromSwPort` function in `SaiPortManager.cpp`, this change ensures that `SAI_PORT_ATTR_FDR_ENABLE` is correctly applied to these critical physical ports. This **feature enhancement** and **fix** allows for the proper collection of FEC codeword statistics, significantly improving **error monitoring and visibility** on high-speed network links.

This commit **implements hyper port verification** within the `verifyDsfCluster()` utility, specifically for **DSF cluster multi-node tests**. It adds checks to ensure hyper port members are operationally up with correct LACP forwarding, and that the aggregate hyper port is enabled and up. This **enhances the test coverage and reliability** of the DSF cluster by validating critical hyper port states using `getAggregatePortTable` and `getPortIdToPortInfo` Thrift APIs. This is a **new capability** for the testing framework, addressing previously unimplemented TODOs in `fboss/agent/test/agent_multinode_tests/AgentMultiNodeDsfUtils.cpp`.

This commit **fixes a bug** where transmit (TX) settings were incorrectly reported as zero in the `fboss2 show interface ethXXX phy` CLI output. It **enhances the `SaiSwitch` component** within the **FBOSS agent's hardware abstraction layer** by modifying `SaiSwitch::updatePmdInfo()` to correctly populate `txSettings` information. A new helper function, `SaiPortManager::getTxSettings`, was introduced to retrieve these settings directly from **SAI attributes** for PMD lanes. This ensures that the **physical interface diagnostic information** is accurate and complete, providing network operators with **correct and complete interface statistics**.

This commit **fixes a critical bug** in the **FBOSS agent's fabric overdrain monitoring**, specifically for **VOQ switches with hyper ports**. Previously, the `computeFabricOverdrainPct` function in `fboss/agent/PortUpdateHandler.cpp` incorrectly excluded `HYPER_PORT` types from its Network Interface (NIF) bandwidth calculation, leading to a misleading 0% overdrain report and masking actual fabric capacity issues. By **including `HYPER_PORT` in the bandwidth accounting**, this change ensures **accurate reporting of fabric utilization**, providing operators with correct telemetry for capacity planning and problem detection. `HYPER_PORT_MEMBER` types are deliberately omitted to prevent double-counting, as the hyper port's speed already represents its aggregate members.

This commit introduces a **defensive programming improvement** within the **FBOSS agent's Link Aggregation Manager**. It **initializes** the `newPortState` variable to `cfg::PortState::DISABLED` in the `updateHyperPortState()` function. This change **prevents the potential reading of an uninitialized variable**, enhancing the **robustness** and **stability** of port state management. Although current logic guards against immediate issues, this **pre-emptive bug fix** mitigates future fragility and potential errors in **hyper-port state updates**.

This commit **fixes a critical bug** in the **FBOSS CLI `port details` command** where the `_convert_bps` method would crash with an `AssertionError` when processing port speeds of 1 Tbps or higher. The **port speed conversion logic** within `fboss/py/fboss/cli/commands/port.py` has been enhanced to correctly handle **Tbps (terabits per second)** by adding it to the unit conversion table. Furthermore, the problematic assertion has been replaced with a graceful fallback, ensuring the CLI command now **displays all port speeds accurately and robustly**, preventing crashes and improving user experience for high-speed network interfaces.

This commit **enhances the `fboss2 show aggregate-port` CLI command** by providing a clearer distinction between the physical link status and LACP forwarding state for aggregate port members. Previously, only a single "Up" field represented LACP forwarding; now, a dedicated "Link" field (Up/Down) is displayed alongside a renamed "Fwding" field (True/False). This **usability improvement** involves **updating the CLI's internal Thrift model** to include an `isLinkUp` field and **correcting test data** to reflect realistic port states. The change significantly **improves network troubleshooting capabilities** by offering more granular diagnostic information for aggregate ports.

This commit performs a **configuration alignment and feature enablement** for the **Jericho4 simulator (J4SIM)**, bringing its behavior closer to the MERU800BIA platform. It **enables Virtual Output Queueing (VOQ) support** for J4SIM by moving the feature out of a simulator-excluded block in `Jericho4Asic.cpp`, ensuring consistent VOQ configuration. Additionally, it **adjusts Control Plane Policing (CoPP) configurations** by removing DHCP and DHCPv6 Rx reason traps for Jericho4 in `CoppTestUtils.cpp`, anticipating a future ACL-based handling for these packets. Finally, support for the `PORT_MTU_ERROR_TRAP` feature is **temporarily disabled** for J4SIM due to current recycle port limitations.

This commit **updates the configuration** for the **`j4sim` agent**, specifically within the `fboss/oss/hw_test_configs/j4sim.agent.materialized_JSON` file. It introduces two new vendor-specific parameters, `rif_id_max` set to `0x2000` and `AI23_mode` set to `4`, into the `vendor_config` section. This **configuration update** is essential for the proper simulation and operation of the **J4 hardware platform**, enabling necessary settings for its functionality.

This commit introduces a **new capability** to the **FBOSS agent's Thrift API** by adding `flushNeighborEntries`. This new method allows for the **bulk flushing of multiple neighbor entries** in a single API call, addressing performance issues and rate limiting encountered when flushing many entries individually. It impacts the **neighbor management subsystem** by providing a more efficient mechanism for clearing neighbor states. The change includes updates to `ctrl.thrift` for API definition, `ThriftHandler.cpp` for implementation, and comprehensive tests in `ThriftTest.cpp`.

This commit **refactors** the `setPrbsOnAllInterfaces` function within the **AgentEnsemblePrbsTest** to enhance the efficiency of PRBS configuration. It now leverages the new `setInterfacesPrbs` bulk API, which groups **ASIC ports** by polynomial and sets their PRBS in a single RPC call per group, replacing the previous loop of individual calls. This **optimization** significantly streamlines the PRBS setup for ASIC interfaces, while **non-ASIC (transceiver) ports** continue to be handled individually via `QsfpService`. The change primarily impacts the **PRBS testing framework**, making the test setup for ASIC interfaces more performant and reducing RPC overhead.