Midhun Somasundaran

This commit **removes** the `HwArsSprayTest.VerifyEcmpIdManagement` test case from the suite. This specific test, which verified `hwswitch`'s behavior when running out of Dynamic Load Balancing (DLB) capacity and transitioning to spray mode, is now **obsolete**. The underlying **ECMP resource manager** has been enhanced to effectively handle this scenario, rendering the explicit test for `hwswitch`'s spray mode redundant. This is a **maintenance** task that **simplifies the test suite** by eliminating unnecessary and outdated verification.

This commit **implements** the previously missing `getAllEcmpDetails` functionality within the **SAI hardware abstraction layer**, specifically in the `SaiNextHopGroupManager` and `SaiSwitch` components. This **new capability** enables the retrieval of ECMP and flowlet details from SAI-based hardware, returning the SAI object ID as the ECMP ID. The primary impact is a **test fix**, allowing the `HwArsFlowletTest.VerifyGetEcmpDetails` test to run and pass correctly on **TH6 platforms**, which was previously unported due to the absence of this API.

This commit **fixes intermittent failures** in the **SaiAclTableGroupTrafficTests** on the `th6` platform, addressing test flakiness where ACL stat checks were timing out on boundary conditions. This **maintenance** work **refactors the ACL stat verification logic** within tests like `testAclTableGroupTraffic` and `testDscpTtlAclTables`. Specifically, it now utilizes `WITH_RETRIES` and `EXPECT_EVENTUALLY_EQ` to allow for transient delays in stat updates. This change significantly **improves the reliability of hardware tests** by ensuring more stable and accurate validation of **ACL functionality** within the **SAI hardware agent**.

This commit provides a **bug fix** for the **PFC (Priority Flow Control) subsystem** on **Tomahawk6 (TH6) ASICs**, specifically addressing issues with watchdog timer granularity. It **adjusts buffer pool and PFC buffer parameters** within `fboss/agent/test/utils/PfcTestUtils.cpp` to correctly account for SDK-reserved memory and utilize cell-aligned values. This crucial configuration change enables the `AgentPfcWatchdogGranularityTest.verifyPfcWatchdogTimerGranularity` test to pass, ensuring accurate validation of 1ms PFC watchdog timer granularity on TH6 platforms. The work improves the reliability of **PFC functionality** and its associated testing by providing appropriate pool settings.

This commit **fixes failing `AgentSflowMirror` tests** on the **Tomahawk6 (th6) ASIC platform**. The tests were previously failing because the `MANAGEMENT_PORT` was not properly configured, preventing the capture of mirrored packets. This **bug fix** updates the `fboss/agent/hw/switch_asics/Tomahawk6Asic.cpp` configuration, specifically changing the `desiredLoopbackModes` for the `MANAGEMENT_PORT` from `NONE` to `MAC`. This ensures the management port is operational during testing, thereby **resolving test flakiness** and validating the sFlow mirroring functionality on th6.

This commit **enables support for bulk ECMP group APIs** on **Tomahawk6 ASICs**, specifically activating the `BULK_CREATE_ECMP_MEMBER` feature. It updates the `isSupported` method within the `Tomahawk6Asic.cpp` file, part of the `fboss/agent`'s hardware abstraction layer. This **new feature enablement** leverages recent software updates (14.2) that make these bulk APIs functional on TH6. The change allows for **more efficient and scalable management of ECMP groups** on Tomahawk6 devices by utilizing bulk operations, improving performance for network configurations involving many ECMP members.

This commit **enhances the test coverage** for the **VLAN subsystem** within `fboss/agent`. It introduces **new test cases** in `VlanTests.cpp` to validate the functionality of recently added fields related to **VLAN priority tagging**. Furthermore, it includes tests specifically designed to verify the correct **warmboot migration** of these new VLAN configurations, ensuring data integrity across system restarts. This work is a **testing enhancement** that improves the reliability and stability of VLAN management by thoroughly validating new features and their persistence.

This commit **enhances the `AgentDot1qMappingTest` hardware tests** within the `fboss/agent` suite to improve coverage for **dot1q mapping**. Specifically, it introduces test cases that utilize **VLAN ID 0 packets**, which are used for priority tagging. The updated test now punts these packets to the CPU and employs a snooper to accurately verify the VLAN ID and Priority Code Point (PCP) of the captured frames. This **test enhancement** ensures robust validation of the network device's ability to correctly process and map priority-tagged traffic.

This commit introduces **support for VLAN priority tagging**, enabling more granular control over traffic prioritization within VLANs. This **new capability** involves a significant **refactoring of the internal VLAN state representation**, replacing simple port-tagging pairs with a new `VlanInfo` struct that explicitly includes a `priorityTagged` flag. The changes span the **SAI and BCM hardware abstraction layers** for correct attribute setting, update the **FBOSS agent's VLAN state management** (including Thrift schema and warm boot migration), and necessitate widespread updates across various **agent components** to correctly access and utilize the updated VLAN port information. This ensures that network devices can properly handle and prioritize VLAN-tagged frames based on configuration.

This commit **adds comprehensive unit tests** to validate the recently introduced **virtual and non-virtual ECMP group support** within the **`EcmpResourceManager`**. The new test cases in `fboss/agent/test/BaseEcmpResourceManagerTest.cpp` specifically verify the correct configuration and limit handling for these advanced routing groups. This **enhances the robustness and reliability** of the `EcmpResourceManager` by ensuring the new functionality behaves as expected, preventing regressions and validating its core logic. This **test-only change** is crucial for maintaining the stability of the ECMP management subsystem.

This commit introduces a **new feature** to the **resource manager** by adding comprehensive support for **virtual ARS groups**, also referred to as virtual DLBs. It **enhances resource accounting** within the `fboss/agent/ResourceAccountant` module, implementing new checks and updating existing ECMP/ARS resource calculations to correctly manage these dynamic entities. Specifically, methods like `isVirtualArsGroup`, `checkArsResource`, and `updateArsVirtualGroupConfig` are introduced or modified to ensure proper resource allocation. This change allows the system to accurately account for and protect resources utilized by virtual ARS groups, with corresponding test updates in `AgentFlowletSwitchingTests.cpp` to validate the new DLB resource checks.

This commit introduces a **new capability** to the **`EcmpResourceManager`** within `fboss/agent` by adding the `isVirtualArsGroup()` helper function, which enables the classification of virtual ARS groups. It enhances **ECMP resource accounting** by updating the `getPrimaryEcmpAndMemberCounts()` method to return a tuple, providing more granular details including virtual group classification and member counts. This change refines how ECMP groups and their members are counted and managed, ensuring more accurate resource allocation. The updated return type of `getPrimaryEcmpAndMemberCounts()` necessitates corresponding adjustments in consuming code, as demonstrated by the updated test utilities.

This commit introduces a **new capability** by adding `virtualEcmpGroupsCnt` to the `InputOutputState` struct within the **ECMP resource management** subsystem. This new member allows the `EcmpResourceManager` to track the count of virtual ECMP groups. The logic in `fboss/agent/EcmpResourceManager.cpp` is updated to utilize this count in various operations, including `consolidate()`, `checkPrimaryGroupAndMemberCounts()`, and during **limit calculations**, thereby enhancing the accuracy and robustness of ECMP resource management.

This commit introduces a **new capability** to configure and manage **virtual ECMP group limits** within the `fboss/agent` component's **ECMP resource management**. It **enhances** the `EcmpResourceManager` constructor in `fboss/agent/EcmpResourceManager.h` to accept optional parameters for virtual ECMP group limits and width. Consequently, the `makeEcmpResourceManager` factory function in `fboss/agent/EcmpResourceManager.cpp` is updated to pass these new virtual group parameters, ensuring proper initialization. A related test in `fboss/agent/test/BaseEcmpResourceManagerTest.cpp` is also adjusted to correctly assert primary ECMP group limits based on ASIC values, reflecting the updated configuration logic. This allows for more precise control over virtual ECMP resource allocation.

This commit introduces a **new feature** to the **ECMP resource management** system, enabling the `EcmpResourceManager` to dynamically determine the maximum number of ARS groups. Specifically, the `makeEcmpResourceManager` function in `EcmpResourceManager.cpp` is updated to prioritize the `maxArsVirtualGroups` setting from the flowlet switching configuration when **flowlet switching** is enabled, overriding the ASIC's default `maxArsGroups`. This enhancement provides greater flexibility in managing **virtual ARS groups** and ensures the resource manager correctly adapts to configured virtual ARS group limits. The change is thoroughly validated by **new unit tests** added to `BaseEcmpResourceManagerTest.cpp`.

This commit introduces a **new feature** to the **ECMP resource management configuration** by adding explicit support for **virtual ECMP group configuration**. It modifies the `EcmpResourceManagerConfig` class in `fboss/agent` to distinguish between virtual and non-virtual ARS groups, updating the constructor and the `ecmpLimitReached` logic. New member variables, getter methods like `getMaxVirtualEcmpGroups`, and a `computeMaxVirtualEcmpGroups` function are added to manage specific parameters for virtual groups. This enhancement provides greater flexibility in ECMP resource allocation by allowing distinct configuration and limit checks for virtual ECMP groups.

This commit introduces **new unit tests** for the **ResourceAccountant** component, specifically validating the resource allocation and limits for **virtual ARS groups**. A `getMaxArsWidth` method was also added to `MockAsic` to provide necessary mock data, supporting these new tests. This **testing enhancement** significantly improves the **test coverage** and reliability of resource management logic within the `fboss/agent` subsystem.

This commit **fixes a test failure** in the **FBOSS agent hardware tests** related to **Flowlet Switching**. The `wideArsFlowlet` test was failing because its configuration, specifically the combination of width and depth, resulted in a total ARS profile count (1024) exceeding the maximum supported value (1023). To resolve this, the commit **adjusts the maximum number of virtual ARS groups** from 256 to 255 within `fboss/agent/test/agent_hw_tests/AgentFlowletSwitchingTests.cpp`. This **test maintenance** ensures the `wideArsFlowlet` test passes reliably by respecting system limits, thereby preventing false negatives during validation of **Flowlet Switching** functionality.

This commit introduces a **new capability** to the **FBOSS agent's switch configuration**, enabling explicit control over priority tagging on a per-VLAN port basis. A new boolean field, `emitPriorityTags`, has been added to the `VlanPort` struct in `switch_config.thrift`. This allows the agent to be configured to emit priority-tagged frames, specifically those with VLAN ID 0 and PCP bits set, for enhanced QoS signaling. Consequently, the `ChildThriftPath` specialization for `VlanPort` in `fsdb_model_thriftpath.h` was **refactored** to incorporate this new configuration option, ensuring the FSDB model accurately reflects the updated Thrift schema.

This commit introduces **new configuration parameters** for the **FBOSS agent's flowlet switching** mechanism. It adds setter and getter APIs for `maxArsVirtualGroupWidth` and `maxArsVirtualGroups` to the `FlowletSwitchingConfig` class, enabling programmatic access to these values. This **new feature** also includes the necessary logic in `ApplyThriftConfig.cpp` to apply these virtual group field configurations from the Thrift configuration. The change provides more granular control over Adaptive Routing System (ARS) virtual group settings within the flowlet switching subsystem.