[vslib] Add High Frequency Telemetry (HFT) support for virtual SAI by 50n1c-rnsft · Pull Request #1812 · sonic-net/sonic-sairedis

50n1c-rnsft · 2026-03-25T00:59:35Z

Description

Add High Frequency Telemetry (HFT) / Stream Telemetry support to the virtual SAI (vslib), enabling the full HFT pipeline to run in virtual/test environments.

What this PR does

Task 1 — TAM object initialization:

Add SAI_SWITCH_ATTR_TAM_ST_REPORT_CHUNK_SIZE (default: 65535) and SAI_SWITCH_ATTR_TAM_ST_CHUNK_COUNT (default: 0) to set_initial_tam_objects()
Add SAI_OBJECT_TYPE_TAM_TEL_TYPE and SAI_OBJECT_TYPE_TAM_COUNTER_SUBSCRIPTION to supported object types
TAM_COUNTER_SUBSCRIPTION CRUD works through the existing generic create_internal() path

Task 2 — IPFIX template generation:

Replace placeholder random data in refresh_tam_tel_ipfix_templates() with real IPFIX template binary per HLD 7.2.2
Enterprise number encoding: (stat_id << 16) | object_type
Element ID: label | 0x8000 (enterprise bit set)
First field: observationTimeNanoseconds (Element ID=325, 8 bytes)
Label range validation with SWSS_LOG_WARN for values exceeding 15-bit range

Task 3 — STEL genetlink sender:

SwitchStateBaseStel.cpp: RAII GenlConnection wrapper using libnl-genl (genl_connect(), genl_ctrl_resolve(), genlmsg_put(), nla_put(), nl_send_auto())
Worker thread generates fake IPFIX data records at configurable intervals
Wire setTamTelType() to start/stop STEL stream on START_STREAM/STOP_STREAM state changes
Proper cleanup in ~SwitchStateBase() destructor

Task 4 — Stats capability extension:

Add queryBufferPoolStatsCapability() (15 stats) and queryIngressPriorityGroupStatsCapability() (9 stats)
Extend queryStatsCapability() to dispatch BUFFER_POOL and INGRESS_PRIORITY_GROUP object types

Files changed (14 files, +1089/-156 lines)

File	Description
`vslib/SwitchStateBase.cpp`	Task 1 + Task 2 + Task 4 + destructor fix
`vslib/SwitchStateBase.h`	New method declarations + STEL thread members
`vslib/SwitchStateBaseStel.cpp`	New — STEL genetlink sender + worker thread
`vslib/sonic_stel_uapi.h`	New — UAPI header (shared with kernel module)
`vslib/Makefile.am`	Add `-lnl-genl-3 -lnl-3` link flags
`vslib/tests.cpp`	Update expected supported object type count (8→10)
`syncd/Makefile.am`	Add `-lnl-genl-3 -lnl-3` to SAILIB
`syncd/tests/Makefile.am`	Add `-lnl-genl-3 -lnl-3` link flags
`saidiscovery/Makefile.am`	Add `-lnl-genl-3 -lnl-3` to SAILIB
`saisdkdump/Makefile.am`	Add `-lnl-genl-3 -lnl-3` to SAILIB
`tests/Makefile.am`	Add `-lnl-genl-3 -lnl-3` to SAILIB
`tests/aspell.en.pws`	Add new words (libnl, syscall, templatehdr)
`unittest/vslib/TestTAMIpfixTemplate.cpp`	New — 4 unit tests for IPFIX template generation
`unittest/vslib/Makefile.am`	Add test file

Kernel module

The sonic_stel kernel module (genetlink family registration) is in a separate PR:

[platform/vs] Add sonic_stel kernel module for virtual stream telemetry sonic-buildimage#26390 — placed in platform/vs/sonic-stel-module/

Testing

Build: dpkg-buildpackage -Psyncd,vs,nopython2 passes in sonic-slave-bookworm:master-amd64 CI container
Unit tests: 203/203 passed (including 4 new TAMIpfixTemplateTest cases)
- EmptySubscription_GeneratesMinimalTemplate
- SingleSubscription_CorrectBinaryFormat
- MultipleSubscriptions_CorrectFieldCountAndEncoding
- UnmatchedSubscription_Filtered

References

mssonicbld · 2026-03-25T00:59:42Z

/azp run

linux-foundation-easycla · 2026-03-25T00:59:48Z

The committers listed above are authorized under a signed CLA.

✅ login: 50n1c-rnsft / name: Ze Gan agent (15fc3e3, 1bbfb9e, 28477b6, 2c80d87, 2fc9ddc, 5527bb2, 563241f, 5f6d2a4, 64f2250, 6575e80, 6b4881c, 7bb4977, 7f25549, 88299a5, 9683082, af43e00, b007856, b54b8dc, c0f34d4, cb87a3f)

azure-pipelines · 2026-03-25T00:59:55Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:04:39Z

/azp run

azure-pipelines · 2026-03-25T01:04:50Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:06:25Z

/azp run

azure-pipelines · 2026-03-25T01:06:32Z

Azure Pipelines will not run the associated pipelines, because the pull request was updated after the run command was issued. Review the pull request again and issue a new run command.

mssonicbld · 2026-03-25T01:06:36Z

/azp run

azure-pipelines · 2026-03-25T01:06:46Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:10:15Z

/azp run

azure-pipelines · 2026-03-25T01:10:27Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:12:21Z

/azp run

azure-pipelines · 2026-03-25T01:12:33Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:15:25Z

/azp run

azure-pipelines · 2026-03-25T01:15:36Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:16:47Z

/azp run

azure-pipelines · 2026-03-25T01:16:57Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:20:03Z

/azp run

azure-pipelines · 2026-03-25T01:20:14Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:22:34Z

/azp run

azure-pipelines · 2026-03-25T01:22:50Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T01:25:49Z

/azp run

azure-pipelines · 2026-03-25T01:26:05Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T03:06:43Z

/azp run

azure-pipelines · 2026-03-25T03:06:57Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T03:16:38Z

/azp run

azure-pipelines · 2026-03-25T03:16:49Z

Azure Pipelines successfully started running 1 pipeline(s).

- Add 'libnl', 'syscall', 'templatehdr' to tests/aspell.en.pws - Add SWSS_LOG_ENTER() to test fixture methods in TestTAMIpfixTemplate.cpp (SetUp, getIpfixTemplates, createCounterSubscription, validateTemplateHeader) - Add swss/logger.h include to test file Signed-off-by: Ze Gan agent <ganze_12345@qq.com>

mssonicbld · 2026-03-25T14:45:32Z

/azp run

azure-pipelines · 2026-03-25T14:45:47Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T14:55:50Z

/azp run

azure-pipelines · 2026-03-25T14:56:03Z

Azure Pipelines successfully started running 1 pipeline(s).

Add a minimal generic netlink kernel module (sonic_stel) for the VS platform that registers the 'sonic_stel' family with an 'ipfix' multicast group. This module acts as a relay: vslib (virtual SAI) sends IPFIX data records via SONIC_STEL_CMD_SEND_IPFIX, and the module multicasts them to countersyncd via genlmsg_multicast. Files: - platform/vs/sonic-stel-module/ - kernel module source + debian packaging - platform/vs/sonic-stel-ko.mk - buildimage rule Related PR: sonic-net/sonic-sairedis#1812 Signed-off-by: Ze Gan agent <ganze_12345@qq.com>

mssonicbld · 2026-03-25T15:01:37Z

/azp run

azure-pipelines · 2026-03-25T15:01:52Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T15:04:02Z

/azp run

Add a minimal generic netlink kernel module (sonic_stel) for the VS platform that registers the 'sonic_stel' family with an 'ipfix' multicast group. This module acts as a relay: vslib (virtual SAI) sends IPFIX data records via SONIC_STEL_CMD_SEND_IPFIX, and the module multicasts them to countersyncd via genlmsg_multicast. Files: - platform/vs/sonic-stel-module/ - kernel module source + debian packaging - platform/vs/sonic-stel-ko.mk - buildimage rule Related PR: sonic-net/sonic-sairedis#1812 Signed-off-by: Ze Gan agent <ganze_12345@qq.com>

azure-pipelines · 2026-03-25T15:04:16Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-25T15:45:13Z

/azp run

azure-pipelines · 2026-03-25T15:45:25Z

Azure Pipelines successfully started running 1 pipeline(s).

- Add 'genl' and 'ko' to tests/aspell.en.pws - Add '// SWSS_LOG_ENTER() omitted' comments to static helper functions: - SwitchStateBaseStel.cpp: put_u16_be, put_u32_be, put_u64_be, build_ipfix_data_message - SwitchStateBase.cpp: write_u16_be, write_u32_be - TestTAMIpfixTemplate.cpp: read_u16_be, read_u32_be Signed-off-by: Ze Gan agent <ganze_12345@qq.com>

mssonicbld · 2026-03-25T15:46:44Z

/azp run

azure-pipelines · 2026-03-25T15:46:57Z

Azure Pipelines successfully started running 1 pipeline(s).

vslib/SwitchStateBase.cpp

+        switch (attr->value.s32)
+        {
+            case SAI_TAM_TEL_TYPE_STATE_CREATE_CONFIG:
+                send_tam_tel_type_config_change(tam_tel_type_id);
+                break;
+
+            case SAI_TAM_TEL_TYPE_STATE_START_STREAM:
+            {
+                // Count counter subscriptions for this tel_type to determine num_counters
+                size_t num_counters = 0;
+                auto it = m_objectHash.find(SAI_OBJECT_TYPE_TAM_COUNTER_SUBSCRIPTION);
+                if (it != m_objectHash.end())
+                {
+                    for (auto &kv : it->second)
+                    {
+                        auto tel_it = kv.second.find(sai_serialize_attr_id(
+                            *sai_metadata_get_attr_metadata(
+                                SAI_OBJECT_TYPE_TAM_COUNTER_SUBSCRIPTION,
+                                SAI_TAM_COUNTER_SUBSCRIPTION_ATTR_TEL_TYPE)));
+
+                        if (tel_it != kv.second.end() &&
+                            tel_it->second->getAttr()->value.oid == tam_tel_type_id)
+                        {
+                            num_counters++;
+                        }
+                    }
+                }
+
+                // Default poll interval: 100ms (100000 us)
+                // In real HW this comes from TAM report config; for vslib use a reasonable default
+                uint32_t poll_interval_us = 100000;
+                uint16_t template_id = 256;
+
+                SWSS_LOG_NOTICE("Starting STEL stream for tel_type %s: %zu counters, %u us interval",
+                                sai_serialize_object_id(tam_tel_type_id).c_str(),
+                                num_counters, poll_interval_us);
+
+                if (num_counters > 0)
+                {
+                    startStelStream(poll_interval_us, template_id, num_counters);
+                }
+                break;
+            }
+
+            case SAI_TAM_TEL_TYPE_STATE_STOP_STREAM:
+                SWSS_LOG_NOTICE("Stopping STEL stream for tel_type %s",
+                                sai_serialize_object_id(tam_tel_type_id).c_str());
+                stopStelStream();
+                break;
+
+            default:
+                break;
+        }


Per RFC 7011, an IPFIX Exporting Process must send Template Records before any Data Records. The STEL worker thread was only sending Data Records (Set ID=256), causing countersyncd to fail with: 'Failed to parse IPFIX data message: Error: Missing Template at 0x14' Fix: - Add build_ipfix_template_message() to wrap template set in IPFIX header - Read IPFIX template from SAI_TAM_TEL_TYPE_ATTR_IPFIX_TEMPLATES on START_STREAM and pass to worker thread - Send Template Record before first Data Record - Re-send Template Record every ~30 seconds per RFC 7011 recommendation - Update startStelStream/stelWorkerThread signatures to accept template Tested: 203/203 unit tests pass, aspell/swsslogenter clean. Bug found by msft-internal-linux during VS testbed integration testing. Signed-off-by: Ze Gan agent <ganze_12345@qq.com>

mssonicbld · 2026-03-26T11:39:01Z

/azp run

azure-pipelines · 2026-03-26T11:39:13Z

Azure Pipelines successfully started running 1 pipeline(s).

Revert commit af43e00. Per Ze's clarification, IPFIX templates are NOT sent via the data plane (genetlink). Instead: 1. orchagent queries SAI_TAM_TEL_TYPE_ATTR_IPFIX_TEMPLATES 2. orchagent writes template to STATE_DB (session_config field) 3. countersyncd reads template from STATE_DB The genetlink channel only carries IPFIX Data Records. The 'Missing Template' error from countersyncd testing was because orchagent had not yet written the template to STATE_DB, not because vslib needed to send it. Signed-off-by: Ze Gan agent <ganze_12345@qq.com>

mssonicbld · 2026-03-26T11:41:45Z

/azp run

azure-pipelines · 2026-03-26T11:41:56Z

Azure Pipelines successfully started running 1 pipeline(s).

Signed-off-by: Ze Gan agent <ganze_12345@qq.com>

mssonicbld · 2026-03-26T15:43:39Z

/azp run

azure-pipelines · 2026-03-26T15:43:52Z

Azure Pipelines successfully started running 1 pipeline(s).

50n1c-rnsft force-pushed the feature/vslib-hft-tam-ipfix branch from 9fb21e0 to f73d1ed Compare March 25, 2026 01:25

50n1c-rnsft force-pushed the feature/vslib-hft-tam-ipfix branch from 7621c8f to 76dc23e Compare March 25, 2026 03:16

50n1c-rnsft changed the title ~~[vslib] Add HFT TAM object support and IPFIX template generation~~ [vslib] Add High Frequency Telemetry (HFT) support for virtual SAI Mar 25, 2026

50n1c-rnsft force-pushed the feature/vslib-hft-tam-ipfix branch from 6c00dcd to 17862f1 Compare March 25, 2026 14:55

50n1c-rnsft force-pushed the feature/vslib-hft-tam-ipfix branch from 17862f1 to 64d2196 Compare March 25, 2026 15:01

50n1c-rnsft force-pushed the feature/vslib-hft-tam-ipfix branch from 64d2196 to 5527bb2 Compare March 25, 2026 15:03

50n1c-rnsft force-pushed the feature/vslib-hft-tam-ipfix branch from 186e94e to 2c80d87 Compare March 25, 2026 15:46

github-advanced-security bot found potential problems Mar 25, 2026

View reviewed changes

[ci] Retrigger CI (FlexCounter.bulkChunksize flaky test)

15fc3e3

Signed-off-by: Ze Gan agent <ganze_12345@qq.com>

Conversation

50n1c-rnsft commented Mar 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

What this PR does

Files changed (14 files, +1089/-156 lines)

Kernel module

Testing

References

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

linux-foundation-easycla bot commented Mar 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

Uh oh!

mssonicbld commented Mar 25, 2026

Uh oh!

azure-pipelines bot commented Mar 25, 2026

50n1c-rnsft commented Mar 25, 2026 •

edited

Loading

linux-foundation-easycla bot commented Mar 25, 2026 •

edited

Loading