[vpp] Dedicated binary-API socket and thread for asynchronous VPP events by nicholasching · Pull Request #1951 · sonic-net/sonic-sairedis

nicholasching · 2026-06-23T16:06:11Z

Description of PR

Summary:
Part of the SAIVPP unit-test framework (docker-sai-test-vpp, #1950, see its devdocs/). Carries the asynchronous VPP event stream on a dedicated binary-API socket and background thread, separate from the synchronous request/reply command path. This resolves a mutex deadlock between the command path and the async event path that blocked bring-up of the OCP sai_test suite at PORT_COUNT=32. Reviewers can start in vslib/vpp/vppxlate/SaiVppXlate.c.

Fixes # (N/A — no upstream issue)

Type of change

Approach

What is the motivation for this PR?

The synchronous command path and the asynchronous VPP event stream shared a single binary-API socket and the global __plugin_msg_base. At scale this deadlocked between the command path and the async event path, preventing the standalone SAIVPP UT (SaiSanityTest) from passing.

How did you do it?

In vslib/vpp/vppxlate/SaiVppXlate.c: introduced a dedicated event binary-API socket (event_socket_client_main / vat_event_main) and background thread; added event-connection message macros (M_EV / PING_EV / S_EV / WR_EV) that operate on the explicit per-connection socket via the vl_socket_client_*2() APIs; added a separate lock for the event connection; and made __plugin_msg_base thread-local so the command and event threads cannot race between setting the base and the following M()/PING().

How did you verify/test it?

Built into the docker-sai-test-vpp image and run against the OCP sai_test suite; the command-vs-event deadlock at 32 ports no longer occurs and SaiSanityTest passes. The change is orthogonal to the recently-merged route-flow-counter context-table rework (#1894) and ipinip support (#1860); rebased on current master and verified both coexist.

Any platform specific information?

VPP platform only (vslib/vpp/). No change to the synchronous command path or SAI semantics; the event connection is additive.

Documentation

Development notes for this work are in .azure-pipelines/docker-sai-test-vpp/devdocs/ in #1950.

linux-foundation-easycla · 2026-06-23T16:06:19Z

The committers listed above are authorized under a signed CLA.

✅ login: nicholasching / name: Nicholas Ching (f3ab3f9)

mssonicbld · 2026-06-23T16:07:01Z

/azp run

azure-pipelines · 2026-06-23T16:07:14Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-06-23T17:59:15Z

/azp run

azure-pipelines · 2026-06-23T17:59:27Z

Azure Pipelines successfully started running 1 pipeline(s).

nicholasching · 2026-06-24T18:04:04Z

It appears the checks are failing due to an upstream issue which should be resolved when #4696 is merged into sonic-swss.

yue-fred-gao

Please trigger t1-lag-vpp sanity through PR in sonic-buildimage with this PR branch.

yue-fred-gao · 2026-06-29T14:13:28Z

+        if (now >= hard_deadline || now >= idle_deadline)       \
+            break;                                              \
+        if (scm && scm->socket_enable) {                        \
+            vl_socket_client_read (1);                          \


Miss the code to extend idle_deadline
_wr_rv = vl_socket_client_read (1);
if (_wr_rv == 0) {
idle_deadline = vat_time_now (vam) + 1.0;
if (idle_deadline > hard_deadline)
idle_deadline = hard_deadline;
}

Good catch, thank you Fred. Fixed in 145026f now.

yue-fred-gao · 2026-06-29T14:52:49Z

    uword *ptr;

-    sw_if_index = htonl(mp->sw_if_index);
+    sw_if_index = ntohl(mp->sw_if_index);


Good catch. Not sure how it works today or it never worked

Thank you, Fred!

yue-fred-gao · 2026-06-29T15:21:19Z

@@ -2972,21 +3160,23 @@ int vpp_get_interface_speed (const char *hwif_name, uint32_t *speed)

 int vpp_sync_for_events ()


Please add a comment that event handler running under EVENT_LOCK should never acquire VPP_LOCK to avoid cross lock.

Done! Added in 15a5298

mssonicbld · 2026-06-30T18:06:44Z

/azp run

azure-pipelines · 2026-06-30T18:06:57Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-07-02T17:54:33Z

/azp run

azure-pipelines · 2026-07-02T17:54:44Z

Azure Pipelines successfully started running 1 pipeline(s).

yue-fred-gao · 2026-07-02T20:54:02Z

The change looks good. If t1-lag-vpp sanity passes, I will approve the PR

nicholasching · 2026-07-02T20:55:19Z

Thanks @yue-fred-gao. Test running now; should be ready by tomorrow morning!

mssonicbld · 2026-07-03T14:24:56Z

/azp run

azure-pipelines · 2026-07-03T14:25:08Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-07-07T21:07:18Z

/azp run

azure-pipelines · 2026-07-07T21:07:31Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-07-09T14:56:28Z

/azp run

azure-pipelines · 2026-07-09T14:56:41Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-07-09T17:08:06Z

/azp run

azure-pipelines · 2026-07-09T17:08:23Z

Azure Pipelines successfully started running 1 pipeline(s).

nicholasching · 2026-07-09T19:00:58Z

/azp run

azure-pipelines · 2026-07-09T19:01:12Z

Azure Pipelines successfully started running 1 pipeline(s).

azure-pipelines · 2026-07-10T14:36:34Z

Azure Pipelines successfully started running 1 pipeline(s).

nicholasching · 2026-07-10T17:49:08Z

/azp run

azure-pipelines · 2026-07-10T17:49:41Z

Azure Pipelines successfully started running 1 pipeline(s).

nicholasching · 2026-07-10T19:27:01Z

/azp run

azure-pipelines · 2026-07-10T19:27:18Z

Azure Pipelines successfully started running 1 pipeline(s).

nicholasching · 2026-07-11T03:50:48Z

/azp run

azure-pipelines · 2026-07-11T03:50:59Z

Azure Pipelines successfully started running 1 pipeline(s).

nicholasching · 2026-07-13T14:01:28Z

/azp run

azure-pipelines · 2026-07-13T14:01:48Z

Azure Pipelines successfully started running 1 pipeline(s).

nicholasching · 2026-07-13T15:26:59Z

The t1-lag-vpp sanity test has passed on the associated sonic-buildimage PR: sonic-net/sonic-buildimage#28145

I have attached the results below:
Github Run: https://github.com/sonic-net/sonic-buildimage/runs/86538349140
Azure Build: https://dev.azure.com/mssonic/build/_build/results?buildId=1162024&view=logs&jobId=314bc92a-d36a-5eac-5c45-43e8f1288876

nicholasching · 2026-07-13T15:43:52Z

/azpw run

mssonicbld · 2026-07-13T15:43:54Z

⚠️ Notice: /azpw run only runs failed jobs now. If you want to trigger a whole pipline run, please rebase your branch or close and reopen the PR.
💡 Tip: You can also use /azpw retry to retry failed jobs directly.

Retrying failed(or canceled) jobs...

mssonicbld · 2026-07-13T15:43:57Z

Retrying failed(or canceled) stages in build 1163729:

✅Stage TestAsan:

Job vstest: retried.

✅Stage Test:

Job vstest: retried.
Job vstest: retried.

nicholasching · 2026-07-13T16:56:20Z

/azpw run

mssonicbld · 2026-07-13T16:56:22Z

⚠️ Notice: /azpw run only runs failed jobs now. If you want to trigger a whole pipline run, please rebase your branch or close and reopen the PR.
💡 Tip: You can also use /azpw retry to retry failed jobs directly.

Retrying failed(or canceled) jobs...

mssonicbld · 2026-07-13T16:56:25Z

Retrying failed(or canceled) stages in build 1163729:

✅Stage TestAsan:

Job vstest: retried.

✅Stage Test:

Job vstest: retried.
Job vstest: retried.

nicholasching · 2026-07-14T13:42:38Z

/azpw run

mssonicbld · 2026-07-14T13:42:41Z

⚠️ Notice: /azpw run only runs failed jobs now. If you want to trigger a whole pipline run, please rebase your branch or close and reopen the PR.
💡 Tip: You can also use /azpw retry to retry failed jobs directly.

Retrying failed(or canceled) jobs...

mssonicbld · 2026-07-14T13:42:44Z

Retrying failed(or canceled) stages in build 1163729:

✅Stage TestAsan:

Job vstest: retried.

✅Stage Test:

Job vstest: retried.
Job vstest: retried.

nicholasching · 2026-07-14T17:09:41Z

Hello @saiarcot895 @lolyu! The changes in this PR have previously been reviewed by @yue-fred-gao, received final approval from @dypet, and are now passing checks. Our team at Cisco currently does not have anyone with merge permissions. When you have a chance, could you please take a quick look and help us merge into master for SAIVPP UT feature?
https://github.com/yue-fred-gao/sonic-platform-vpp/blob/sai_vpp_unittest_hld/docs/HLD/saivpp-unittest-hld.md
Thanks!

Signed-off-by: Nicholas Ching <nicholaslching@gmail.com>

mssonicbld · 2026-07-15T21:50:04Z

/azp run

azure-pipelines · 2026-07-15T21:50:16Z

Azure Pipelines: Successfully started running 1 pipeline(s).

nicholasching · 2026-07-21T15:10:37Z

The t1-lag-vpp sanity test, applied to both #1951 and #1952 has passed on the associated sonic-buildimage PR: sonic-net/sonic-buildimage#28145

I have attached the results below:
Github Run: https://github.com/sonic-net/sonic-buildimage/runs/86968365539
Azure Build: https://dev.azure.com/mssonic/build/_build/results?buildId=1163779&view=results

nicholasching force-pushed the vpp_event_socket branch from f42fbc6 to f3ab3f9 Compare June 23, 2026 17:59

yue-fred-gao reviewed Jun 29, 2026

View reviewed changes

nicholasching closed this Jun 30, 2026

nicholasching reopened this Jun 30, 2026

nicholasching mentioned this pull request Jul 6, 2026

[vpp] SAIVPP unit-test harness (docker-sai-test-vpp) + devdocs #1950

Open

5 tasks

nicholasching closed this Jul 9, 2026

nicholasching reopened this Jul 9, 2026

nicholasching closed this Jul 9, 2026

dypet previously approved these changes Jul 14, 2026

View reviewed changes

fix: resolve WR macro declaration after master update

4389524

Signed-off-by: Nicholas Ching <nicholaslching@gmail.com>

nicholasching dismissed dypet’s stale review via 4389524 July 15, 2026 21:49

dypet approved these changes Jul 15, 2026

View reviewed changes

		@@ -2972,21 +3160,23 @@ int vpp_get_interface_speed (const char hwif_name, uint32_t speed)

		int vpp_sync_for_events ()

Uh oh!

Conversation

nicholasching commented Jun 23, 2026

Description of PR

Type of change

Approach

What is the motivation for this PR?

How did you do it?

How did you verify/test it?

Any platform specific information?

Documentation

Uh oh!

linux-foundation-easycla Bot commented Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mssonicbld commented Jun 23, 2026

Uh oh!

azure-pipelines Bot commented Jun 23, 2026

Uh oh!

mssonicbld commented Jun 23, 2026

Uh oh!

azure-pipelines Bot commented Jun 23, 2026

Uh oh!

nicholasching commented Jun 24, 2026

Uh oh!

yue-fred-gao left a comment

Choose a reason for hiding this comment

Uh oh!

yue-fred-gao Jun 29, 2026

Choose a reason for hiding this comment

Uh oh!

nicholasching Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

yue-fred-gao Jun 29, 2026

Choose a reason for hiding this comment

Uh oh!

nicholasching Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

yue-fred-gao Jun 29, 2026

Choose a reason for hiding this comment

Uh oh!

nicholasching Jul 2, 2026

Choose a reason for hiding this comment

Uh oh!

mssonicbld commented Jun 30, 2026

Uh oh!

azure-pipelines Bot commented Jun 30, 2026

Uh oh!

mssonicbld commented Jul 2, 2026

Uh oh!

azure-pipelines Bot commented Jul 2, 2026

Uh oh!

yue-fred-gao commented Jul 2, 2026

Uh oh!

nicholasching commented Jul 2, 2026

Uh oh!

mssonicbld commented Jul 3, 2026

Uh oh!

azure-pipelines Bot commented Jul 3, 2026

Uh oh!

mssonicbld commented Jul 7, 2026

Uh oh!

azure-pipelines Bot commented Jul 7, 2026

Uh oh!

mssonicbld commented Jul 9, 2026

Uh oh!

azure-pipelines Bot commented Jul 9, 2026

Uh oh!

mssonicbld commented Jul 9, 2026

Uh oh!

azure-pipelines Bot commented Jul 9, 2026

Uh oh!

nicholasching commented Jul 9, 2026

Uh oh!

azure-pipelines Bot commented Jul 9, 2026

Uh oh!

azure-pipelines Bot commented Jul 10, 2026

Uh oh!

linux-foundation-easycla Bot commented Jun 23, 2026 •

edited

Loading