[Celestica] Leh800bcls: Fix SRv6 decapsulation QoS mapping and Multi-NPU packet routing by gang-tao · Pull Request #1609 · facebook/fboss

gang-tao · 2026-09-07T07:40:14Z

Pre-submission checklist

[ ✓ ] I've ran the linters locally and fixed lint errors related to the files I modified in this PR. You can install the linters by running pip install -r requirements-dev.txt && pre-commit install
[ ✓ ] pre-commit run

[INFO] Stashing unstaged files to /work/home/gangtao/.cache/pre-commit/patch1788758710-2761128.
clang-format.............................................................Passed
shellcheck...........................................(no files to check)Skipped
shfmt................................................(no files to check)Skipped
trim trailing whitespace.................................................Passed
fix end of files.........................................................Passed
check yaml...........................................(no files to check)Skipped
check json...........................................(no files to check)Skipped
check for merge conflicts................................................Passed
ruff check...........................................(no files to check)Skipped
ruff format..........................................(no files to check)Skipped
Prevent sai_impl in fboss manifest.......................................Passed
[INFO] Restored changes from /work/home/gangtao/.cache/pre-commit/patch1788758710-2761128.

Summary

In AgentSrv6DecapTest switched test packets were previously injected via this->getSw()->sendPacketSwitchedAsync(...) without specifying the target SwitchID.
In multi-NPU split-agent environments (e.g. dual-NPU leh800bcls), this caused switched traffic to fallback and route exclusively to Switch 0. Consequently, test cases running on secondary NPUs (Switch 1) timed out with 0 packets received.
This PR replaces getSw()->sendPacketSwitchedAsync with this->sendPacketSwitchedAsync, which dynamically routes traffic to the active SwitchID under test.

Test Plan

Case passed on both NPU0 and NPU1:

NPU0

Removed "/etc/systemd/system/fboss_hw_agent_oss@1.service".
Cleaning up FBOSS HW Agent Service for index=0...
Cleaning up FBOSS HW Agent Service for index=1...
Running all tests took 0:01:39.649817 between 2026-10-14 11:19:53.266924 and 2026-10-14 11:21:32.916741
[ PASSED ] cold_boot.AgentSrv6DecapTest/0.VerifyDscpQueueMapping (17764 ms)
[ PASSED ] warm_boot.AgentSrv6DecapTest/0.VerifyDscpQueueMapping (11399 ms)
[ PASSED ] cold_boot.AgentSrv6DecapTest/1.VerifyDscpQueueMapping (17034 ms)
[ PASSED ] warm_boot.AgentSrv6DecapTest/1.VerifyDscpQueueMapping (11371 ms)
Summary:
   PASSED : 4
   FAILED : 0
   SKIPPED : 0
   TIMEOUT : 0

Test output stored at: hwtest_results_2026_Oct_14-11_21_32_AM.csv

NPU1

Running all tests took 0:01:40.154005 between 2026-10-14 11:21:48.291563 and 2026-10-14 11:23:28.445568
[ PASSED ] cold_boot.AgentSrv6DecapTest/0.VerifyDscpQueueMapping (18162 ms)
[ PASSED ] warm_boot.AgentSrv6DecapTest/0.VerifyDscpQueueMapping (11314 ms)
[ PASSED ] cold_boot.AgentSrv6DecapTest/1.VerifyDscpQueueMapping (17416 ms)
[ PASSED ] warm_boot.AgentSrv6DecapTest/1.VerifyDscpQueueMapping (10409 ms)
Summary:
   PASSED : 4
   FAILED : 0
   SKIPPED : 0
   TIMEOUT : 0

Test output stored at: hwtest_results_2026_Oct_14-11_23_28_AM.csv

meta-codesync · 2026-09-09T17:59:55Z

@jchallag has imported this pull request. If you are a Meta employee, you can view this in D119364585.

jchallag · 2026-09-09T22:56:30Z

+    auto config = this->getAgentEnsemble()->getCurrentConfig();
+    auto aclName = folly::to<std::string>("trap-", this->kV6RouteDstIp.str());
+    utility::delAcl(&config, aclName);
+    utility::delMatcher(&config, aclName);


I think you should use delCPUMatcher() to rightly remove the Action. delmatcher() seems to be a NO-op here

Hi @jchallag
Thank you for your incredibly sharp and precise review! You are 100% correct, utility::delMatcher only cleans up config->dataPlaneTrafficPolicy().
Now I added a new utility::delCPUMatcher() helper in AclTestUtils to target the CPU traffic policy and updated the VerifyDscpQueueMapping setup block to use utility::delCPUMatcher(&config, aclName).
Retested the case and all passed.

NPU0

Running all tests took 0:01:47.259575 between 2026-09-10 05:05:22.888876 and 2026-09-10 05:07:10.148451 [ PASSED ] cold_boot.AgentSrv6DecapTest/0.VerifyDscpQueueMapping (16878 ms) [ PASSED ] warm_boot.AgentSrv6DecapTest/0.VerifyDscpQueueMapping (10908 ms) [ PASSED ] cold_boot.AgentSrv6DecapTest/1.VerifyDscpQueueMapping (17789 ms) [ PASSED ] warm_boot.AgentSrv6DecapTest/1.VerifyDscpQueueMapping (10816 ms) Summary: PASSED : 4 FAILED : 0 SKIPPED : 0 TIMEOUT : 0 Test output stored at: hwtest_results_2026_Sep_10-05_07_10_AM.csv

NPU1

Running all tests took 0:01:47.729228 between 2026-09-10 05:07:38.133611 and 2026-09-10 05:09:25.862839 [ PASSED ] cold_boot.AgentSrv6DecapTest/0.VerifyDscpQueueMapping (16976 ms) [ PASSED ] warm_boot.AgentSrv6DecapTest/0.VerifyDscpQueueMapping (10943 ms) [ PASSED ] cold_boot.AgentSrv6DecapTest/1.VerifyDscpQueueMapping (17485 ms) [ PASSED ] warm_boot.AgentSrv6DecapTest/1.VerifyDscpQueueMapping (10812 ms) Summary: PASSED : 4 FAILED : 0 SKIPPED : 0 TIMEOUT : 0 Test output stored at: hwtest_results_2026_Sep_10-05_09_25_AM.csv

facebook-github-tools · 2026-09-10T02:13:18Z

@gang-tao has updated the pull request. You must reimport the pull request before landing.

msomasundaran · 2026-09-17T16:50:35Z

It is not clear why the acl failure is ladakh specific. Adding @jasmeetbagga to review that part.

gang-tao · 2026-09-18T01:56:43Z

It is not clear why the acl failure is ladakh specific. Adding @jasmeetbagga to review that part.
Hi @msomasundaran
I didn't see this passed on other TH6 platform. let me know if I am wrong. Thanks.

jchallag · 2026-10-13T06:30:07Z

Hi @gang-tao , it looks like this issue is fixed on TH platforms in a different way and we may not need this fix. We will still need multi-NPU part of this PR. can you please rebase and update?

…i-NPU in AgentSrv6DecapTest

facebook-github-tools · 2026-10-13T12:53:06Z

@gang-tao has updated the pull request. You must reimport the pull request before landing.

gang-tao · 2026-10-13T12:54:37Z

Hi @gang-tao , it looks like this issue is fixed on TH platforms in a different way and we may not need this fix. We will still need multi-NPU part of this PR. can you please rebase and update?

@jchallag updated as you suggested. thanks

jchallag · 2026-10-13T17:11:27Z

@gang-tao can you update the summary, title and testplan as well? hope it is verified on the latest rebase.

gang-tao · 2026-10-13T22:57:47Z

@gang-tao can you update the summary, title and testplan as well? hope it is verified on the latest rebase.
@jchallag

After rebasing and making the changes, I ran into two regression issues:
1 'Update : update mySids application to HW failed'
2 Check failed: intfID == 0 (10 vs. 0) Portless interface 10 on vlan 4094 has no SwitchInfo::loopbackIntfId, but this platform has 2 l3 switches

After the fix, I hit the original issue again.

I wanted to check if this was already tested and verified on a single NPU? Could you let me know which single NPU platform this was fixed on? I'll double-check it on another platform (TH6) as well

jchallag · 2026-10-14T05:34:26Z

@gang-tao this was tested on icecube platform. commit hash 0ee51b9 .

gang-tao · 2026-10-14T08:36:54Z

@gang-tao this was tested on icecube platform. commit hash 0ee51b9 .

@jchallag
I found the root cause. The previous fix for the first regression wasn't complete, which caused the TunnelId on MySidEntry to not be set. Now that I've applied the fix, it works. I have updated the summary and test plan.
Thanks!

meta-codesync · 2026-10-14T20:27:23Z

@jchallag merged this pull request in 9a797dd.

gang-tao requested a review from a team as a code owner September 7, 2026 07:40

meta-cla Bot added the CLA Signed label Sep 7, 2026

github-actions Bot added the celestica label Sep 7, 2026

jchallag reviewed Sep 9, 2026

View reviewed changes

gang-tao force-pushed the leh800bcls_agent_fix25 branch from 6299016 to 99ee053 Compare September 10, 2026 02:13

jchallag reviewed Oct 13, 2026

View reviewed changes

Comment thread fboss/agent/test/agent_hw_tests/AgentSrv6DecapTests.cpp

[Celestica] Leh800bcls: Route switched packets with SwitchID for Mult…

59e70ad

…i-NPU in AgentSrv6DecapTest

gang-tao force-pushed the leh800bcls_agent_fix25 branch from 99ee053 to 59e70ad Compare October 13, 2026 12:53

github-actions Bot removed the celestica label Oct 14, 2026

meta-codesync Bot closed this in 9a797dd Oct 14, 2026

meta-codesync Bot added the Merged label Oct 14, 2026

Conversation

gang-tao commented Sep 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test Plan

Uh oh!

meta-codesync Bot commented Sep 9, 2026

Uh oh!

jchallag Sep 9, 2026

Choose a reason for hiding this comment

Uh oh!

gang-tao Sep 10, 2026

Choose a reason for hiding this comment

Uh oh!

facebook-github-tools Bot commented Sep 10, 2026

Uh oh!

msomasundaran commented Sep 17, 2026

Uh oh!

gang-tao commented Sep 18, 2026

Uh oh!

jchallag commented Oct 13, 2026

Uh oh!

Uh oh!

facebook-github-tools Bot commented Oct 13, 2026

Uh oh!

gang-tao commented Oct 13, 2026

Uh oh!

jchallag commented Oct 13, 2026

Uh oh!

gang-tao commented Oct 13, 2026

Uh oh!

jchallag commented Oct 14, 2026

Uh oh!

gang-tao commented Oct 14, 2026

Uh oh!

meta-codesync Bot commented Oct 14, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

gang-tao commented Sep 7, 2026 •

edited

Loading