[Hw PFCWD feature] Skeleton hw pfcwd by abhishek-nexthop · Pull Request #4403 · sonic-net/sonic-swss

abhishek-nexthop · 2026-03-28T18:01:51Z

What I did
Added hardware-based PFC watchdog support with SKU-specific enablement. Created PfcWdHwOrch skeleton class that inherits from PfcWdBaseOrch, refactored shared functionality into the base class, and implemented runtime SKU detection to automatically choose between hardware and software watchdog implementations.

Why I did it
Modern switch ASICs support hardware-based PFC deadlock detection/recovery which is more efficient than software-based approaches. Need infrastructure to:
• Enable hardware watchdog only on supported SKUs
• Share common code between software and hardware implementations
• Gracefully fall back to software watchdog on unsupported platforms

How I verified it

Code compiles successfully with new files added to Makefile
SKU detection logic correctly identifies supported hardware:
• Check logs: grep "PFC hardware watchdog" /var/log/syslog
• Supported SKU shows: "HWSKU 'nh-4010' supports PFC hardware watchdog"
• Unsupported SKU shows: "does not support PFC hardware watchdog"
Correct watchdog instance created based on SKU and DLR init flag

mssonicbld · 2026-03-28T18:01:59Z

/azp run

azure-pipelines · 2026-03-28T18:02:10Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-28T18:05:31Z

/azp run

azure-pipelines · 2026-03-28T18:05:42Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-28T18:11:33Z

/azp run

azure-pipelines · 2026-03-28T18:11:43Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-28T18:44:49Z

/azp run

azure-pipelines · 2026-03-28T18:45:00Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-29T02:20:00Z

/azp run

azure-pipelines · 2026-03-29T02:20:11Z

Azure Pipelines successfully started running 1 pipeline(s).

mssonicbld · 2026-03-29T02:56:14Z

/azp run

azure-pipelines · 2026-03-29T02:56:24Z

Azure Pipelines successfully started running 1 pipeline(s).

Signed-off-by: Abhishek <abhishek@nexthop.ai>

… functions. Signed-off-by: Abhishek <abhishek@nexthop.ai>

Signed-off-by: Abhishek <abhishek@nexthop.ai>

mssonicbld · 2026-05-27T06:38:08Z

/azp run

azure-pipelines · 2026-05-27T06:38:18Z

Azure Pipelines successfully started running 1 pipeline(s).

vmittal-msft · 2026-06-20T00:45:42Z

@abhishek-nexthop please help resolve conflicts

Copilot

Pull request overview

This PR introduces an initial framework for hardware-based PFC watchdog (PFCWD) support in orchagent. It refactors existing PFCWD logic into a new PfcWdBaseOrch, moves the software watchdog implementation into PfcWdSwOrch, adds a skeleton PfcWdHwOrch, and wires runtime SKU detection to select HW vs SW watchdog at startup.

Changes:

Refactor PFC watchdog common logic into PfcWdBaseOrch and move software watchdog into new pfcwdsworch.* files.
Add PfcWdHwOrch skeleton and build plumbing (Makefiles, includes) for HW watchdog compilation.
Add HWSKU read + regex-based SKU support check in SwitchOrch, and select HW vs SW watchdog in OrchDaemon::init().

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tests/mock_tests/mock_orchagent_main.h	Includes SW PFCWD header for mock test builds.
tests/mock_tests/Makefile.am	Adds new PFCWD SW/HW source files to mock test build.
orchagent/switchorch.h	Exposes HWSKU getter and HW PFCWD SKU support query.
orchagent/switchorch.cpp	Implements HWSKU read from CONFIG_DB and SKU regex match helper.
orchagent/pfcwdsworch.h	Introduces `PfcWdSwOrch` (software watchdog) as a separate class.
orchagent/pfcwdsworch.cpp	Contains the migrated software watchdog implementation.
orchagent/pfcwdorch.h	Introduces `PfcWdBaseOrch` to host shared PFCWD logic.
orchagent/pfcwdorch.cpp	Implements the new base class and moves shared helpers into it.
orchagent/pfcwdhworch.h	Adds HW PFCWD orchestrator class declaration (skeleton).
orchagent/pfcwdhworch.cpp	Adds HW PFCWD orchestrator stub implementation (TODOs).
orchagent/orchdaemon.h	Wires in new SW/HW PFCWD headers for orchagent build.
orchagent/orchdaemon.cpp	Chooses HW vs SW watchdog based on SKU + DLR INIT enablement.
orchagent/Makefile.am	Adds new SW/HW PFCWD source files to orchagent build.

Comments suppressed due to low confidence (1)

orchagent/orchdaemon.h:35

Both pfcwdorch.h and countercheckorch.h define PFC_WD_TC_MAX. Since orchdaemon.h includes both, this can trigger a macro redefinition warning (the build uses -Werror/-Wextra) and break the build depending on compiler settings. Consider consolidating the constant in one header; as a minimal mitigation here, undefine the macro before including countercheckorch.h.

#include "pfcwdorch.h"
#include "pfcwdsworch.h"
#include "pfcwdhworch.h"
#include "switchorch.h"
#include "crmorch.h"
#include "vrforch.h"
#include "vxlanorch.h"
#include "vnetorch.h"
#include "countercheckorch.h"

+task_process_status PfcWdHwOrch::createEntry(const string& key, const vector<FieldValueTuple>& data)
+{
+    SWSS_LOG_ENTER();
+    // TODO: Implementation
+    return task_process_status::task_success;
+}


+task_process_status PfcWdHwOrch::deleteEntry(const string& key)
+{
+    SWSS_LOG_ENTER();
+    // TODO: Implementation
+    return task_process_status::task_success;
+}


+    // TODO: Implementation
+    PfcWdQueueStats stats;
+    memset(&stats, 0, sizeof(PfcWdQueueStats));
+    return stats;


+        // Check if SKU supports PFC hardware watchdog
+        bool pfcHwWdSupportedSku = gSwitchOrch->isHwPfcWdSupportedSku();
+
+        if (pfcHwWdSupportedSku)
+        {
+            SWSS_LOG_NOTICE("HWSKU '%s' supports PFC hardware watchdog", gSwitchOrch->getHwSku().c_str());
+        }
+        else
+        {
+            SWSS_LOG_NOTICE("HWSKU '%s' does not support PFC hardware watchdog", gSwitchOrch->getHwSku().c_str());
+        }
+
+        if (pfcDlrInit && pfcHwWdSupportedSku)
+        {
+            SWSS_LOG_NOTICE("Starting hardware-based pfc watchdog");
+            m_orchList.push_back(new PfcWdHwOrch(
+                        m_configDb,
+                        pfc_wd_tables,
+                        portStatIds,
+                        queueStatIds,
+                        queueAttrIds));
+        }


+    // Read HWSKU from CONFIG_DB
+    DBConnector configDb("CONFIG_DB", 0);
+    Table deviceMetadataTable(&configDb, CFG_DEVICE_METADATA_TABLE_NAME);
+    deviceMetadataTable.hget("localhost", "hwsku", m_hwSku);
+    SWSS_LOG_NOTICE("HWSKU: %s", m_hwSku.c_str());
+


prsunny · 2026-06-20T01:37:00Z

What is this PR about when there are other hw pfcwd PRs.? Folks, if this is experimental, please keep this as "draft".

Signed-off-by: Abhishek <abhishek@nexthop.ai>

mssonicbld · 2026-06-23T10:57:44Z

/azp run

azure-pipelines · 2026-06-23T10:57:56Z

Azure Pipelines successfully started running 1 pipeline(s).

abhishek-nexthop · 2026-06-23T11:14:59Z

What is this PR about when there are other hw pfcwd PRs.? Folks, if this is experimental, please keep this as "draft".

@prsunny We have 3 PRs for pfcwd h/w orch.

Refactor the existing code so it can be modularized for sw and hw classes,
Boilerplate code for pfcwd hw orch(this PR)
Implementation of pfcwd hw orch.
There are other PR which are related to cli/sonic-mgmt/sonic-buildimage Please refer Feature/ PFCWD Hardware Recovery HLD SONiC#2159 for more details on PRs.

mssonicbld · 2026-06-24T11:20:50Z

Hi, there are workflow run(s) waiting for approval, you may be first-time contributor. I will notify maintainers to help approve once PR is approved. Thanks!

_{---Powered by SONiC BuildBot}

vmittal-msft · 2026-07-10T16:13:50Z

@abhishek-nexthop please take care of copilot comments. once depending PRs are in. we can do further review.

abhishek-nexthop requested review from kperumalbfn and prsunny as code owners March 28, 2026 18:01

abhishek-nexthop force-pushed the skeleton-hw-pfcwd branch from 28da074 to dd3fc1c Compare March 28, 2026 18:05

abhishek-nexthop force-pushed the skeleton-hw-pfcwd branch from dd3fc1c to d1e15c4 Compare March 28, 2026 18:11

abhishek-nexthop changed the title ~~Skeleton hw pfcwd~~ [Hw PFCWD feature] Skeleton hw pfcwd Mar 28, 2026

abhishek-nexthop force-pushed the skeleton-hw-pfcwd branch from d1e15c4 to 4c5ce7a Compare March 28, 2026 18:44

abhishek-nexthop force-pushed the skeleton-hw-pfcwd branch from 4c5ce7a to 9d92949 Compare March 29, 2026 02:19

abhishek-nexthop force-pushed the skeleton-hw-pfcwd branch from 9d92949 to cfbfb45 Compare March 29, 2026 02:56

pinky-nexthop mentioned this pull request Apr 23, 2026

Feature/ PFCWD Hardware Recovery HLD sonic-net/SONiC#2159

Open

abhishek-nexthop added 3 commits May 27, 2026 06:34

Refactor pfcwd orch code.

384f50e

Signed-off-by: Abhishek <abhishek@nexthop.ai>

Add helper function counterIdstoStr and comment stating common helper…

8d08bfa

… functions. Signed-off-by: Abhishek <abhishek@nexthop.ai>

Skeleton code for hw based PFCWD.

03e418b

Signed-off-by: Abhishek <abhishek@nexthop.ai>

abhishek-nexthop force-pushed the skeleton-hw-pfcwd branch from cfbfb45 to 03e418b Compare May 27, 2026 06:38

vmittal-msft requested a review from Copilot June 20, 2026 00:45

Copilot started reviewing on behalf of vmittal-msft June 20, 2026 00:45 View session

Copilot AI reviewed Jun 20, 2026

View reviewed changes

abhishek-nexthop added 2 commits June 23, 2026 10:49

Merge branch 'master' into refactor-pfcwd

5b9248b

Signed-off-by: Abhishek <abhishek@nexthop.ai>

Merge branch 'refactor-pfcwd' into skeleton-hw-pfcwd

9535ba0

Signed-off-by: Abhishek <abhishek@nexthop.ai>

Uh oh!

Conversation

abhishek-nexthop commented Mar 28, 2026

Uh oh!

mssonicbld commented Mar 28, 2026

Uh oh!

azure-pipelines Bot commented Mar 28, 2026

Uh oh!

mssonicbld commented Mar 28, 2026

Uh oh!

azure-pipelines Bot commented Mar 28, 2026

Uh oh!

mssonicbld commented Mar 28, 2026

Uh oh!

azure-pipelines Bot commented Mar 28, 2026

Uh oh!

mssonicbld commented Mar 28, 2026

Uh oh!

azure-pipelines Bot commented Mar 28, 2026

Uh oh!

mssonicbld commented Mar 29, 2026

Uh oh!

azure-pipelines Bot commented Mar 29, 2026

Uh oh!

mssonicbld commented Mar 29, 2026

Uh oh!

azure-pipelines Bot commented Mar 29, 2026

Uh oh!

mssonicbld commented May 27, 2026

Uh oh!

azure-pipelines Bot commented May 27, 2026

Uh oh!

vmittal-msft commented Jun 20, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

prsunny commented Jun 20, 2026

Uh oh!

mssonicbld commented Jun 23, 2026

Uh oh!

azure-pipelines Bot commented Jun 23, 2026

Uh oh!

abhishek-nexthop commented Jun 23, 2026

Uh oh!

mssonicbld commented Jun 24, 2026

Uh oh!

vmittal-msft commented Jul 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants